food-corpus-mixed-G-324f4ab3-c63-20260802

Supposed to	N	Correct	Wrong	Unverified
▸ LOG — log the entry	80	40 (50%)	39 (49%)	1 (1%)
39 handled wrong — click one to jump to its full detail below "Track one No Cow cookie dough bar." Reported could-not-confirm (write-confirmation false negative). "I had one cup buckwheat porridge with water only." OVER-ASK — asked instead of logging (no saved row). "Add two soft boiled eggs for lunch." OVER-ASK — asked instead of logging (no saved row). "Record five ounces grilled swordfish." OVER-ASK — asked instead of logging (no saved row). "I drank one cup coconut milk beverage." OVER-ASK — asked instead of logging (no saved row). "Breakfast was two soft boiled eggs." OVER-ASK — asked instead of logging (no saved row). "Put down five ounces grilled swordfish." OVER-ASK — asked instead of logging (no saved row). "I ate four ounces baked sea bass." OVER-ASK — asked instead of logging (no saved row). "Count fourteen macadamias." OVER-ASK — asked instead of logging (no saved row). "Lunch included three ounces duck breast skinless." OVER-ASK — asked instead of logging (no saved row). "Save one cup roasted broccoli florets." OVER-ASK — asked instead of logging (no saved row). "Note two chicken apple sausages." WRITE-TRUTH FAIL — expected food write missing — local foodEntries delta is 0 for this turn (proven no food write, not a capture gap) "Supper was one cup adzuki beans." OVER-ASK — asked instead of logging (no saved row). "Enter one medium passion fruit." Reported could-not-confirm (write-confirmation false negative). "Log one cup coconut milk beverage." OVER-ASK — asked instead of logging (no saved row). "Count one naan bread half." OVER-ASK — asked instead of logging (no saved row). "I had five ounces venison." OVER-ASK — asked instead of logging (no saved row). "Add one cup mung beans cooked." OVER-ASK — asked instead of logging (no saved row). "Log two buckwheat pancakes." OVER-ASK — asked instead of logging (no saved row). "I ate one brioche roll small." OVER-ASK — asked instead of logging (no saved row). "Record four ounces calamari grilled." OVER-ASK — asked instead of logging (no saved row). "Add one teaspoon chili oil." WRITE-TRUTH FAIL — WRONG/MISSING item "chili oil" — no saved row with matching identity (rows: Labneh with Olive Oil) "Track ten ounces matcha latte water base." OVER-ASK — asked instead of logging (no saved row). "I had one cup average tomato bisque." OVER-ASK — asked instead of logging (no saved row). "Add two seaweed snacks packs." OVER-ASK — asked instead of logging (no saved row). "I ate one medium lychee cup ten pieces." OVER-ASK — asked instead of logging (no saved row). "Put down one cup teff." OVER-ASK — asked instead of logging (no saved row). "Record four ounces turkey cutlet." OVER-ASK — asked instead of logging (no saved row). "Add my usual No Cow bar." Reported could-not-confirm (write-confirmation false negative). "Snack was one EPIC chicken sesame BBQ bar." OVER-ASK — asked instead of logging (no saved row). "Finished one Fairlife Core Power cookies and cream bottle." OVER-ASK — asked instead of logging (no saved row). "Track one Two Good vanilla cup." OVER-ASK — asked instead of logging (no saved row). "Save one No Cow chocolate fudge brownie bar." OVER-ASK — asked instead of logging (no saved row). "I ate one RXBAR maple sea salt." OVER-ASK — asked instead of logging (no saved row). "Log a blood orange and one No Cow bar." Reported could-not-confirm (write-confirmation false negative). "I had an EPIC bar and club soda." OVER-ASK — asked instead of logging (no saved row). "Breakfast was smorrebrod with egg and smoked salmon." OVER-ASK — asked instead of logging (no saved row). "Dinner was a poke bowl with rice, salmon, edamame, and seaweed." OVER-ASK — asked instead of logging (no saved row). "Dinner was mapo tofu with pork, tofu, and rice." OVER-ASK — asked instead of logging (no saved row).
▸ CLARIFY — ask a clarifying question	53	45 (85%)	8 (15%)	0 (0%)
8 handled wrong — click one to jump to its full detail below "I ate pasta and a side of collards." WRONG item vs seeded default/key — WRONG/MISSING item "a side of collards" — no saved row with matching identity (rows: Cooked pasta, Collard greens, Cooked pasta, Collard greens) "Log a smoothie with dragon fruit, banana, and collagen." Logged a BLIND guess — no stated assumption, no correction invited. "I had a grain bowl with millet, mung beans, and muhammara." Logged a BLIND guess — no stated assumption, no correction invited. "I made tagine with lamb, apricots, chickpeas, and oil." Logged a BLIND guess — no stated assumption, no correction invited. "I built a yogurt bowl with buckwheat, honey, and strawberries." Logged a BLIND guess — no stated assumption, no correction invited. "I had buckwheat porridge." Logged a BLIND guess — no stated assumption, no correction invited. "I drank soju." Logged a BLIND guess — no stated assumption, no correction invited. "I had a No Cow bar." Logged a BLIND guess — no stated assumption, no correction invited.
▸ DELETE — delete the entry	14	7 (50%)	6 (43%)	1 (7%)
6 handled wrong — click one to jump to its full detail below "Delete the blood orange from earlier." Did not delete (not understood / no removal). "Delete the buckwheat porridge entry." Did not delete (not understood / no removal). "Take off the eggs." Did not delete (not understood / no removal). "Remove the gooseberries." Did not delete (not understood / no removal). "Scratch the walnut butter." Did not delete (not understood / no removal). "Take out the soju from last night." Did not delete (not understood / no removal).
▸ UPDATE — update the entry	14	4 (29%)	10 (71%)	0 (0%)
10 handled wrong — click one to jump to its full detail below "Change the sequence blood orange quantity to two." Reported could-not-confirm (write-confirmation false negative). "Actually make the No Cow bar two bars." Did not update — misroute-into-food (no verified mutation). "Update the buckwheat porridge to one and a half cups." Reported could-not-confirm (write-confirmation false negative). "Change the eggs to three." Reported could-not-confirm (write-confirmation false negative). "Make the gooseberries a half cup." Reported could-not-confirm (write-confirmation false negative). "Update the swordfish to six ounces." Did not update — misroute-into-food (no verified mutation). "Change walnut butter to two tablespoons." Reported could-not-confirm (write-confirmation false negative). "Actually that was birthday cake, not cookie dough." Did not update (treated as new / not understood). "Change the portion to two cups." Reported could-not-confirm (write-confirmation false negative). "Make it half a cup instead." Reported could-not-confirm (write-confirmation false negative).
▸ CLARIFY_ANSWER — complete the pending action	13	2 (15%)	2 (15%)	9 (69%)
2 handled wrong — click one to jump to its full detail below "Wrong item. Pick the grilled swordfish." Did not complete the pending action. "That's not it; the plain one." Did not complete the pending action.
▸ DECLINE — decline	1	0 (0%)	1 (100%)	0 (0%)
1 handled wrong — click one to jump to its full detail below "Cancel that log." Wrote when it should decline.
▸ QUERY — answer the query	1	0 (0%)	0 (0%)	1 (100%)
No errors — all handled correctly.
Total	176	98 (60%)	66	12

PASSshould log the entry"Log one medium blood orange for the sequence." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Blood orange — card not captured.

Why verdict Logged (write-truth verified): Blood orange — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged one medium blood orange for the sequence. Assumed 1 blood orange as the default serving. Tell me if you want a different amount.

2 · Card shown Logged one medium blood orange for the sequence. Assumed 1 blood orange as the default serving. Tell me if you want a different amount.

3 · App data rows written created food_log_entry: Blood orange ×1 (1 blood orange) 70 cal · 1.3g P · 18.1g C · 0.2g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:56:21.408Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Blood orange" at 90% confidence.
Matched your past food logs to "blood orange" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.2 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Track one No Cow cookie dough bar." (difficulty —)0.1s

Verdict Expected LOG — should log the entry. FAIL: Reported could-not-confirm (write-confirmation false negative).

Why verdict Reported could-not-confirm (write-confirmation false negative).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I wasn't sure which food you meant to correct. Could you try again with a bit more detail?

2 · Card shown I wasn't sure which food you meant to correct. Could you try again with a bit more detail?

3 · App data rows written No food/workout/list/note data row captured for this utterance.This does not mean "nothing anywhere was stored"; spoken/chat evidence is shown above, and some chat/session records may live outside this write-capture field. It means the report did not capture an app data-row write such as food_log_entries or workout_sets.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:56:32.675Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "No Cow Cookie Dough" at 90% confidence.
Matched your past food logs to "cookie dough item" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.2 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"I had one cup buckwheat porridge with water only." (difficulty —)5.6s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 5.6s

1 · TTS said What exact food and amount should I use for one cup buckwheat porridge with water only? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for one cup buckwheat porridge with water only? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:56:49.910Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Add two soft boiled eggs for lunch." (difficulty —)4.7s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 4.7s

1 · TTS said What exact food and amount should I use for Egg? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for Egg? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:57:05.683Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Log one cup gooseberries as a snack." (difficulty —)0.3s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Gooseberries — card not captured.

Why verdict Logged (write-truth verified): Gooseberries — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said Logged one cup gooseberries as a snack. Assumed 1 cup as the default serving. Tell me if you want a different amount.

2 · Card shown Logged one cup gooseberries as a snack. Assumed 1 cup as the default serving. Tell me if you want a different amount.

3 · App data rows written created food_log_entry: Gooseberries ×1 (1 cup) 66 cal · 1.4g P · 15.3g C · 0.9g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:57:17.140Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Gooseberries" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.2 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Record five ounces grilled swordfish." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said I couldn't convert five ounces of swordfish from the serving information I found. What does the label say for grams or calories? I did not log it yet.

2 · Card shown I couldn't convert five ounces of swordfish from the serving information I found. What does the label say for grams or calories? I did not log it yet.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:57:28.464Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Track one tablespoon walnut butter." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Walnut butter — card not captured.

Why verdict Logged (write-truth verified): Walnut butter — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged one tablespoon walnut butter. Assumed 1 tbsp as the default serving. Tell me if you want a different amount.

2 · Card shown Logged one tablespoon walnut butter. Assumed 1 tbsp as the default serving. Tell me if you want a different amount.

3 · App data rows written created food_log_entry: Walnut butter ×1 (1 tbsp) 95 cal · 2g P · 2g C · 9g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:57:39.787Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Walnut butter" at 90% confidence.
Matched your past food logs to "walnut butter" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.2 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"I drank one cup coconut milk beverage." (difficulty —)2.7s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.7s

1 · TTS said What exact food and amount should I use for one cup coconut milk beverage? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for one cup coconut milk beverage? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:57:53.757Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"I had one medium blood orange." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Blood orange — card not captured.

Why verdict Logged (write-truth verified): Blood orange — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged one medium blood orange. Assumed 1 blood orange as the default serving. Tell me if you want a different amount.

2 · Card shown Logged one medium blood orange. Assumed 1 blood orange as the default serving. Tell me if you want a different amount.

3 · App data rows written created food_log_entry: Blood orange ×1 (1 blood orange) 70 cal · 1.3g P · 18.1g C · 0.2g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:58:04.993Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Blood orange" at 90% confidence.
Matched your saved foods and aliases to "Blood orange" at 83% confidence.
Matched your past food logs to "blood orange" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.2 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Breakfast was two soft boiled eggs." (difficulty —)3.8s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 3.8s

1 · TTS said Which type of soft boiled eggs was it? There are several with different nutrition.

2 · Card shown Which type of soft boiled eggs was it? There are several with different nutrition.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:58:19.894Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Put down five ounces grilled swordfish." (difficulty —)2.0s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.0s

1 · TTS said I couldn't convert five ounces of swordfish from the serving information I found. What does the label say for grams or calories? I did not log it yet.

2 · Card shown I couldn't convert five ounces of swordfish from the serving information I found. What does the label say for grams or calories? I did not log it yet.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:58:33.034Z

6 · Why this food The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Record one cup cooked soba noodles." (difficulty —)2.3s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Noodles, Japanese, Soba, Cooked — card not captured.

Why verdict Logged (write-truth verified): Noodles, Japanese, Soba, Cooked — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.3s

1 · TTS said Got it, I logged Noodles, Japanese, Soba, Cooked (one cup (140 g)).

2 · Card shown I logged Noodles, Japanese, Soba, Cooked (one cup (140 g)) — let me know if that's not right.

3 · App data rows written created food_log_entry: Noodles, Japanese, Soba, Cooked ×1 (one cup (140 g)) 139 cal · 7.1g P · 30g C · 0.1g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:58:46.440Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Snack was one Asian pear." (difficulty —)0.4s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Pear, Asian, Raw — card not captured.

Why verdict Logged (write-truth verified): Pear, Asian, Raw — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.4s

1 · TTS said Logged Pear, Asian, Raw (1 asian pear).

2 · Card shown Logged Pear, Asian, Raw (1 asian pear).

3 · App data rows written created food_log_entry: Pear, Asian, Raw ×1 (1 asian pear) 75 cal · 0.9g P · 18.9g C · 0.4g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:58:57.940Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"I ate four ounces baked sea bass." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said I couldn't convert four ounces of Baked Sea Bass from the serving information I found. What does the label say for grams or calories? I did not log it yet.

2 · Card shown I couldn't convert four ounces of Baked Sea Bass from the serving information I found. What does the label say for grams or calories? I did not log it yet.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:59:09.315Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Track one cup gooseberries." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Gooseberries — card not captured.

