acceptance-corpus-conversational-324f4ab3-c63-20260802

Supposed to	N	Correct	Wrong	Unverified
▸ OTHER — handle appropriately	68	63 (93%)	5 (7%)	0 (0%)
5 handled wrong — click one to jump to its full detail below "it was the the one with the" OTHER ambient_non_command — wrote workout on a non-workout request "they added squats to the class routine" OTHER ambient_non_command — misrouted ambient speech into food/workout log "I read an article about todo apps" OTHER ambient_non_command — performed a command action on non-command speech "I heard deadlifts are good for your back" OTHER ambient_non_command — misrouted ambient speech into food/workout log "I've been meaning to start a workout journal" OTHER ambient_non_command — performed a command action on non-command speech
▸ CLARIFY_ANSWER — complete the pending action	40	9 (23%)	4 (10%)	27 (68%)
4 handled wrong — click one to jump to its full detail below "the grocery list" Reported could-not-confirm (write-confirmation false negative). "mute the voice readback" Did not complete the pending action. "open the workouts tab" Did not complete the pending action. "turn off voice responses" Did not complete the pending action.
▸ CLARIFY — ask a clarifying question	39	34 (87%)	5 (13%)	0 (0%)
5 handled wrong — click one to jump to its full detail below "add a task for tomorrow" Reported could-not-confirm (write-confirmation false negative). "make a note" Logged a BLIND guess — no stated assumption, no correction invited. "jot something down" Logged a BLIND guess — no stated assumption, no correction invited. "save a note about the meeting" Logged a BLIND guess — no stated assumption, no correction invited. "show me the other one" Neither clarified nor logged.
▸ DECLINE — decline	27	26 (96%)	1 (4%)	0 (0%)
1 handled wrong — click one to jump to its full detail below "forget the note" Reported could-not-confirm (write-confirmation false negative).
▸ NAVIGATE — navigate	1	0 (0%)	1 (100%)	0 (0%)
1 handled wrong — click one to jump to its full detail below "sorry I meant the food tab" Did not navigate — misrouted into a food/workout clarification or dead-ended.
Total	175	132 (89%)	16	27

PASSshould ask a clarifying question"log a sandwich" (difficulty —)0.7s

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 0.7s

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

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

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 2026-08-02T21:05:34.596Z

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 some pasta" (difficulty —)3.1s

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 3.1s

1 · TTS said How much some pasta did you have?

2 · Card shown How much some pasta 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 2026-08-02T21:05:48.831Z

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 my lunch" (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 I heard "my lunch" as a personal default or usual item, but I do not have a matching saved food yet. Which exact food and amount should I log?

2 · Card shown I heard "my lunch" as a personal default or usual item, but I do not have a matching saved food yet. Which exact food and amount should I 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 2026-08-02T21:06:00.077Z

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 0.0 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"log a coffee" (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 coffee. Assumed 100 g because no amount was stated.

2 · Card shown Logged a coffee. Assumed 100 g because no amount was stated.

3 · App data rows written 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 2026-08-02T21:06:11.318Z

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 "Coffee" 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 0.0 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 ate a burger" (difficulty —)2.2s

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 2.2s

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

2 · Card shown Which type of burger 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 2026-08-02T21:06:24.655Z

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 fruit" (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 some fruit? 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 some fruit? 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 2026-08-02T21:06:37.073Z

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 dinner from earlier" (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 Coffee. Using your recent Coffee history. Tell me if that is wrong.

2 · Card shown Logged Coffee. Using your recent Coffee history. Tell me if that is wrong.

3 · App data rows written created food_log_entry: Coffee ×0.01 (100 g) 0 cal · 0g 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 2026-08-02T21:06:48.286Z

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 0.0 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 a smoothie" (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 Which type of smoothie was it? There are several with different nutrition.

2 · Card shown Which type of smoothie 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 2026-08-02T21:07:00.437Z

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 snack" (difficulty —)0.8s

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 0.8s

1 · TTS said What exact food and amount should I use for a snack? 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 snack? 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 2026-08-02T21:07:12.407Z

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 drank something" (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 What did you drink?