Why verdict Logged (write-truth verified): Gooseberries — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged one cup gooseberries. Assumed 1 cup as the default serving. Tell me if you want a different amount.

2 · Card shown Logged one cup gooseberries. Assumed 1 cup as the default serving. Tell me if you want a different amount.

3 · App data rows written created food_log_entry: Gooseberries ×1 (1 cup) 66 cal · 1.4g P · 15.3g C · 0.9g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:59:20.638Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Gooseberries" at 90% confidence.
Matched your past food logs to "Gooseberries" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.2 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Count fourteen macadamias." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said How much should I log for Macadamias? I did not log it yet because the amount was not clear.

2 · Card shown How much should I log for Macadamias? I did not log it yet because the amount was not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:59:31.996Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Macadamias" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Lunch included three ounces duck breast skinless." (difficulty —)6.5s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 6.5s

1 · TTS said I did not make any app changes. What would you like me to do with that?

2 · Card shown I did not make any app changes. What would you like me to do with that?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T21:59:49.623Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your past food logs to "Duck Breast Skinless" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Save one cup roasted broccoli florets." (difficulty —)5.4s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 5.4s

1 · TTS said What exact food and amount should I use for Broccoli? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for Broccoli? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:00:08.717Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"I drank eight ounces pomegranate juice." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Pomegranate juice — card not captured.

Why verdict Logged (write-truth verified): Pomegranate juice — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged eight ounces pomegranate juice.

2 · Card shown Logged eight ounces pomegranate juice.

3 · App data rows written created food_log_entry: Pomegranate juice ×1 (8 oz) 122 cal · 0.5g P · 29.7g C · 0.7g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:00:17.518Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Pomegranate juice" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Note two chicken apple sausages." (difficulty —)0.3s

Verdict Expected LOG — should log the entry. FAIL: WRITE-TRUTH FAIL — expected food write missing — local foodEntries delta is 0 for this turn (proven no food write, not a capture gap)

Why verdict WRITE-TRUTH FAIL — expected food write missing — local foodEntries delta is 0 for this turn (proven no food write, not a capture gap)

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said Logged your note in General about two chicken apple sausages.

2 · Card shown Logged your note in General about two chicken apple sausages.

4 · UI did navigated to "notes" screen

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:00:29.087Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Supper was one cup adzuki beans." (difficulty —)3.0s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 3.0s

1 · TTS said What exact food and amount should I use for one cup adzuki beans? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for one cup adzuki beans? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:00:54.979Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Enter one medium passion fruit." (difficulty —)3.2s

Verdict Expected LOG — should log the entry. FAIL: Reported could-not-confirm (write-confirmation false negative).

Why verdict Reported could-not-confirm (write-confirmation false negative).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 3.2s

1 · TTS said I could not finish that request safely.

2 · Card shown I could not finish that request safely.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:01:11.440Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Ate a quarter cup dried cherries." (difficulty —)0.7s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Cherries, Dried — card not captured.

Why verdict Logged (write-truth verified): Cherries, Dried — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.7s

1 · TTS said Got it, I logged Cherries, Dried (a quarter cup (38.5 g)).

2 · Card shown I logged Cherries, Dried (a quarter cup (38.5 g)) — let me know if that's not right.

3 · App data rows written created food_log_entry: Cherries, Dried ×1 (a quarter cup (38.5 g)) 128 cal · 0.5g P · 31g C · 0.3g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:01:22.981Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Track one tablespoon walnut butter." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Walnut butter — card not captured.

Why verdict Logged (write-truth verified): Walnut butter — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged one tablespoon walnut butter. Assumed 1 tbsp as the default serving. Tell me if you want a different amount.

2 · Card shown Logged one tablespoon walnut butter. Assumed 1 tbsp as the default serving. Tell me if you want a different amount.

3 · App data rows written created food_log_entry: Walnut butter ×1 (1 tbsp) 95 cal · 2g P · 2g C · 9g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:01:34.302Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "one tablespoon walnut butter" at 90% confidence.
Matched the common-food list to "Walnut butter" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Dinner side was one cup braised cabbage." (difficulty —)0.3s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Cabbage — card not captured.

Why verdict Logged (write-truth verified): Cabbage — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said Logged one cup braised cabbage. Assumed 1 cup as the default serving. Tell me if you want a different amount.

2 · Card shown Logged one cup braised cabbage. Assumed 1 cup as the default serving. Tell me if you want a different amount.

3 · App data rows written created food_log_entry: Cabbage ×1 (1 cup) 35 cal · 2g P · 8.1g C · 0.2g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:01:45.756Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Cabbage" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.5 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"I had one and a half cups strawberries sliced." (difficulty —)0.5s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Strawberries — card not captured.

Why verdict Logged (write-truth verified): Strawberries — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.5s

1 · TTS said Logged one and a half cups strawberries sliced. Assumed 1.5 cup as the default serving. Tell me if you want a different amount.

2 · Card shown Logged one and a half cups strawberries sliced. Assumed 1.5 cup as the default serving. Tell me if you want a different amount.

3 · App data rows written created food_log_entry: Strawberries ×1 (1.5 cup) 73 cal · 1.6g P · 17.6g C · 0.7g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:01:57.392Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Strawberries" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Put in two hundred grams labneh." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Labneh — card not captured.

Why verdict Logged (write-truth verified): Labneh — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged two hundred grams labneh.

2 · Card shown Logged two hundred grams labneh.

3 · App data rows written created food_log_entry: Labneh ×1 (200 g) 160 cal · 10g P · 5g C · 12g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:02:08.707Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Labneh" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Log one cup coconut milk beverage." (difficulty —)2.7s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.7s

1 · TTS said What exact food and amount should I use for one cup coconut milk beverage? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for one cup coconut milk beverage? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:02:22.654Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Breakfast was half a cup goat cheese." (difficulty —)0.3s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Goat cheese — card not captured.

Why verdict Logged (write-truth verified): Goat cheese — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said Logged half a cup goat cheese. Assumed 0.5 cup as the default serving. Tell me if you want a different amount.

2 · Card shown Logged half a cup goat cheese. Assumed 0.5 cup as the default serving. Tell me if you want a different amount.

3 · App data rows written created food_log_entry: Goat cheese ×1 (0.5 cup) 200 cal · 11.9g P · 0.1g C · 16.4g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:02:34.094Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Goat cheese" at 90% confidence.
Matched your past food logs to "Codex Test Dairy Soft Goat Cheese" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Count one naan bread half." (difficulty —)1.0s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.0s

1 · TTS said What food did you have?

2 · Card shown What food did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:02:46.226Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"I had five ounces venison." (difficulty —)0.3s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said Which type of venison was it? There are several with different nutrition.

2 · Card shown Which type of venison was it? There are several with different nutrition.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:02:57.785Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Add one cup mung beans cooked." (difficulty —)2.7s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.7s

1 · TTS said What exact food and amount should I use for one cup mung beans cooked? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for one cup mung beans cooked? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:03:12.226Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Track half a cup kimchi." (difficulty —)1.0s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Kimchi — card not captured.

Why verdict Logged (write-truth verified): Kimchi — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.0s

1 · TTS said Got it, I logged Kimchi (half a cup (75 g)).

2 · Card shown I logged Kimchi (half a cup (75 g)) — let me know if that's not right.

3 · App data rows written created food_log_entry: Kimchi ×1 (half a cup (75 g)) 11 cal · 0.8g P · 1.8g C · 0.4g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:03:23.844Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Log two buckwheat pancakes." (difficulty —)1.0s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.0s

1 · TTS said What food did you have?

2 · Card shown What food did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:03:35.975Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"I ate one brioche roll small." (difficulty —)2.7s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.7s

1 · TTS said Was it plain, or did you have anything on it like butter or cream cheese? I did not log it yet because add-ons can change the nutrition.

2 · Card shown Was it plain, or did you have anything on it like butter or cream cheese? I did not log it yet because add-ons can change the nutrition.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:03:49.973Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Record four ounces calamari grilled." (difficulty —)2.1s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.1s

1 · TTS said What exact food and amount should I use for four ounces calamari grilled? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for four ounces calamari grilled? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:04:03.195Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Snack was two Babybel original cheeses." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Babybel — card not captured.

Why verdict Logged (write-truth verified): Babybel — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged two babybel original cheeses.

2 · Card shown Logged two babybel original cheeses.

3 · App data rows written created food_log_entry: Babybel ×1 (2 babybel original) 140 cal · 10g P · 0g C · 11g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:04:14.597Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Babybel" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.5 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Add one teaspoon chili oil." (difficulty —)2.3s

Verdict Expected LOG — should log the entry. FAIL: WRITE-TRUTH FAIL — WRONG/MISSING item "chili oil" — no saved row with matching identity (rows: Labneh with Olive Oil)

Why verdict WRITE-TRUTH FAIL — WRONG/MISSING item "chili oil" — no saved row with matching identity (rows: Labneh with Olive Oil)

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.3s

1 · TTS said Got it, I logged Codex Test Dairy Labneh with Olive Oil.

2 · Card shown I logged Codex Test Dairy Labneh with Olive Oil — let me know if that's not right.

3 · App data rows written created food_log_entry: Labneh with Olive Oil ×0.021 (cup) 4 cal · 0.2g P · 0.1g C · 0.3g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:04:28.111Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"I had one cup millet." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Cooked millet — card not captured.

Why verdict Logged (write-truth verified): Cooked millet — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged one cup millet. Assumed 1 cup as the default serving. Tell me if you want a different amount.

2 · Card shown Logged one cup millet. Assumed 1 cup as the default serving. Tell me if you want a different amount.

3 · App data rows written created food_log_entry: Cooked millet ×1 (1 cup) 207 cal · 6.1g P · 41.2g C · 1.7g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:04:39.320Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Cooked millet" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Log one medium yam." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Yam — card not captured.

Why verdict Logged (write-truth verified): Yam — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged one medium yam. Assumed 1 yam as the default serving. Tell me if you want a different amount.

2 · Card shown Logged one medium yam. Assumed 1 yam as the default serving. Tell me if you want a different amount.

3 · App data rows written created food_log_entry: Yam ×1 (1 yam) 174 cal · 2.3g P · 41.3g C · 0.3g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:04:50.746Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Yam" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Track ten ounces matcha latte water base." (difficulty —)2.7s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.7s

1 · TTS said What exact food and amount should I use for ten ounces matcha latte water base? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for ten ounces matcha latte water base? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:05:16.231Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Count one cup sugar snap peas raw." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Snap peas — card not captured.

Why verdict Logged (write-truth verified): Snap peas — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged one cup sugar snap peas raw. Assumed 1 cup as the default serving. Tell me if you want a different amount.

2 · Card shown Logged one cup sugar snap peas raw. Assumed 1 cup as the default serving. Tell me if you want a different amount.

3 · App data rows written created food_log_entry: Snap peas ×1 (1 cup) 41 cal · 2.7g P · 7.4g C · 0.2g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:05:27.616Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Snap peas" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.7 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Dinner was five ounces lamb loin." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Lamb loin — card not captured.

Why verdict Logged (write-truth verified): Lamb loin — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged five ounces lamb loin.

2 · Card shown Logged five ounces lamb loin.

3 · App data rows written created food_log_entry: Lamb loin ×1 (5 oz) 220 cal · 28.1g P · 0g C · 12g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:05:38.981Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Lamb loin" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.6 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"I had one cup average tomato bisque." (difficulty —)4.5s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 4.5s

1 · TTS said I couldn't find Average Tomato Bisque Packaged Food in my nutrition database. What exact name is on the package? I did not log it because I'd only be guessing the nutrition. If you prefer, tell me to log a generic packaged food.

2 · Card shown I couldn't find Average Tomato Bisque Packaged Food in my nutrition database. What exact name is on the package? I did not log it because I'd only be guessing the nutrition. If you prefer, tell me to log a generic packaged food.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:05:54.641Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Add two seaweed snacks packs." (difficulty —)3.9s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 3.9s

1 · TTS said I couldn't find Add Two Seaweed Snacks Packaged Food in my nutrition database. What exact name is on the package? I did not log it because I'd only be guessing the nutrition. If you prefer, tell me to log a generic packaged food.

2 · Card shown I couldn't find Add Two Seaweed Snacks Packaged Food in my nutrition database. What exact name is on the package? I did not log it because I'd only be guessing the nutrition. If you prefer, tell me to log a generic packaged food.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:06:09.714Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Log one cup mango chunks." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Mango — card not captured.

Why verdict Logged (write-truth verified): Mango — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged one cup mango chunks. Assumed 1 cup as the default serving. Tell me if you want a different amount.

2 · Card shown Logged one cup mango chunks. Assumed 1 cup as the default serving. Tell me if you want a different amount.

3 · App data rows written created food_log_entry: Mango ×1 (1 cup) 99 cal · 1.3g P · 24.8g C · 0.7g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:06:21.022Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Mango" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.5 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Track two tablespoons muhammara." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Muhammara — card not captured.

Why verdict Logged (write-truth verified): Muhammara — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged two tablespoons muhammara.

2 · Card shown Logged two tablespoons muhammara.

3 · App data rows written created food_log_entry: Muhammara ×1 (2 tbsp) 70 cal · 1.5g P · 6g C · 5g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:06:32.428Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Muhammara" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.5 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"I ate one medium lychee cup ten pieces." (difficulty —)4.1s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 4.1s

1 · TTS said I couldn't find Medium Lychee Packaged Food in my nutrition database. What exact name is on the package? I did not log it because I'd only be guessing the nutrition. If you prefer, tell me to log a generic packaged food.