2 · Card shown What did you drink?

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 2026-08-02T21:07:24.151Z

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 eggs this morning" (difficulty —)2.0s

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 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: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:07:37.302Z

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"log the salad I just ate" (difficulty —)2.6s

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 2.6s

1 · TTS said Did the salad have dressing or anything calorie-heavy on it? I did not log it yet because add-ons can change the nutrition.

2 · Card shown Did the salad have dressing or anything calorie-heavy on it? 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 2026-08-02T21:07:51.048Z

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 some squats" (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 Which exercise, weight, and reps should I log?

2 · Card shown Which exercise, weight, and reps should I 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 2026-08-02T21:08:02.246Z

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 exercise list to "squats" at 88% 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 did bench press" (difficulty —)0.0s

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 0.0s

1 · TTS said Which exercise, weight, and reps should I log?

2 · Card shown Which exercise, weight, and reps should I 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 2026-08-02T21:08:13.437Z

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 exercise list to "Bench Press" at 88% 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"track my run" (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 my run? 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 my run? 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 2026-08-02T21:08:27.236Z

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 deadlifts" (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 Which exercise, weight, and reps should I log?

2 · Card shown Which exercise, weight, and reps should I 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 2026-08-02T21:08:38.425Z

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 exercise list to "deadlifts" at 88% 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 did some curls" (difficulty —)2.6s

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 2.6s

1 · TTS said What weight and how many reps for the curls? I'll help you log it once I have those details.

2 · Card shown What weight and how many reps for the curls? I'll help you log it once I have those details.

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 2026-08-02T21:08:52.154Z

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"log cardio" (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 What exact food and amount should I use for cardio? 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 cardio? 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 2026-08-02T21:09:03.907Z

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"did pull-ups" (difficulty —)0.0s

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 0.0s

1 · TTS said Which exercise, weight, and reps should I log?

2 · Card shown Which exercise, weight, and reps should I 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 2026-08-02T21:09:15.069Z

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 exercise list to "pull ups" at 88% 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"log my leg day" (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 my leg day? 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 my leg day? 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 2026-08-02T21:09:28.914Z

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 benched earlier" (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 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 2026-08-02T21:09:51.576Z

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"track the set I just did" (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 the set i just did? 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 the set i just did? 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 2026-08-02T21:10:05.443Z

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 milk" (difficulty —)1.4s

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 1.4s

1 · TTS said How much milk did you have?

2 · Card shown How much milk 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 2026-08-02T21:10:17.965Z

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"put batteries on the list" (difficulty —)0.1s

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 0.1s

1 · TTS said Which list should I add batteries to?

2 · Card shown Which list should I add batteries to?

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 2026-08-02T21:10:29.152Z

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"remind me about that thing" (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 What should the reminder be about?

2 · Card shown What should the reminder be about?

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 2026-08-02T21:10:40.342Z

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 ask a clarifying question"add a task for tomorrow" (difficulty —)0.1s

Verdict Expected CLARIFY — should ask a clarifying question. 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'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or a to-do.

2 · Card shown I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or 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 2026-08-02T21:10:51.526Z

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 ask a clarifying question"make a note" (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 Logged your note in General about for tomorrow.

2 · Card shown Logged your note in General about for tomorrow.

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 2026-08-02T21:11:02.858Z

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 ask a clarifying question"jot something down" (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 Logged your note in General about something down.

2 · Card shown Logged your note in General about something down.

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 2026-08-02T21:11:14.219Z

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 it to my list" (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 Which list should I add that to, and what is the item called?

2 · Card shown Which list should I add that to, and what is the item called?

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 2026-08-02T21:11:25.452Z

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"create a todo" (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 What should the to-do say?

2 · Card shown What should the to-do say?

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 2026-08-02T21:11:36.651Z

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 ask a clarifying question"save a note about the meeting" (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 Logged your note in General about about the meeting.

2 · Card shown Logged your note in General about about the meeting.

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 2026-08-02T21:11:48.024Z

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 eggs and 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 Egg and Butter? I did not log them yet because the amounts were not clear.