2 · Card shown I couldn't find Medium Lychee Packaged Food in my nutrition database. What exact name is on the package? I did not log it because I'd only be guessing the nutrition. If you prefer, tell me to log a generic packaged food.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:06:47.818Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Put down one cup teff." (difficulty —)0.8s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.8s

1 · TTS said I couldn't convert one cup of Teff, Cooked from the serving information I found. What does the label say for grams or calories? I did not log it yet.

2 · Card shown I couldn't convert one cup of Teff, Cooked from the serving information I found. What does the label say for grams or calories? I did not log it yet.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:06:59.610Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Record four ounces turkey cutlet." (difficulty —)1.2s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.2s

1 · TTS said What exact food and amount should I use for four ounces turkey cutlet? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for four ounces turkey cutlet? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:07:12.018Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Add my usual No Cow bar." (difficulty —)0.1s

Verdict Expected LOG — should log the entry. FAIL: Reported could-not-confirm (write-confirmation false negative).

Why verdict Reported could-not-confirm (write-confirmation false negative).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I wasn't sure which food you meant to correct. Could you try again with a bit more detail?

2 · Card shown I wasn't sure which food you meant to correct. Could you try again with a bit more detail?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: No Cow No Cow Cookie Dough

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-nocow",
    "name": "No Cow Cookie Dough",
    "brand": "No Cow",
    "source": "manual",
    "servingLabel": "bar",
    "servingQuantity": 1,
    "servingUnit": "bar",
    "macrosPerServing": {
      "calories": 190,
      "proteinGrams": 20,
      "fatGrams": 5,
      "carbsGrams": 19
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

food_default

{
  "type": "food_default",
  "phrase": "my usual No Cow bar",
  "foodItemId": "seed-g-nocow",
  "settings": {
    "quickLogMode": "fixed",
    "defaultQuantity": 1,
    "defaultServingLabel": "bar",
    "displayLabel": "my usual No Cow bar",
    "confirmBeforeLog": false
  }
}

snapshot captured 2026-08-02T22:07:23.234Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "my usual no cow bar" at 96% confidence.
Matched your saved foods and aliases to "No Cow Bar" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.2 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"I had my usual labneh dip." (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Labneh with Olive Oil — card not captured.

Why verdict Logged (write-truth verified): Labneh with Olive Oil — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Logged Codex Test Dairy Labneh with Olive Oil. Using your saved "my usual labneh dip". Tell me if that is wrong.

2 · Card shown Logged Codex Test Dairy Labneh with Olive Oil. Using your saved "my usual labneh dip". Tell me if that is wrong.

3 · App data rows written created food_log_entry: Labneh with Olive Oil ×1 (cup) 180 cal · 10g P · 5g C · 14g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Codex Test Dairy Labneh with Olive Oil

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-labneh",
    "name": "Labneh with Olive Oil",
    "brand": "Codex Test Dairy",
    "source": "manual",
    "servingLabel": "cup",
    "servingQuantity": 1,
    "servingUnit": "cup",
    "macrosPerServing": {
      "calories": 180,
      "proteinGrams": 10,
      "fatGrams": 14,
      "carbsGrams": 5
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

food_default

{
  "type": "food_default",
  "phrase": "my usual labneh dip",
  "foodItemId": "seed-g-labneh",
  "settings": {
    "quickLogMode": "fixed",
    "defaultQuantity": 1,
    "defaultServingLabel": "cup",
    "displayLabel": "my usual labneh dip",
    "confirmBeforeLog": false
  }
}

snapshot captured 2026-08-02T22:07:34.513Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "my usual labneh dip" at 96% confidence.
Matched the common-food list to "Labneh" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Lunch was my usual poke cup." (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Salmon Poke Cup — card not captured.

Why verdict Logged (write-truth verified): Salmon Poke Cup — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Logged Codex Test Kitchen Salmon Poke Cup. Using your saved "my usual poke cup". Tell me if that is wrong.

2 · Card shown Logged Codex Test Kitchen Salmon Poke Cup. Using your saved "my usual poke cup". Tell me if that is wrong.

3 · App data rows written created food_log_entry: Salmon Poke Cup ×1 (cup) 380 cal · 28g P · 32g C · 14g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Codex Test Kitchen Salmon Poke Cup

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-poke",
    "name": "Salmon Poke Cup",
    "brand": "Codex Test Kitchen",
    "source": "manual",
    "servingLabel": "cup",
    "servingQuantity": 1,
    "servingUnit": "cup",
    "macrosPerServing": {
      "calories": 380,
      "proteinGrams": 28,
      "fatGrams": 14,
      "carbsGrams": 32
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

food_default

{
  "type": "food_default",
  "phrase": "my usual poke cup",
  "foodItemId": "seed-g-poke",
  "settings": {
    "quickLogMode": "fixed",
    "defaultQuantity": 1,
    "defaultServingLabel": "cup",
    "displayLabel": "my usual poke cup",
    "confirmBeforeLog": false
  }
}

snapshot captured 2026-08-02T22:07:45.763Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "my usual poke cup" at 96% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Dinner side was my usual rice." (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Cooked Sushi Rice — card not captured.

Why verdict Logged (write-truth verified): Cooked Sushi Rice — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Logged Cooked Sushi Rice. Using your saved "my usual rice". Tell me if that is wrong.

2 · Card shown Logged Cooked Sushi Rice. Using your saved "my usual rice". Tell me if that is wrong.

3 · App data rows written created food_log_entry: Cooked Sushi Rice ×1 (cup) 200 cal · 4g P · 44g C · 0.3g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Cooked Sushi Rice

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-rice",
    "name": "Cooked Sushi Rice",
    "brand": null,
    "source": "manual",
    "servingLabel": "cup",
    "servingQuantity": 1,
    "servingUnit": "cup",
    "macrosPerServing": {
      "calories": 200,
      "proteinGrams": 4,
      "fatGrams": 0.3,
      "carbsGrams": 44
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

food_default

{
  "type": "food_default",
  "phrase": "my usual rice",
  "foodItemId": "seed-g-rice",
  "settings": {
    "quickLogMode": "fixed",
    "defaultQuantity": 1,
    "defaultServingLabel": "cup",
    "displayLabel": "my usual rice",
    "confirmBeforeLog": false
  }
}

snapshot captured 2026-08-02T22:07:57.021Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "my usual rice" at 96% confidence.
Matched the common-food list to "Cooked white rice" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Track my usual macadamias." (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Macadamia Nuts — card not captured.

Why verdict Logged (write-truth verified): Macadamia Nuts — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Logged Macadamia Nuts. Using your saved "my usual macadamias". Tell me if that is wrong.

2 · Card shown Logged Macadamia Nuts. Using your saved "my usual macadamias". Tell me if that is wrong.

3 · App data rows written created food_log_entry: Macadamia Nuts ×1 (ounce) 204 cal · 2.2g P · 4g C · 21g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Macadamia Nuts

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-mac",
    "name": "Macadamia Nuts",
    "brand": null,
    "source": "manual",
    "servingLabel": "ounce",
    "servingQuantity": 1,
    "servingUnit": "oz",
    "macrosPerServing": {
      "calories": 204,
      "proteinGrams": 2.2,
      "fatGrams": 21,
      "carbsGrams": 4
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

food_default

{
  "type": "food_default",
  "phrase": "my usual macadamias",
  "foodItemId": "seed-g-mac",
  "settings": {
    "quickLogMode": "fixed",
    "defaultQuantity": 1,
    "defaultServingLabel": "ounce",
    "displayLabel": "my usual macadamias",
    "confirmBeforeLog": false
  }
}

snapshot captured 2026-08-02T22:08:08.258Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "my usual macadamias" at 96% confidence.
Matched the common-food list to "Macadamias" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Log my default coconut milk." (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Coconut Milk Beverage — card not captured.

Why verdict Logged (write-truth verified): Coconut Milk Beverage — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Logged Coconut Milk Beverage. Using your saved "my default coconut milk". Tell me if that is wrong.

2 · Card shown Logged Coconut Milk Beverage. Using your saved "my default coconut milk". Tell me if that is wrong.

3 · App data rows written created food_log_entry: Coconut Milk Beverage ×1 (cup) 45 cal · 0g P · 1g C · 4.5g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Coconut Milk Beverage

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-cocomilk",
    "name": "Coconut Milk Beverage",
    "brand": null,
    "source": "manual",
    "servingLabel": "cup",
    "servingQuantity": 1,
    "servingUnit": "cup",
    "macrosPerServing": {
      "calories": 45,
      "proteinGrams": 0,
      "fatGrams": 4.5,
      "carbsGrams": 1
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

food_default

{
  "type": "food_default",
  "phrase": "my default coconut milk",
  "foodItemId": "seed-g-cocomilk",
  "settings": {
    "quickLogMode": "fixed",
    "defaultQuantity": 1,
    "defaultServingLabel": "cup",
    "displayLabel": "my default coconut milk",
    "confirmBeforeLog": false
  }
}

snapshot captured 2026-08-02T22:08:19.743Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "my default coconut milk" at 96% confidence.
Matched the common-food list to "Milk" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Add my usual morning matcha latte." (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Matcha Latte Oat — card not captured.

Why verdict Logged (write-truth verified): Matcha Latte Oat — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Logged Codex Test Cafe Matcha Latte Oat. Using your saved "my usual morning matcha latte". Tell me if that is wrong.

2 · Card shown Logged Codex Test Cafe Matcha Latte Oat. Using your saved "my usual morning matcha latte". Tell me if that is wrong.

3 · App data rows written created food_log_entry: Matcha Latte Oat ×1 (cup) 140 cal · 2g P · 22g C · 4g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Codex Test Cafe Matcha Latte Oat

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-matcha",
    "name": "Matcha Latte Oat",
    "brand": "Codex Test Cafe",
    "source": "manual",
    "servingLabel": "cup",
    "servingQuantity": 1,
    "servingUnit": "cup",
    "macrosPerServing": {
      "calories": 140,
      "proteinGrams": 2,
      "fatGrams": 4,
      "carbsGrams": 22
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

food_default

{
  "type": "food_default",
  "phrase": "my usual morning matcha latte",
  "foodItemId": "seed-g-matcha",
  "settings": {
    "quickLogMode": "fixed",
    "defaultQuantity": 1,
    "defaultServingLabel": "cup",
    "displayLabel": "my usual morning matcha latte",
    "confirmBeforeLog": false
  }
}

snapshot captured 2026-08-02T22:08:30.791Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "my usual morning matcha latte" at 96% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"I had my normal buckwheat porridge." (difficulty —)0.1s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Decisive guess with stated assumption + invited correction (good-guessing).

Why verdict Decisive guess with stated assumption + invited correction (good-guessing).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Logged Buckwheat Porridge. Using your saved "my normal buckwheat porridge". Tell me if that is wrong.

2 · Card shown Logged Buckwheat Porridge. Using your saved "my normal buckwheat porridge". Tell me if that is wrong.

3 · App data rows written created food_log_entry: Buckwheat Porridge ×1 (cup) 155 cal · 5.7g P · 33g C · 1g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Buckwheat Porridge

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-buck",
    "name": "Buckwheat Porridge",
    "brand": null,
    "source": "manual",
    "servingLabel": "cup",
    "servingQuantity": 1,
    "servingUnit": "cup",
    "macrosPerServing": {
      "calories": 155,
      "proteinGrams": 5.7,
      "fatGrams": 1,
      "carbsGrams": 33
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

food_default

{
  "type": "food_default",
  "phrase": "my normal buckwheat porridge",
  "foodItemId": "seed-g-buck",
  "settings": {
    "quickLogMode": "fixed",
    "defaultQuantity": 1,
    "defaultServingLabel": "cup",
    "displayLabel": "my normal buckwheat porridge",
    "confirmBeforeLog": false
  }
}

snapshot captured 2026-08-02T22:08:42.047Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "my normal buckwheat porridge" at 96% confidence.
Matched the common-food list to "Cooked buckwheat" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.6 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Track my go-to miso soup." (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Miso Soup Cup — card not captured.

Why verdict Logged (write-truth verified): Miso Soup Cup — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Logged Codex Test Kitchen Miso Soup Cup. Using your saved "my go to miso soup". Tell me if that is wrong.

2 · Card shown Logged Codex Test Kitchen Miso Soup Cup. Using your saved "my go to miso soup". Tell me if that is wrong.

3 · App data rows written created food_log_entry: Miso Soup Cup ×1 (cup) 40 cal · 3g P · 5g C · 1g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Codex Test Kitchen Miso Soup Cup

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-miso",
    "name": "Miso Soup Cup",
    "brand": "Codex Test Kitchen",
    "source": "manual",
    "servingLabel": "cup",
    "servingQuantity": 1,
    "servingUnit": "cup",
    "macrosPerServing": {
      "calories": 40,
      "proteinGrams": 3,
      "fatGrams": 1,
      "carbsGrams": 5
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

food_default

{
  "type": "food_default",
  "phrase": "my go-to miso soup",
  "foodItemId": "seed-g-miso",
  "settings": {
    "quickLogMode": "fixed",
    "defaultQuantity": 1,
    "defaultServingLabel": "cup",
    "displayLabel": "my go-to miso soup",
    "confirmBeforeLog": false
  }
}

snapshot captured 2026-08-02T22:08:53.278Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "my go to miso soup" at 96% confidence.
Matched the common-food list to "Miso soup" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Log my usual evening rooibos." (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Rooibos Tea — card not captured.

Why verdict Logged (write-truth verified): Rooibos Tea — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Logged Rooibos Tea. Using your saved "my usual evening rooibos". Tell me if that is wrong.

2 · Card shown Logged Rooibos Tea. Using your saved "my usual evening rooibos". Tell me if that is wrong.

3 · App data rows written created food_log_entry: Rooibos Tea ×1 (cup) 0 cal · 0g P · 0g C · 0g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Rooibos Tea

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-rooibos",
    "name": "Rooibos Tea",
    "brand": null,
    "source": "manual",
    "servingLabel": "cup",
    "servingQuantity": 1,
    "servingUnit": "cup",
    "macrosPerServing": {
      "calories": 0,
      "proteinGrams": 0,
      "fatGrams": 0,
      "carbsGrams": 0
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

food_default

{
  "type": "food_default",
  "phrase": "my usual evening rooibos",
  "foodItemId": "seed-g-rooibos",
  "settings": {
    "quickLogMode": "fixed",
    "defaultQuantity": 1,
    "defaultServingLabel": "cup",
    "displayLabel": "my usual evening rooibos",
    "confirmBeforeLog": false
  }
}

snapshot captured 2026-08-02T22:09:04.512Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "my usual evening rooibos" at 96% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Add the same goat cheese as yesterday." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Goat cheese — card not captured.

Why verdict Logged (write-truth verified): Goat cheese — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged Goat cheese. Using your recent Goat cheese history. Tell me if that is wrong.

2 · Card shown Logged Goat cheese. Using your recent Goat cheese history. Tell me if that is wrong.

3 · App data rows written created food_log_entry: Goat cheese ×1 (0.5 cup) 200 cal · 11.9g P · 0.1g C · 16.4g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Codex Test Dairy Soft Goat Cheese

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-hist-goat",
    "name": "Soft Goat Cheese",
    "brand": "Codex Test Dairy",
    "source": "manual",
    "servingLabel": "cup",
    "servingQuantity": 1,
    "servingUnit": "cup",
    "macrosPerServing": {
      "calories": 200,
      "proteinGrams": 12,
      "fatGrams": 16,
      "carbsGrams": 1
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 0.5 for seed-g-hist-goat at 2026-07-17T13:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-hist-goat",
  "quantity": 0.5,
  "loggedAt": "2026-07-17T13:00:00.000Z"
}

snapshot captured 2026-08-02T22:09:27.380Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Goat cheese" at 90% confidence.
Matched your past food logs to "Goat cheese" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

UNVERIFIEDunclassifiedshould log the entry"Same No Cow bar as yesterday." (difficulty —)0.1s

Verdict Expected LOG — should log the entry. UNVERIFIED: CONTEXT-FRAGMENT — app asked for missing context; needs multi-turn replay (not a single-shot fail).

Why verdict CONTEXT-FRAGMENT — app asked for missing context; needs multi-turn replay (not a single-shot fail).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Which exact same no cow bar as yesterday should replace Goat cheese?

2 · Card shown Which exact same no cow bar as yesterday should replace Goat cheese?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: No Cow No Cow Bar

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-hist-nocow",
    "name": "No Cow Bar",
    "brand": "No Cow",
    "source": "manual",
    "servingLabel": "bar",
    "servingQuantity": 1,
    "servingUnit": "bar",
    "macrosPerServing": {
      "calories": 190,
      "proteinGrams": 20,
      "fatGrams": 5,
      "carbsGrams": 19
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-hist-nocow at 2026-07-17T16:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-hist-nocow",
  "quantity": 1,
  "loggedAt": "2026-07-17T16:00:00.000Z"
}

snapshot captured 2026-08-02T22:09:38.703Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "No Cow Bar" at 90% confidence.
Matched your saved foods and aliases to "my usual no cow bar" at 65% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Log the same tomato bisque I had last night." (difficulty —)0.3s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Tomato Bisque — card not captured.

Why verdict Logged (write-truth verified): Tomato Bisque — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said Logged Codex Test Kitchen Tomato Bisque. Using your recent Codex Test Kitchen Tomato Bisque history. Tell me if that is wrong.

2 · Card shown Logged Codex Test Kitchen Tomato Bisque. Using your recent Codex Test Kitchen Tomato Bisque history. Tell me if that is wrong.

3 · App data rows written created food_log_entry: Tomato Bisque ×1 (bowl) 130 cal · 3g P · 18g C · 5g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Codex Test Kitchen Tomato Bisque

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-hist-bisque",
    "name": "Tomato Bisque",
    "brand": "Codex Test Kitchen",
    "source": "manual",
    "servingLabel": "bowl",
    "servingQuantity": 1,
    "servingUnit": "bowl",
    "macrosPerServing": {
      "calories": 130,
      "proteinGrams": 3,
      "fatGrams": 5,
      "carbsGrams": 18
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-hist-bisque at 2026-07-17T19:30:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-hist-bisque",
  "quantity": 1,
  "loggedAt": "2026-07-17T19:30:00.000Z"
}

snapshot captured 2026-08-02T22:09:50.149Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Tomato bisque" at 90% confidence.
Matched your past food logs to "Codex Test Kitchen Tomato Bisque" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Same oatmeal as this morning." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Oatmeal Bowl — card not captured.

Why verdict Logged (write-truth verified): Oatmeal Bowl — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged Oatmeal Bowl. Using your recent Oatmeal Bowl history. Tell me if that is wrong.

2 · Card shown Logged Oatmeal Bowl. Using your recent Oatmeal Bowl history. Tell me if that is wrong.

3 · App data rows written created food_log_entry: Oatmeal Bowl ×1 (bowl) 150 cal · 5g P · 27g C · 2.5g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Oatmeal Bowl

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-hist-oats",
    "name": "Oatmeal Bowl",
    "brand": null,
    "source": "manual",
    "servingLabel": "bowl",
    "servingQuantity": 1,
    "servingUnit": "bowl",
    "macrosPerServing": {
      "calories": 150,
      "proteinGrams": 5,
      "fatGrams": 2.5,
      "carbsGrams": 27
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-hist-oats at 2026-07-18T07:30:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-hist-oats",
  "quantity": 1,
  "loggedAt": "2026-07-18T07:30:00.000Z"
}

snapshot captured 2026-08-02T22:10:01.490Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Add the same sea bass I had earlier." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Baked Sea Bass — card not captured.

Why verdict Logged (write-truth verified): Baked Sea Bass — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged Baked Sea Bass. Using your recent Baked Sea Bass history. Tell me if that is wrong.

2 · Card shown Logged Baked Sea Bass. Using your recent Baked Sea Bass history. Tell me if that is wrong.

3 · App data rows written created food_log_entry: Baked Sea Bass ×1 (fillet) 160 cal · 30g P · 0g C · 4g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Baked Sea Bass

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-hist-bass",
    "name": "Baked Sea Bass",
    "brand": null,
    "source": "manual",
    "servingLabel": "fillet",
    "servingQuantity": 1,
    "servingUnit": "fillet",
    "macrosPerServing": {
      "calories": 160,
      "proteinGrams": 30,
      "fatGrams": 4,
      "carbsGrams": 0
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-hist-bass at 2026-07-18T12:30:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-hist-bass",
  "quantity": 1,
  "loggedAt": "2026-07-18T12:30:00.000Z"
}

snapshot captured 2026-08-02T22:10:12.863Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your past food logs to "Baked Sea Bass" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.2 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Same smoothie as yesterday." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Dragon Fruit Smoothie — card not captured.

Why verdict Logged (write-truth verified): Dragon Fruit Smoothie — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged Codex Test Recipe Dragon Fruit Smoothie. Using your recent Codex Test Recipe Dragon Fruit Smoothie history. Tell me if that is wrong.

2 · Card shown Logged Codex Test Recipe Dragon Fruit Smoothie. Using your recent Codex Test Recipe Dragon Fruit Smoothie history. Tell me if that is wrong.

3 · App data rows written created food_log_entry: Dragon Fruit Smoothie ×1 (smoothie) 250 cal · 20g P · 35g C · 3g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Codex Test Recipe Dragon Fruit Smoothie

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-hist-smooth",
    "name": "Dragon Fruit Smoothie",
    "brand": "Codex Test Recipe",
    "source": "manual",
    "servingLabel": "smoothie",
    "servingQuantity": 1,
    "servingUnit": "smoothie",
    "macrosPerServing": {
      "calories": 250,
      "proteinGrams": 20,
      "fatGrams": 3,
      "carbsGrams": 35
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-hist-smooth at 2026-07-17T10:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-hist-smooth",
  "quantity": 1,
  "loggedAt": "2026-07-17T10:00:00.000Z"
}

snapshot captured 2026-08-02T22:10:24.200Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.2 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Log the duck I had last time." (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Duck Breast Skinless — card not captured.

Why verdict Logged (write-truth verified): Duck Breast Skinless — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Logged Duck Breast Skinless. Using your saved "Duck Breast Skinless". Tell me if that is wrong.

2 · Card shown Logged Duck Breast Skinless. Using your saved "Duck Breast Skinless". Tell me if that is wrong.

3 · App data rows written created food_log_entry: Duck Breast Skinless ×1 (serving) 140 cal · 22g P · 0g C · 5g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Duck Breast Skinless

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-hist-duck",
    "name": "Duck Breast Skinless",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 140,
      "proteinGrams": 22,
      "fatGrams": 5,
      "carbsGrams": 0
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-hist-duck at 2026-07-16T18:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-hist-duck",
  "quantity": 1,
  "loggedAt": "2026-07-16T18:00:00.000Z"
}

snapshot captured 2026-08-02T22:10:35.502Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.9 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Same naan as earlier today." (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Naan Half — card not captured.

Why verdict Logged (write-truth verified): Naan Half — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged Naan Half. Using your recent Naan Half history. Tell me if that is wrong.

2 · Card shown Logged Naan Half. Using your recent Naan Half history. Tell me if that is wrong.

3 · App data rows written created food_log_entry: Naan Half ×1 (piece) 160 cal · 5g P · 26g C · 4g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Naan Half

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-hist-naan",
    "name": "Naan Half",
    "brand": null,
    "source": "manual",
    "servingLabel": "piece",
    "servingQuantity": 1,
    "servingUnit": "piece",
    "macrosPerServing": {
      "calories": 160,
      "proteinGrams": 5,
      "fatGrams": 4,
      "carbsGrams": 26
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-hist-naan at 2026-07-18T07:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-hist-naan",
  "quantity": 1,
  "loggedAt": "2026-07-18T07:00:00.000Z"
}

snapshot captured 2026-08-02T22:10:46.831Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Naan" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Snack was one EPIC chicken sesame BBQ bar." (difficulty —)1.5s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.5s

1 · TTS said I couldn't find Snack Was One EPIC Chicken Sesame BBQ Snack Bar in my nutrition database. What exact name is on the package? I did not log it because I'd only be guessing the nutrition. If you prefer, tell me to log a generic snack bar.

2 · Card shown I couldn't find Snack Was One EPIC Chicken Sesame BBQ Snack Bar in my nutrition database. What exact name is on the package? I did not log it because I'd only be guessing the nutrition. If you prefer, tell me to log a generic snack bar.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:10:59.477Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Finished one Fairlife Core Power cookies and cream bottle." (difficulty —)2.1s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.1s

1 · TTS said What food did you have?

2 · Card shown What food did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:11:12.850Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Track one Two Good vanilla cup." (difficulty —)1.9s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.9s

1 · TTS said Which one did you mean? 1. TWO GOOD Vanilla Lowfat Greek Yogurt, Vanilla, 2. Vanilla Blended Greek Nonfat Yogurt — Good & Gather

2 · Card shown Which one did you mean? 1. TWO GOOD Vanilla Lowfat Greek Yogurt, Vanilla, 2. Vanilla Blended Greek Nonfat Yogurt — Good & Gather

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:11:26.135Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Save one No Cow chocolate fudge brownie bar." (difficulty —)0.7s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.7s

1 · TTS said Which exact save one no cow chocolate fudge brownie bar should replace Naan Half?

2 · Card shown Which exact save one no cow chocolate fudge brownie bar should replace Naan Half?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:11:37.964Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"I ate one RXBAR maple sea salt." (difficulty —)0.5s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.5s

1 · TTS said What exact food and amount should I use for one rxbar maple sea salt? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for one rxbar maple sea salt? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:11:49.571Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Log a blood orange and one No Cow bar." (difficulty —)0.1s

Verdict Expected LOG — should log the entry. FAIL: Reported could-not-confirm (write-confirmation false negative).

Why verdict Reported could-not-confirm (write-confirmation false negative).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I wasn't sure which food you meant to correct. Could you try again with a bit more detail?

2 · Card shown I wasn't sure which food you meant to correct. Could you try again with a bit more detail?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:12:00.862Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "No Cow Bar" at 90% confidence.
Matched your past food logs to "blood orange" at 82% confidence.
Matched your saved foods and aliases to "my usual no cow bar" at 65% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.0 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"I had oatmeal and matcha." (difficulty —)1.9s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.9s

1 · TTS said I need to resolve matcha before I log this meal. What should I use for matcha?

2 · Card shown I need to resolve matcha before I log this meal. What should I use for matcha?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:12:13.836Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Breakfast was soft boiled eggs and naan." (difficulty —)0.3s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Decisive guess with stated assumption + invited correction (good-guessing).

Why verdict Decisive guess with stated assumption + invited correction (good-guessing).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said Logged Egg and Naan. Assumed catalog default servings where you did not say an amount — tell me if that is not right.

2 · Card shown Logged Egg and Naan. Assumed catalog default servings where you did not say an amount — tell me if that is not right.

3 · App data rows written created food_log_entry: Egg ×1 (50 g) 72 cal · 6.3g P · 0.4g C · 4.8g F
created food_log_entry: Naan ×1 (90 g) 262 cal · 8.1g P · 45g C · 5.1g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:12:25.316Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Lunch was a banh mi and an Asian pear." (difficulty —)0.3s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Decisive guess with stated assumption + invited correction (good-guessing).

Why verdict Decisive guess with stated assumption + invited correction (good-guessing).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said Logged a banh mi and Pear. Assumed catalog default servings where you did not say an amount — tell me if that is not right.

2 · Card shown Logged a banh mi and Pear. Assumed catalog default servings where you did not say an amount — tell me if that is not right.

3 · App data rows written created food_log_entry: Banh mi ×1 (1 banh mi) 690 cal · 30g P · 90g C · 24g F
created food_log_entry: Pear ×1 (178 g) 101 cal · 0.7g P · 27.1g C · 0.2g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:12:36.809Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Snack was macadamias and gooseberries." (difficulty —)0.3s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said How much should I log for Macadamias and Gooseberries? I did not log them yet because the amounts were not clear.

2 · Card shown How much should I log for Macadamias and Gooseberries? I did not log them yet because the amounts were not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:12:48.220Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Gooseberries" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.9 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Dinner was sea bass and broccoli florets." (difficulty —)0.5s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Decisive guess with stated assumption + invited correction (good-guessing).

Why verdict Decisive guess with stated assumption + invited correction (good-guessing).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.5s

1 · TTS said Logged Baked Sea Bass and Broccoli. Assumed catalog default servings where you did not say an amount — tell me if that is not right.

2 · Card shown Logged Baked Sea Bass and Broccoli. Assumed catalog default servings where you did not say an amount — tell me if that is not right.

3 · App data rows written created food_log_entry: Baked Sea Bass ×1 (fillet) 160 cal · 30g P · 0g C · 4g F
created food_log_entry: Broccoli ×1 (100 g) 35 cal · 2.4g P · 7.2g C · 0.4g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:12:59.858Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"I ate labneh and strawberries." (difficulty —)0.3s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said How much should I log for Labneh and Strawberries? I did not log them yet because the amounts were not clear.

2 · Card shown How much should I log for Labneh and Strawberries? I did not log them yet because the amounts were not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:13:11.265Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Strawberries" at 90% confidence.
Matched your past food logs to "Labneh" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.9 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Track swordfish and millet." (difficulty —)1.0s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.0s

1 · TTS said I need to resolve swordfish before I log this meal. What should I use for swordfish?

2 · Card shown I need to resolve swordfish before I log this meal. What should I use for swordfish?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:13:34.931Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Add goat cheese and endive." (difficulty —)0.4s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.4s

1 · TTS said How much should I log for Goat cheese and Endive? I did not log them yet because the amounts were not clear.

2 · Card shown How much should I log for Goat cheese and Endive? I did not log them yet because the amounts were not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:13:47.501Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "goat cheese" at 90% confidence.
Matched the common-food list to "Goat cheese" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.8 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"I had an EPIC bar and club soda." (difficulty —)2.0s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.0s

1 · TTS said Which an EPIC bar should I use? I did not log it yet because more than one match fits.

2 · Card shown Which an EPIC bar should I use? I did not log it yet because more than one match fits.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:13:59.703Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Log tomato bisque and watercress." (difficulty —)0.4s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.4s

1 · TTS said How much should I log for Tomato bisque and Watercress? I did not log them yet because the amounts were not clear.

2 · Card shown How much should I log for Tomato bisque and Watercress? I did not log them yet because the amounts were not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:14:11.206Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Tomato bisque" at 90% confidence.
Matched your past food logs to "Codex Test Kitchen Tomato Bisque" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.0 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Breakfast was oatmeal, mango, and coffee." (difficulty —)0.4s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Decisive guess with stated assumption + invited correction (good-guessing).

Why verdict Decisive guess with stated assumption + invited correction (good-guessing).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.4s

1 · TTS said Logged Cooked oatmeal, Mango, and Coffee. Assumed catalog default servings where you did not say an amount — tell me if that is not right.

2 · Card shown Logged Cooked oatmeal, Mango, and Coffee. Assumed catalog default servings where you did not say an amount — tell me if that is not right.

3 · App data rows written created food_log_entry: Cooked oatmeal ×1 (100 g) 71 cal · 2.5g P · 12g C · 1.5g F
created food_log_entry: Mango ×1 (100 g) 60 cal · 0.8g P · 15g C · 0.4g F
created food_log_entry: Coffee ×1 (100 g) 1 cal · 0.1g P · 0g C · 0g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:14:22.819Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"I ate a dragon fruit, Babybel, and seaweed snacks." (difficulty —)0.4s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.4s

1 · TTS said How much should I log for Babybel and Seaweed snacks? I did not log them yet because the amounts were not clear.

2 · Card shown How much should I log for Babybel and Seaweed snacks? I did not log them yet because the amounts were not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:14:34.374Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Seaweed snacks" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Dinner was lamb loin, millet, and cabbage." (difficulty —)0.4s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.4s

1 · TTS said How much should I log for Lamb loin, Cooked millet, and Cabbage? I did not log them yet because the amounts were not clear.

2 · Card shown How much should I log for Lamb loin, Cooked millet, and Cabbage? I did not log them yet because the amounts were not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:14:45.981Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Lamb loin" at 90% confidence.
Matched your past food logs to "Lamb loin" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Snack was snap peas and muhammara." (difficulty —)0.3s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said How much should I log for Snap peas and Muhammara? I did not log them yet because the amounts were not clear.

2 · Card shown How much should I log for Snap peas and Muhammara? I did not log them yet because the amounts were not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:14:57.446Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Muhammara" at 90% confidence.
Matched your past food logs to "Snap peas" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Lunch was a grain bowl and rooibos." (difficulty —)1.8s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.8s

1 · TTS said I need to resolve rooibos before I log this meal. What should I use for rooibos?

2 · Card shown I need to resolve rooibos before I log this meal. What should I use for rooibos?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:15:10.414Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"I had soba, adzuki beans, and duck." (difficulty —)1.2s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.2s

1 · TTS said I need to resolve soba and duck before I log this meal. What should I use for soba and duck?

2 · Card shown I need to resolve soba and duck before I log this meal. What should I use for soba and duck?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:15:22.779Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Track labneh, honey, and macadamias." (difficulty —)0.6s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.6s

1 · TTS said How much should I log for Labneh, Honey, and Macadamias? I did not log them yet because the amounts were not clear.

2 · Card shown How much should I log for Labneh, Honey, and Macadamias? I did not log them yet because the amounts were not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:15:34.548Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Macadamias" at 90% confidence.
Matched your past food logs to "Labneh" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Add eggs, chicken apple sausage, and pomegranate juice." (difficulty —)0.3s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said How much should I log for Egg, Chicken apple sausage, and Pomegranate juice? I did not log them yet because the amounts were not clear.

2 · Card shown How much should I log for Egg, Chicken apple sausage, and Pomegranate juice? I did not log them yet because the amounts were not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:15:46.044Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "egg" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould ask a clarifying question"I ate pasta and a side of collards." (difficulty —)0.7s

Verdict Expected CLARIFY — should ask a clarifying question. FAIL: WRONG item vs seeded default/key — WRONG/MISSING item "a side of collards" — no saved row with matching identity (rows: Cooked pasta, Collard greens, Cooked pasta, Collard greens)

Why verdict WRONG item vs seeded default/key — WRONG/MISSING item "a side of collards" — no saved row with matching identity (rows: Cooked pasta, Collard greens, Cooked pasta, Collard greens)

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.7s

1 · TTS said Logged Cooked pasta and Collard greens. Assumed catalog default servings where you did not say an amount — tell me if that is not right.

2 · Card shown Logged Cooked pasta and Collard greens. Assumed catalog default servings where you did not say an amount — tell me if that is not right.

3 · App data rows written created food_log_entry: Cooked pasta ×1 (140 g) 220 cal · 8.1g P · 43.3g C · 1.3g F
created food_log_entry: Collard greens ×1 (100 g) 33 cal · 2.7g P · 5.6g C · 0.7g F
created food_log_entry: Cooked pasta ×1 (140 g) 220 cal · 8.1g P · 43.3g C · 1.3g F
created food_log_entry: Collard greens ×1 (100 g) 33 cal · 2.7g P · 5.6g C · 0.7g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:15:57.936Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Log lychees and walnut butter." (difficulty —)0.4s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.4s

1 · TTS said How much should I log for Lychees and Walnut butter? I did not log them yet because the amounts were not clear.

2 · Card shown How much should I log for Lychees and Walnut butter? I did not log them yet because the amounts were not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:16:09.540Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Walnut butter" at 90% confidence.
Matched your past food logs to "Walnut butter" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.2 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Breakfast was cereal and coconut milk." (difficulty —)1.0s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.0s

1 · TTS said I need to resolve cereal before I log this meal. What should I use for cereal?

2 · Card shown I need to resolve cereal before I log this meal. What should I use for cereal?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:16:21.715Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"I had a bibimbap with rice, egg, spinach, and gochujang" (difficulty —)1.5s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Egg, Bibimbap, Cooked white rice, Spinach, Gochujang — card not captured.

Why verdict Logged (write-truth verified): Egg, Bibimbap, Cooked white rice, Spinach, Gochujang — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.5s

1 · TTS said Logged Bibimbap. Includes Cooked white rice, one egg, Spinach, and Gochujang.

2 · Card shown Logged Bibimbap. Includes Cooked white rice, one egg, Spinach, and Gochujang.

3 · App data rows written created food_log_entry: Bibimbap ×1 (serving) 249 cal · 10.8g P · 37.8g C · 5.6g F
created food_log_entry: Cooked white rice ×1 (100 g) 130 cal · 2.7g P · 28g C · 0.3g F
created food_log_entry: Egg ×1 (1 egg) 72 cal · 6.3g P · 0.4g C · 4.8g F
created food_log_entry: Spinach ×1 (30 g) 7 cal · 0.9g P · 1.1g C · 0.1g F
created food_log_entry: Gochujang ×1 (18 g) 40 cal · 0.9g P · 8.3g C · 0.4g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:16:34.386Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould ask a clarifying question"Log a smoothie with dragon fruit, banana, and collagen." (difficulty —)2.7s

Verdict Expected CLARIFY — should ask a clarifying question. FAIL: Logged a BLIND guess — no stated assumption, no correction invited.

Why verdict Logged a BLIND guess — no stated assumption, no correction invited.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.7s

1 · TTS said Logged Smoothie. Includes Dragon Fruit, one banana, and Collagen peptides.

2 · Card shown Logged Smoothie. Includes Dragon Fruit, one banana, and Collagen peptides.

3 · App data rows written created food_log_entry: Smoothie ×1 (serving) 587 cal · 92.6g P · 56.3g C · 0.8g F
created food_log_entry: Dragon Fruit ×1 (a) 122 cal · 1.3g P · 29.2g C · 0.4g F
created food_log_entry: Banana ×1 (1 banana) 105 cal · 1.3g P · 27.1g C · 0.4g F
created food_log_entry: Collagen peptides ×1 (100 g) 360 cal · 90g P · 0g C · 0g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:16:48.232Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould ask a clarifying question"I had a grain bowl with millet, mung beans, and muhammara." (difficulty —)8.8s

Verdict Expected CLARIFY — should ask a clarifying question. FAIL: Logged a BLIND guess — no stated assumption, no correction invited.

Why verdict Logged a BLIND guess — no stated assumption, no correction invited.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 8.8s

1 · TTS said Logged Grain Bowl. Includes Millet, Cooked, Mung Beans, Cooked, and Muhammara.

2 · Card shown Logged Grain Bowl. Includes Millet, Cooked, Mung Beans, Cooked, and Muhammara.

3 · App data rows written pending_created food_log_entry: a grain bowl with millet, mung beans, and muhammara (pending)
created food_log_entry: Grain Bowl ×1 (serving) 566 cal · 19.6g P · 80.1g C · 19.5g F
created food_log_entry: Millet, Cooked ×1 (1 serving) 207 cal · 6.1g P · 41.2g C · 1.7g F
created food_log_entry: Mung Beans, Cooked ×1 (1 cup) 289 cal · 12g P · 32.9g C · 12.8g F
created food_log_entry: Muhammara ×1 (2 tbsp) 70 cal · 1.5g P · 6g C · 5g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:17:08.231Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Breakfast was smorrebrod with egg and smoked salmon." (difficulty —)0.5s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.5s

1 · TTS said How much should I log for Smoked salmon? I did not log it yet because the amount was not clear.

2 · Card shown How much should I log for Smoked salmon? I did not log it yet because the amount was not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:17:19.916Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Smoked salmon" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.8 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould ask a clarifying question"I made tagine with lamb, apricots, chickpeas, and oil." (difficulty —)3.6s

Verdict Expected CLARIFY — should ask a clarifying question. FAIL: Logged a BLIND guess — no stated assumption, no correction invited.

Why verdict Logged a BLIND guess — no stated assumption, no correction invited.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 3.6s

1 · TTS said Logging your I made tagine with lamb, apricots, chickpeas, and oil — still looking up the nutrition details; they'll appear in your log shortly.

2 · Card shown Logging your I made tagine with lamb, apricots, chickpeas, and oil — still looking up the nutrition details; they'll appear in your log shortly.

3 · App data rows written pending_created food_log_entry: I made tagine with lamb, apricots, chickpeas, and oil (pending)

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:17:41.277Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Dinner was a poke bowl with rice, salmon, edamame, and seaweed." (difficulty —)1.1s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.1s

1 · TTS said What should I use for rice?

2 · Card shown What should I use for rice?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:17:58.398Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"I had laksa with shrimp and tofu." (difficulty —)2.5s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Laksa, Shrimp, Firm tofu — card not captured.

Why verdict Logged (write-truth verified): Laksa, Shrimp, Firm tofu — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.5s

1 · TTS said Logged Laksa. Includes Shrimp and Firm tofu.

2 · Card shown Logged Laksa. Includes Shrimp and Firm tofu.

3 · App data rows written created food_log_entry: Laksa ×1 (serving) 172 cal · 23.7g P · 2.9g C · 8.8g F
created food_log_entry: Shrimp ×1 (1 oz) 28 cal · 6.7g P · 0.1g C · 0.1g F
created food_log_entry: Firm tofu ×1 (100 g) 144 cal · 17g P · 2.8g C · 8.7g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:18:12.101Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"Log a duck wrap with hoisin and cucumber." (difficulty —)1.7s

Verdict Expected LOG — should log the entry. PASS: Logged (write-truth verified): Duck Wrap, Hoisin Sauce, Cucumber — card not captured.

Why verdict Logged (write-truth verified): Duck Wrap, Hoisin Sauce, Cucumber — card not captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.7s

1 · TTS said Logged Duck Wrap. Includes Hoisin Sauce and Cucumber.

2 · Card shown Logged Duck Wrap. Includes Hoisin Sauce and Cucumber.

3 · App data rows written created food_log_entry: Duck Wrap ×1 (serving) 50 cal · 1.2g P · 10.7g C · 0.6g F
created food_log_entry: Hoisin Sauce ×1 (1 tablespoon) 35 cal · 0.5g P · 7.1g C · 0.5g F
created food_log_entry: Cucumber ×1 (100 g) 15 cal · 0.7g P · 3.6g C · 0.1g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:18:24.970Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould ask a clarifying question"I built a yogurt bowl with buckwheat, honey, and strawberries." (difficulty —)4.5s

Verdict Expected CLARIFY — should ask a clarifying question. FAIL: Logged a BLIND guess — no stated assumption, no correction invited.

Why verdict Logged a BLIND guess — no stated assumption, no correction invited.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 4.5s

1 · TTS said Logged Yogurt Bowl. Includes Buckwheat, Honey, and Strawberries, Raw.

2 · Card shown Logged Yogurt Bowl. Includes Buckwheat, Honey, and Strawberries, Raw.

3 · App data rows written created food_log_entry: Yogurt Bowl ×1 (serving) 701 cal · 23.6g P · 150.9g C · 6.1g F
created food_log_entry: Buckwheat ×1 (1 serving) 583 cal · 22.6g P · 121.6g C · 5.8g F
created food_log_entry: Honey ×1 (21 g) 64 cal · 0.1g P · 17.3g C · 0g F
created food_log_entry: Strawberries, Raw ×1 (1 cup) 54 cal · 0.9g P · 12g C · 0.3g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:18:40.649Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"Dinner was mapo tofu with pork, tofu, and rice." (difficulty —)2.1s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of logging (no saved row).

Why verdict OVER-ASK — asked instead of logging (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.1s

1 · TTS said What should I use for tofu?

2 · Card shown What should I use for tofu?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:18:54.031Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Add some millet." (difficulty —)0.1s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said How much should I log for Cooked millet? I did not log it yet because the amount was not clear.

2 · Card shown How much should I log for Cooked millet? I did not log it yet because the amount was not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:19:05.361Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Cooked millet" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould ask a clarifying question"I had buckwheat porridge." (difficulty —)3.3s

Verdict Expected CLARIFY — should ask a clarifying question. FAIL: Logged a BLIND guess — no stated assumption, no correction invited.

Why verdict Logged a BLIND guess — no stated assumption, no correction invited.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 3.3s

1 · TTS said How much buckwheat porridge did you have?

2 · Card shown How much buckwheat porridge did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:19:19.830Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Log soba for dinner." (difficulty —)0.4s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Decisive guess with stated assumption + invited correction (good-guessing).

Why verdict Decisive guess with stated assumption + invited correction (good-guessing).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.4s

1 · TTS said Logged Noodles, Japanese, Soba, Cooked. Using your saved "Noodles, Japanese, Soba, Cooked". Tell me if that is wrong.

2 · Card shown Logged Noodles, Japanese, Soba, Cooked. Using your saved "Noodles, Japanese, Soba, Cooked". Tell me if that is wrong.

3 · App data rows written created food_log_entry: Noodles, Japanese, Soba, Cooked ×1 (1 cup) 113 cal · 5.8g P · 24.4g C · 0.1g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:19:31.410Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Track udon noodles." (difficulty —)3.8s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 3.8s

1 · TTS said How much udon noodles did you have?

2 · Card shown How much udon noodles did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:19:46.363Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"I ate seaweed snacks." (difficulty —)0.1s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said How much should I log for Seaweed snacks? I did not log it yet because the amount was not clear.

2 · Card shown How much should I log for Seaweed snacks? I did not log it yet because the amount was not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:19:57.582Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Seaweed snacks" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Add macadamias." (difficulty —)0.1s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said How much should I log for Macadamias? I did not log it yet because the amount was not clear.

2 · Card shown How much should I log for Macadamias? I did not log it yet because the amount was not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:20:08.832Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Macadamias" at 90% confidence.
Matched your past food logs to "Macadamia Nuts" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.2 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Log a bowl of bisque." (difficulty —)2.7s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.7s

1 · TTS said What exact food and amount should I use for a bowl of bisque? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for a bowl of bisque? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:20:22.700Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"I had a plate of tagine." (difficulty —)2.7s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.7s

1 · TTS said What exact food and amount should I use for a plate of tagine? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for a plate of tagine? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:20:36.676Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Track pomegranate juice." (difficulty —)0.1s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said How much should I log for Pomegranate juice? I did not log it yet because the amount was not clear.

2 · Card shown How much should I log for Pomegranate juice? I did not log it yet because the amount was not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:20:47.978Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Pomegranate juice" at 90% confidence.
Matched your past food logs to "Pomegranate juice" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Add matcha latte." (difficulty —)1.2s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.2s

1 · TTS said How much matcha latte did you have?

2 · Card shown How much matcha latte did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:21:00.348Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould ask a clarifying question"I drank soju." (difficulty —)0.2s

Verdict Expected CLARIFY — should ask a clarifying question. FAIL: Logged a BLIND guess — no stated assumption, no correction invited.

Why verdict Logged a BLIND guess — no stated assumption, no correction invited.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said How much soju did you have?

2 · Card shown How much soju did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:21:11.700Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Log mochi ice cream." (difficulty —)5.6s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 5.6s

1 · TTS said How much mochi ice cream did you have?

2 · Card shown How much mochi ice cream did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:21:28.492Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Had some pumpkin seeds." (difficulty —)0.1s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said How much should I log for Pumpkin seeds? I did not log it yet because the amount was not clear.

2 · Card shown How much should I log for Pumpkin seeds? I did not log it yet because the amount was not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:21:39.738Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Pumpkin seeds" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.5 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Add seaweed snacks." (difficulty —)0.1s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said How much should I log for Seaweed snacks? I did not log it yet because the amount was not clear.

2 · Card shown How much should I log for Seaweed snacks? I did not log it yet because the amount was not clear.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:21:50.988Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Seaweed snacks" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould ask a clarifying question"I had a No Cow bar." (difficulty —)0.5s

Verdict Expected CLARIFY — should ask a clarifying question. FAIL: Logged a BLIND guess — no stated assumption, no correction invited.

Why verdict Logged a BLIND guess — no stated assumption, no correction invited.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.5s

1 · TTS said What food did you have?

2 · Card shown What food did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:22:02.660Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Log a labneh tub." (difficulty —)1.7s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.7s

1 · TTS said What food did you have?

2 · Card shown What food did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:22:27.072Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Track a coconut drink." (difficulty —)1.2s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Decisive guess with stated assumption + invited correction (good-guessing).

Why verdict Decisive guess with stated assumption + invited correction (good-guessing).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.2s

1 · TTS said Got it, I logged Beverages, Coconut Water, Ready-To-Drink, Unsweetened.

2 · Card shown I logged Beverages, Coconut Water, Ready-To-Drink, Unsweetened — let me know if that's not right.

3 · App data rows written created food_log_entry: Beverages, Coconut Water, Ready-To-Drink, Unsweetened ×1 (1 serving) 44 cal · 0.5g P · 10.3g C · 0g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:22:39.498Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"I ate a banh mi." (difficulty —)0.2s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Decisive guess with stated assumption + invited correction (good-guessing).

Why verdict Decisive guess with stated assumption + invited correction (good-guessing).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged a banh mi. Assumed 1 banh mi as the default serving. Tell me if you want a different amount.

2 · Card shown Logged a banh mi. Assumed 1 banh mi as the default serving. Tell me if you want a different amount.

3 · App data rows written created food_log_entry: Banh mi ×1 (1 banh mi) 690 cal · 30g P · 90g C · 24g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:22:50.836Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "a banh mi" at 90% confidence.
Matched the common-food list to "Banh mi" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Add a collagen shake." (difficulty —)1.7s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.7s

1 · TTS said What food did you have?

2 · Card shown What food did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:23:03.823Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"I had matcha latte." (difficulty —)1.2s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.2s

1 · TTS said How much matcha latte did you have?

2 · Card shown How much matcha latte did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:23:16.421Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Log a roll." (difficulty —)0.4s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Decisive guess with stated assumption + invited correction (good-guessing).

Why verdict Decisive guess with stated assumption + invited correction (good-guessing).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.4s

1 · TTS said Got it, I logged Roll, White, Hot Dog Bun (1 roll (1 hot dog bun)).

2 · Card shown I logged Roll, White, Hot Dog Bun (1 roll (1 hot dog bun)) — let me know if that's not right.

3 · App data rows written created food_log_entry: Roll, White, Hot Dog Bun ×1 (1 roll (1 hot dog bun)) 120 cal · 4.2g P · 22.1g C · 1.6g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:23:27.733Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Track a duck sandwich." (difficulty —)1.1s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.1s

1 · TTS said What food did you have?

2 · Card shown What food did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:23:40.114Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"I had naan pizza." (difficulty —)5.3s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 5.3s

1 · TTS said What exact food and amount should I use for Pizza? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for Pizza? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:23:56.769Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Add a savory muffin." (difficulty —)1.3s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Asked a clarification as expected.

Why verdict Asked a clarification as expected.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.3s

1 · TTS said What exact food and amount should I use for a savory muffin? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for a savory muffin? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:24:09.238Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Log my usual matcha size." (difficulty —)0.1s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Decisive guess with stated assumption + invited correction (good-guessing).

Why verdict Decisive guess with stated assumption + invited correction (good-guessing).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Logged matcha. Using your saved "log my usual matcha size". Tell me if that is wrong.

2 · Card shown Logged matcha. Using your saved "log my usual matcha size". Tell me if that is wrong.

3 · App data rows written created food_log_entry: matcha ×1 (serving) 150 cal · 10g P · 15g C · 5g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: matcha

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-matcha-confirm",
    "name": "matcha",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 150,
      "proteinGrams": 10,
      "fatGrams": 5,
      "carbsGrams": 15
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

food_default

{
  "type": "food_default",
  "phrase": "log my usual matcha size",
  "foodItemId": "seed-g-matcha-confirm",
  "settings": {
    "quickLogMode": "fixed",
    "defaultQuantity": 1,
    "defaultServingLabel": "serving",
    "displayLabel": "log my usual matcha size",
    "confirmBeforeLog": false
  }
}

snapshot captured 2026-08-02T22:24:20.499Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "log my usual matcha size" at 96% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Add the regular supper portion." (difficulty —)0.1s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Decisive guess with stated assumption + invited correction (good-guessing).

Why verdict Decisive guess with stated assumption + invited correction (good-guessing).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Logged supper portion. Using your saved "add the regular supper portion". Tell me if that is wrong.

2 · Card shown Logged supper portion. Using your saved "add the regular supper portion". Tell me if that is wrong.

3 · App data rows written created food_log_entry: supper portion ×1 (serving) 150 cal · 10g P · 15g C · 5g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: supper portion

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-supper-portion",
    "name": "supper portion",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 150,
      "proteinGrams": 10,
      "fatGrams": 5,
      "carbsGrams": 15
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

food_default

{
  "type": "food_default",
  "phrase": "add the regular supper portion",
  "foodItemId": "seed-g-supper-portion",
  "settings": {
    "quickLogMode": "fixed",
    "defaultQuantity": 1,
    "defaultServingLabel": "serving",
    "displayLabel": "add the regular supper portion",
    "confirmBeforeLog": false
  }
}

snapshot captured 2026-08-02T22:24:31.739Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "add the regular supper portion" at 96% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.2 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"Track my default evening snack." (difficulty —)0.1s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Decisive guess with stated assumption + invited correction (good-guessing).

Why verdict Decisive guess with stated assumption + invited correction (good-guessing).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Logged snack. Using your saved "track my default evening snack". Tell me if that is wrong.

2 · Card shown Logged snack. Using your saved "track my default evening snack". Tell me if that is wrong.

3 · App data rows written created food_log_entry: snack ×1 (serving) 150 cal · 10g P · 15g C · 5g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: snack

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-snack-amt",
    "name": "snack",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 150,
      "proteinGrams": 10,
      "fatGrams": 5,
      "carbsGrams": 15
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

food_default

{
  "type": "food_default",
  "phrase": "track my default evening snack",
  "foodItemId": "seed-g-snack-amt",
  "settings": {
    "quickLogMode": "fixed",
    "defaultQuantity": 1,
    "defaultServingLabel": "serving",
    "displayLabel": "track my default evening snack",
    "confirmBeforeLog": false
  }
}

snapshot captured 2026-08-02T22:24:42.998Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "track my default evening snack" at 96% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould ask a clarifying question"I had the usual brunch size." (difficulty —)0.1s

Verdict Expected CLARIFY — should ask a clarifying question. PASS: Decisive guess with stated assumption + invited correction (good-guessing).

Why verdict Decisive guess with stated assumption + invited correction (good-guessing).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Logged brunch. Using your saved "i had the usual brunch size". Tell me if that is wrong.

2 · Card shown Logged brunch. Using your saved "i had the usual brunch size". Tell me if that is wrong.

3 · App data rows written created food_log_entry: brunch ×1 (serving) 150 cal · 10g P · 15g C · 5g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: brunch

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-brunch-size",
    "name": "brunch",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 150,
      "proteinGrams": 10,
      "fatGrams": 5,
      "carbsGrams": 15
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

food_default

{
  "type": "food_default",
  "phrase": "i had the usual brunch size",
  "foodItemId": "seed-g-brunch-size",
  "settings": {
    "quickLogMode": "fixed",
    "defaultQuantity": 1,
    "defaultServingLabel": "serving",
    "displayLabel": "i had the usual brunch size",
    "confirmBeforeLog": false
  }
}

snapshot captured 2026-08-02T22:24:54.388Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "i had the usual brunch size" at 96% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

UNVERIFIEDunclassifiedshould complete the pending action"Half a cup." (difficulty —)1.4s

Verdict Expected CLARIFY_ANSWER — should complete the pending action. UNVERIFIED: CLARIFY_ANSWER not gradeable single-shot — no pending clarification in this replay (app re-asked / mis-parsed as fresh); needs multi-turn replay.

Why verdict CLARIFY_ANSWER not gradeable single-shot — no pending clarification in this replay (app re-asked / mis-parsed as fresh); needs multi-turn replay.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 1.4s

1 · TTS said What food did you have?

2 · Card shown What food did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:25:06.960Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

UNVERIFIEDunclassifiedshould complete the pending action"One cup, plain." (difficulty —)0.1s

Why verdict CLARIFY_ANSWER not gradeable single-shot — no pending clarification in this replay (app re-asked / mis-parsed as fresh); needs multi-turn replay.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Should I add that as a to-do?

2 · Card shown Should I add that as a to-do?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:25:18.172Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

UNVERIFIEDunclassifiedshould complete the pending action"The No Cow chocolate fudge." (difficulty —)0.8s

Why verdict CLARIFY_ANSWER not gradeable single-shot — no pending clarification in this replay (app re-asked / mis-parsed as fresh); needs multi-turn replay.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.8s

1 · TTS said Which exact no cow chocolate fudge should replace brunch?

2 · Card shown Which exact no cow chocolate fudge should replace brunch?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:25:30.065Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

UNVERIFIEDunclassifiedshould complete the pending action"Coconut milk beverage." (difficulty —)0.0s

Why verdict CLARIFY_ANSWER not gradeable single-shot — no pending clarification in this replay (app re-asked / mis-parsed as fresh); needs multi-turn replay.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.0s

1 · TTS said Should I add that as a to-do?

2 · Card shown Should I add that as a to-do?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:25:41.275Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "coconut milk beverage" at 96% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

UNVERIFIEDunclassifiedshould complete the pending action"No, that's not it — the unsweetened one." (difficulty —)0.1s

Why verdict CLARIFY_ANSWER not gradeable single-shot — no pending clarification in this replay (app re-asked / mis-parsed as fresh); needs multi-turn replay.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said What should I change about brunch?

2 · Card shown What should I change about brunch?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:25:52.567Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.6 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould complete the pending action"Wrong one. The blood orange, not navel." (difficulty —)0.8s

Verdict Expected CLARIFY_ANSWER — should complete the pending action. PASS: Completed the pending action.

Why verdict Completed the pending action.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.8s

1 · TTS said Logged Blood orange and Sliced Navel Oranges — NOT A BRANDED ITEM. Assumed catalog default servings where you did not say an amount — tell me if that is not right.

2 · Card shown Logged Blood orange and Sliced Navel Oranges — NOT A BRANDED ITEM. Assumed catalog default servings where you did not say an amount — tell me if that is not right.

3 · App data rows written created food_log_entry: Blood orange ×1 (140 g) 70 cal · 1.3g P · 18.1g C · 0.2g F
created food_log_entry: Sliced Navel Oranges ×1 (5 ONZ) 70 cal · 1g P · 18g C · 0g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:26:04.761Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

UNVERIFIEDunclassifiedshould complete the pending action"No, not the flavored labneh." (difficulty —)0.1s

Why verdict CLARIFY_ANSWER not gradeable single-shot — no pending clarification in this replay (app re-asked / mis-parsed as fresh); needs multi-turn replay.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said What should I change about Sliced Navel Oranges?

2 · Card shown What should I change about Sliced Navel Oranges?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:26:15.945Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Labneh" at 90% confidence.
Matched your past food logs to "Labneh" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.6 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould decline"Cancel that log." (difficulty —)0.5s

Verdict Expected DECLINE — should decline. FAIL: Wrote when it should decline.

Why verdict Wrote when it should decline.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.5s

1 · TTS said Okay — cancelled Blood orange.

2 · Card shown Okay — cancelled Blood orange.

3 · App data rows written deleted food_log_entry:

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: Pending Food Item

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-corr-pending-3",
    "name": "Pending Food Item",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 150,
      "proteinGrams": 10,
      "fatGrams": 5,
      "carbsGrams": 15
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-corr-pending-3 at 2026-07-18T11:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-corr-pending-3",
  "quantity": 1,
  "loggedAt": "2026-07-18T11:00:00.000Z"
}

snapshot captured 2026-08-02T22:26:46.026Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.5 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

UNVERIFIEDunclassifiedshould complete the pending action"No, that's not the bar I meant." (difficulty —)0.1s

Why verdict CLARIFY_ANSWER not gradeable single-shot — no pending clarification in this replay (app re-asked / mis-parsed as fresh); needs multi-turn replay.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said What should I change about Sliced Navel Oranges?

2 · Card shown What should I change about Sliced Navel Oranges?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:27:02.004Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.8 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

UNVERIFIEDunclassifiedshould complete the pending action"Not the large — the medium blood orange." (difficulty —)6.8s

Why verdict CLARIFY_ANSWER not gradeable single-shot — no pending clarification in this replay (app re-asked / mis-parsed as fresh); needs multi-turn replay.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 6.8s

1 · TTS said What food did you have?

2 · Card shown What food did you have?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:27:19.985Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould complete the pending action"Wrong item. Pick the grilled swordfish." (difficulty —)4.7s

Verdict Expected CLARIFY_ANSWER — should complete the pending action. FAIL: Did not complete the pending action.

Why verdict Did not complete the pending action.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 4.7s

1 · TTS said Got it. I’ve selected grilled swordfish instead.

2 · Card shown Got it. I’ve selected grilled swordfish instead.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:27:35.985Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 2.0 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould complete the pending action"That's not it; the plain one." (difficulty —)2.1s

Verdict Expected CLARIFY_ANSWER — should complete the pending action. FAIL: Did not complete the pending action.

Why verdict Did not complete the pending action.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.1s

1 · TTS said Got it — you want the plain chocolate fudge bar, not a branded variety. Say "replace brunch with plain chocolate fudge" to log it correctly.

2 · Card shown Got it — you want the plain chocolate fudge bar, not a branded variety. Say "replace brunch with plain chocolate fudge" to log it correctly.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:28:00.632Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.6 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould complete the pending action"No, the other naan." (difficulty —)3.8s

Verdict Expected CLARIFY_ANSWER — should complete the pending action. PASS: Completed the pending action.

Why verdict Completed the pending action.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 3.8s

1 · TTS said Updated to the other naan.

2 · Card shown Updated the food entry from "Sliced Navel Oranges" to "the other naan".

3 · App data rows written updated food_log_entry: Naan ×1 (90 g) 262 cal · 8.1g P · 45g C · 5.1g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:28:15.580Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

UNVERIFIEDunclassifiedshould complete the pending action"Not those macadamias — the roasted ones." (difficulty —)4.3s

Why verdict CLARIFY_ANSWER not gradeable single-shot — no pending clarification in this replay (app re-asked / mis-parsed as fresh); needs multi-turn replay.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 4.3s

1 · TTS said What exact food and amount should I use for Macadamias? I did not log it yet because I could not safely finish that food log.

2 · Card shown What exact food and amount should I use for Macadamias? I did not log it yet because I could not safely finish that food log.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:28:42.204Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould delete the entry"Delete the blood orange from earlier." (difficulty —)0.1s

Verdict Expected DELETE — should delete the entry. FAIL: Did not delete (not understood / no removal).

Why verdict Did not delete (not understood / no removal).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I could not find a list matching earlier. Which list has blood orange?

2 · Card shown I could not find a list matching earlier. Which list has blood orange?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:29:05.332Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "blood orange" at 90% confidence.
Matched the common-food list to "Blood orange" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.9 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

UNVERIFIEDunclassifiedshould delete the entry"Remove that No Cow bar." (difficulty —)0.3s

Verdict Expected DELETE — should delete the entry. UNVERIFIED: DELETE unverified — delete wording/event exists, but no database removal was captured.

Why verdict DELETE unverified — delete wording/event exists, but no database removal was captured.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said Deleted No Cow Bar.

2 · Card shown Deleted No Cow Bar.

3 · App data rows written deleted food_log_entry:

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:29:13.416Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "No Cow Bar" at 90% confidence.
Matched your saved foods and aliases to "my usual no cow bar" at 65% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 3.3 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould delete the entry"Delete the buckwheat porridge entry." (difficulty —)0.1s

Verdict Expected DELETE — should delete the entry. FAIL: Did not delete (not understood / no removal).

Why verdict Did not delete (not understood / no removal).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Which food entry did you want to correct?

2 · Card shown Which food entry did you want to correct?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:29:28.343Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "buckwheat porridge" at 90% confidence.
Matched the common-food list to "Cooked buckwheat" at 90% confidence.
Matched your past food logs to "Buckwheat" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 2.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould delete the entry"Take off the eggs." (difficulty —)0.1s

Verdict Expected DELETE — should delete the entry. FAIL: Did not delete (not understood / no removal).

Why verdict Did not delete (not understood / no removal).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Which food entry did you want to correct?

2 · Card shown Which food entry did you want to correct?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:29:39.594Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "egg" at 90% confidence.
Matched the common-food list to "Egg" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould delete the entry"Remove the gooseberries." (difficulty —)0.1s

Verdict Expected DELETE — should delete the entry. FAIL: Did not delete (not understood / no removal).

Why verdict Did not delete (not understood / no removal).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Which food entry did you want to correct?

2 · Card shown Which food entry did you want to correct?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:29:50.933Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Gooseberries" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.5 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould delete the entry"Delete the swordfish log." (difficulty —)0.2s

Verdict Expected DELETE — should delete the entry. PASS: Deleted/removed the requested food with database removal proof.

Why verdict Deleted/removed the requested food with database removal proof.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Deleted swordfish.

2 · Card shown Deleted swordfish.

3 · App data rows written deleted food_log_entry:

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:30:09.265Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.7 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould delete the entry"Scratch the walnut butter." (difficulty —)0.1s

Verdict Expected DELETE — should delete the entry. FAIL: Did not delete (not understood / no removal).

Why verdict Did not delete (not understood / no removal).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said Which food entry did you want to correct?

2 · Card shown Which food entry did you want to correct?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:30:13.440Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Walnut butter" at 90% confidence.
Matched your past food logs to "Walnut butter" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.7 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould delete the entry"Remove the coconut milk entry." (difficulty —)0.3s

Verdict Expected DELETE — should delete the entry. PASS: Deleted/removed the requested food with database removal proof.

Why verdict Deleted/removed the requested food with database removal proof.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said Deleted Coconut Milk Beverage.

2 · Card shown Deleted Coconut Milk Beverage.

3 · App data rows written deleted food_log_entry:

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:30:24.859Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Milk" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 2.0 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould delete the entry"Delete that mochi I logged last night." (difficulty —)0.3s

Verdict Expected DELETE — should delete the entry. PASS: Deleted/removed the requested food with database removal proof.

Why verdict Deleted/removed the requested food with database removal proof.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said Deleted mochi.

2 · Card shown Deleted mochi.

3 · App data rows written deleted food_log_entry:

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: mochi

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-del-0-mochi",
    "name": "mochi",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 200,
      "proteinGrams": 5,
      "fatGrams": 10,
      "carbsGrams": 20
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-del-0-mochi at 2026-07-17T20:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-del-0-mochi",
  "quantity": 1,
  "loggedAt": "2026-07-17T20:00:00.000Z"
}

snapshot captured 2026-08-02T22:30:42.779Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 2.0 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould delete the entry"Remove yesterday's tiramisu." (difficulty —)0.2s

Verdict Expected DELETE — should delete the entry. PASS: Deleted/removed the requested food with database removal proof.

Why verdict Deleted/removed the requested food with database removal proof.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Deleted tiramisu.

2 · Card shown Deleted tiramisu.

3 · App data rows written deleted food_log_entry:

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: tiramisu

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-del-1-tiramisu",
    "name": "tiramisu",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 200,
      "proteinGrams": 5,
      "fatGrams": 10,
      "carbsGrams": 20
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-del-1-tiramisu at 2026-07-17T20:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-del-1-tiramisu",
  "quantity": 1,
  "loggedAt": "2026-07-17T20:00:00.000Z"
}

snapshot captured 2026-08-02T22:31:06.403Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.7 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould delete the entry"Delete the leftover sushi entry." (difficulty —)0.3s

Verdict Expected DELETE — should delete the entry. PASS: Deleted/removed the requested food with database removal proof.

Why verdict Deleted/removed the requested food with database removal proof.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said Deleted sushi.

2 · Card shown Deleted sushi.

3 · App data rows written deleted food_log_entry:

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: sushi

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-del-2-sushi",
    "name": "sushi",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 200,
      "proteinGrams": 5,
      "fatGrams": 10,
      "carbsGrams": 20
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-del-2-sushi at 2026-07-17T20:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-del-2-sushi",
  "quantity": 1,
  "loggedAt": "2026-07-17T20:00:00.000Z"
}

snapshot captured 2026-08-02T22:31:17.568Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.9 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould delete the entry"Take out the soju from last night." (difficulty —)2.0s

Verdict Expected DELETE — should delete the entry. FAIL: Did not delete (not understood / no removal).

Why verdict Did not delete (not understood / no removal).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.0s

1 · TTS said I did not make any app changes. What would you like me to do with that?

2 · Card shown I did not make any app changes. What would you like me to do with that?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: soju

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-del-3-soju",
    "name": "soju",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 200,
      "proteinGrams": 5,
      "fatGrams": 10,
      "carbsGrams": 20
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-del-3-soju at 2026-07-17T20:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-del-3-soju",
  "quantity": 1,
  "loggedAt": "2026-07-17T20:00:00.000Z"
}

snapshot captured 2026-08-02T22:31:23.819Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 2.0 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould delete the entry"Remove the croissant log from earlier this week." (difficulty —)0.2s

Verdict Expected DELETE — should delete the entry. PASS: Deleted/removed the requested food with database removal proof.

Why verdict Deleted/removed the requested food with database removal proof.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Deleted croissant.

2 · Card shown Deleted croissant.

3 · App data rows written deleted food_log_entry:

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: croissant

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-del-4-croissant",
    "name": "croissant",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 200,
      "proteinGrams": 5,
      "fatGrams": 10,
      "carbsGrams": 20
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-del-4-croissant at 2026-07-17T20:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-del-4-croissant",
  "quantity": 1,
  "loggedAt": "2026-07-17T20:00:00.000Z"
}

snapshot captured 2026-08-02T22:31:41.794Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 2.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould delete the entry"Delete the candy I shouldn't have logged." (difficulty —)0.2s

Verdict Expected DELETE — should delete the entry. PASS: Deleted/removed the requested food with database removal proof.

Why verdict Deleted/removed the requested food with database removal proof.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Deleted candy.

2 · Card shown Deleted candy.

3 · App data rows written deleted food_log_entry:

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: candy

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-del-5-candy",
    "name": "candy",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 200,
      "proteinGrams": 5,
      "fatGrams": 10,
      "carbsGrams": 20
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-del-5-candy at 2026-07-17T20:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-del-5-candy",
  "quantity": 1,
  "loggedAt": "2026-07-17T20:00:00.000Z"
}

snapshot captured 2026-08-02T22:31:53.066Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.9 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould update the entry"Change the sequence blood orange quantity to two." (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. FAIL: Reported could-not-confirm (write-confirmation false negative).

Why verdict Reported could-not-confirm (write-confirmation false negative).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I could not confirm that was saved, so I did not mark it as done.

2 · Card shown I could not confirm that the change was saved, so I did not mark it as done.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:31:57.849Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "blood orange" at 90% confidence.
Matched the common-food list to "Blood orange" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.9 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould update the entry"Actually make the No Cow bar two bars." (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. FAIL: Did not update — misroute-into-food (no verified mutation).

Why verdict Did not update — misroute-into-food (no verified mutation).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I couldn't identify which food entry you meant. Which one would you like to correct?

2 · Card shown I couldn't identify which food entry you meant. Which one would you like to correct?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:32:09.156Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "No Cow Bar" at 90% confidence.
Matched your saved foods and aliases to "my usual no cow bar" at 60% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 2.0 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould update the entry"Update the buckwheat porridge to one and a half cups." (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. FAIL: Reported could-not-confirm (write-confirmation false negative).

Why verdict Reported could-not-confirm (write-confirmation false negative).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I could not confirm that was saved, so I did not mark it as done.

2 · Card shown I could not confirm that the change was saved, so I did not mark it as done.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:32:20.552Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "buckwheat porridge" at 90% confidence.
Matched the common-food list to "Cooked buckwheat" at 90% confidence.
Matched your past food logs to "Buckwheat" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.7 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould update the entry"Change the eggs to three." (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. FAIL: Reported could-not-confirm (write-confirmation false negative).

Why verdict Reported could-not-confirm (write-confirmation false negative).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I could not confirm that was saved, so I did not mark it as done.

2 · Card shown I could not confirm that the change was saved, so I did not mark it as done.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:32:31.833Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched your saved foods and aliases to "egg" at 90% confidence.
Matched the common-food list to "Egg" at 90% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.4 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould update the entry"Make the gooseberries a half cup." (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. FAIL: Reported could-not-confirm (write-confirmation false negative).

Why verdict Reported could-not-confirm (write-confirmation false negative).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I could not confirm that was saved, so I did not mark it as done.

2 · Card shown I could not confirm that the change was saved, so I did not mark it as done.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:32:43.133Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Gooseberries" at 90% confidence.
Matched your past food logs to "Gooseberries" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.7 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould update the entry"Update the swordfish to six ounces." (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. FAIL: Did not update — misroute-into-food (no verified mutation).

Why verdict Did not update — misroute-into-food (no verified mutation).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I couldn't identify which food entry you meant. Which one would you like to correct?

2 · Card shown I couldn't identify which food entry you meant. Which one would you like to correct?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:32:54.544Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.7 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould update the entry"Change walnut butter to two tablespoons." (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. FAIL: Reported could-not-confirm (write-confirmation false negative).

Why verdict Reported could-not-confirm (write-confirmation false negative).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I could not confirm that was saved, so I did not mark it as done.

2 · Card shown I could not confirm that the change was saved, so I did not mark it as done.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:33:05.711Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched the common-food list to "Walnut butter" at 90% confidence.
Matched your past food logs to "Walnut butter" at 82% confidence.

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.9 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould update the entry"Actually the coconut milk was chocolate, not plain." (difficulty —)2.8s

Verdict Expected UPDATE — should update the entry. PASS: Updated as asked.

Why verdict Updated as asked.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.8s

1 · TTS said Updated to chocolate coconut milk.

2 · Card shown Updated the food entry from "Beverages, Coconut Water, Ready-To-Drink, Unsweetened" to "chocolate coconut milk".

3 · App data rows written updated food_log_entry: Milk ×1 (100 g) 50 cal · 3.3g P · 4.8g C · 2g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:33:19.742Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould update the entry"Actually that was birthday cake, not cookie dough." (difficulty —)2.0s

Verdict Expected UPDATE — should update the entry. FAIL: Did not update (treated as new / not understood).

Why verdict Did not update (treated as new / not understood).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.0s

1 · TTS said Which exact birthday cake should replace Blood orange?

2 · Card shown Which exact birthday cake should replace Blood orange?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: cookie dough item

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-edit-0-cookie-dough-item",
    "name": "cookie dough item",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 180,
      "proteinGrams": 8,
      "fatGrams": 6,
      "carbsGrams": 18
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-edit-0-cookie-dough-item at 2026-07-18T10:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-edit-0-cookie-dough-item",
  "quantity": 1,
  "loggedAt": "2026-07-18T10:00:00.000Z"
}

snapshot captured 2026-08-02T22:33:32.904Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould update the entry"Change the portion to two cups." (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. FAIL: Reported could-not-confirm (write-confirmation false negative).

Why verdict Reported could-not-confirm (write-confirmation false negative).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I could not confirm that was saved, so I did not mark it as done.

2 · Card shown I could not confirm that the change was saved, so I did not mark it as done.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: millet

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-edit-1-millet",
    "name": "millet",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 180,
      "proteinGrams": 8,
      "fatGrams": 6,
      "carbsGrams": 18
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-edit-1-millet at 2026-07-18T10:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-edit-1-millet",
  "quantity": 1,
  "loggedAt": "2026-07-18T10:00:00.000Z"
}

snapshot captured 2026-08-02T22:33:44.181Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.7 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould update the entry"Make it half a cup instead." (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. FAIL: Reported could-not-confirm (write-confirmation false negative).

Why verdict Reported could-not-confirm (write-confirmation false negative).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I could not confirm that was saved, so I did not mark it as done.

2 · Card shown I could not confirm that the change was saved, so I did not mark it as done.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: buckwheat porridge

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-edit-2-buckwheat-porridge",
    "name": "buckwheat porridge",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 180,
      "proteinGrams": 8,
      "fatGrams": 6,
      "carbsGrams": 18
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-edit-2-buckwheat-porridge at 2026-07-18T10:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-edit-2-buckwheat-porridge",
  "quantity": 1,
  "loggedAt": "2026-07-18T10:00:00.000Z"
}

snapshot captured 2026-08-02T22:33:55.536Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.9 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould update the entry"Update that to six ounces." (difficulty —)0.3s

Verdict Expected UPDATE — should update the entry. PASS: Updated as asked.

Why verdict Updated as asked.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.3s

1 · TTS said Updated Blood orange to 6 oz.

2 · Card shown Updated Blood orange to 6 oz.

3 · App data rows written updated food_log_entry: Blood orange ×1.2149785714285712 (140 g) 85 cal · 1.6g P · 22g C · 0.2g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: swordfish

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-edit-3-swordfish",
    "name": "swordfish",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 180,
      "proteinGrams": 8,
      "fatGrams": 6,
      "carbsGrams": 18
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-edit-3-swordfish at 2026-07-18T10:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-edit-3-swordfish",
  "quantity": 1,
  "loggedAt": "2026-07-18T10:00:00.000Z"
}

snapshot captured 2026-08-02T22:34:07.002Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.7 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould update the entry"Actually it was almond butter, not walnut butter." (difficulty —)0.6s

Verdict Expected UPDATE — should update the entry. PASS: Updated as asked.

Why verdict Updated as asked.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.6s

1 · TTS said Updated to almond butter.

2 · Card shown Updated the food entry from "Blood orange" to "almond butter".

3 · App data rows written updated food_log_entry: Almond Butter ×1 (1 tablespoon) 103 cal · 3.3g P · 3.4g C · 8.4g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: walnut butter

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-edit-4-walnut-butter",
    "name": "walnut butter",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 180,
      "proteinGrams": 8,
      "fatGrams": 6,
      "carbsGrams": 18
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-edit-4-walnut-butter at 2026-07-18T10:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-edit-4-walnut-butter",
  "quantity": 1,
  "loggedAt": "2026-07-18T10:00:00.000Z"
}

snapshot captured 2026-08-02T22:34:18.736Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould update the entry"Change the blood orange to a large." (difficulty —)5.0s

Verdict Expected UPDATE — should update the entry. PASS: Updated as asked.

Why verdict Updated as asked.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 5.0s

1 · TTS said Updated Blood orange to 1 large.

2 · Card shown Updated Blood orange to 1 large.

3 · App data rows written updated food_log_entry: Blood orange ×1 (140 g) 70 cal · 1.3g P · 18.1g C · 0.2g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seed food: blood orange

{
  "type": "food_item",
  "foodItem": {
    "id": "seed-g-edit-5-blood-orange",
    "name": "blood orange",
    "brand": null,
    "source": "manual",
    "servingLabel": "serving",
    "servingQuantity": 1,
    "servingUnit": "serving",
    "macrosPerServing": {
      "calories": 180,
      "proteinGrams": 8,
      "fatGrams": 6,
      "carbsGrams": 18
    },
    "aliases": [],
    "timesLogged": 0,
    "createdAt": "2026-07-16T12:00:00.000Z"
  }
}

Seed food history: 1 for seed-g-edit-5-blood-orange at 2026-07-18T10:00:00.000Z

{
  "type": "food_history",
  "foodItemId": "seed-g-edit-5-blood-orange",
  "quantity": 1,
  "loggedAt": "2026-07-18T10:00:00.000Z"
}

snapshot captured 2026-08-02T22:34:41.458Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

UNVERIFIEDunclassifiedshould answer the query"How much protein is on my log today?" (difficulty —)0.1s

Verdict Expected QUERY — should answer the query. UNVERIFIED: answer key has no expected values

Why verdict answer key has no expected values

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I understood the nutrition question, but I could not find a solid number for that food yet. Nothing was logged.

2 · Card shown I understood the nutrition question, but I could not find a solid number for that food yet. Nothing was logged.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

No DB snapshot captured for this utterance; the report cannot prove whether yesterday/personal data existed.

snapshot captured 2026-08-02T22:34:46.072Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 1.9 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

Why the fails happened — comprehension vs execution vs cosmetic

Handled correctly? — by expected action

Accuracy by difficulty

Clarification follow-ups — scored separately

Cosmetic only

System / infra

Latency

Filter — controls the list below

Per-utterance detail