2 · Card shown How much should I log for Egg and 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 2026-08-02T21:11:59.574Z

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 "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 0.1 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"change my settings" (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 Which setting would you like to change?

2 · Card shown Which setting would you like to change?

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 2026-08-02T21:12:10.771Z

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"turn something off" (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 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 2026-08-02T21:12:21.995Z

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"update my preferences" (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 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 2026-08-02T21:12:33.208Z

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"make it quieter" (difficulty —)5.5s

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 5.5s

1 · TTS said Which exact make it quieter should replace Coffee?

2 · Card shown Which exact make it quieter should replace Coffee?

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 2026-08-02T21:12:49.840Z

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"go there" (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 What would you like me to do with that?

2 · Card shown 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 2026-08-02T21:13:01.157Z

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"open that screen" (difficulty —)0.0s

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 0.0s

1 · TTS said What would you like me to do with that?

2 · Card shown 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 2026-08-02T21:13:13.098Z

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"take me back" (difficulty —)0.0s

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 0.0s

1 · TTS said What would you like me to do with that?

2 · Card shown 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 2026-08-02T21:13:23.791Z

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 ask a clarifying question"show me the other one" (difficulty —)0.9s

Verdict Expected CLARIFY — should ask a clarifying question. FAIL: Neither clarified nor logged.

Why verdict Neither clarified nor logged.

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 0.9s

1 · TTS said I can’t show the other one yet. You can tell me to log food, workout, a note, or a to-do, and I can help guide you to the right place.

2 · Card shown I can’t show the other one yet. You can tell me to log food, workout, a note, or a to-do, and I can help guide you to the right place.

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 2026-08-02T21:13:35.895Z

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"the rye bread one" (difficulty —)5.8s

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 5.8s

1 · TTS said How much the rye bread one did you have?

2 · Card shown How much the rye bread one 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 2026-08-02T21:14:17.671Z

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"uh three sets of ten at ninety pounds" (difficulty —)2.3s

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 2.3s

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 2026-08-02T21:14:42.235Z

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"Chobani vanilla nonfat one cup" (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 I couldn't convert one cup of Vanilla Zero Sugar Nonfat Greek Yogurt 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 Vanilla Zero Sugar Nonfat Greek Yogurt 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 2026-08-02T21:15:05.206Z

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"the grocery list" (difficulty —)0.1s

Verdict Expected CLARIFY_ANSWER — should complete the pending action. 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'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or a to-do.

2 · Card shown I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or 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 2026-08-02T21:15:27.493Z

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"call it dentist follow-up for next Tuesday" (difficulty —)3.6s

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 3.6s

1 · TTS said What would you like me to do with that?

2 · Card shown 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 2026-08-02T21:15:53.339Z

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"pack the charger and passport" (difficulty —)2.9s

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 2.9s

1 · TTS said What exact food and amount should I use for pack the charger and passport? 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 pack the charger and passport? 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 2026-08-02T21:16:18.469Z

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 complete the pending action"large oat milk latte no sugar" (difficulty —)0.1s

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 0.1s

1 · TTS said Logged Coffee and Oat milk latte.

2 · Card shown Logged Coffee and Oat milk latte.

3 · App data rows written created food_log_entry: Coffee ×1 (100 g) 1 cal · 0.1g P · 0g C · 0g F
created food_log_entry: Oat milk latte ×1 (355 g) 135 cal · 4.3g P · 18.5g C · 5.3g 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 2026-08-02T21:16:43.556Z

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 "Oat milk latte" 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 complete the pending action"four sets of eight at one thirty five" (difficulty —)0.1s

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 0.1s

1 · TTS said Logged thirty: 1 lb for 8 reps.

2 · Card shown Logged thirty: 1 lb for 8 reps.

3 · App data rows written created workout_set:

4 · UI did navigated to "workouts" 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 2026-08-02T21:17:06.018Z

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 complete the pending action"put batteries on the list" (difficulty —)0.1s

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 0.1s

1 · TTS said Which list should I add batteries to?

2 · Card shown Which list should I add batteries to?

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 2026-08-02T21:17:17.247Z

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 complete the pending action"mute the voice readback" (difficulty —)0.1s

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 0.1s

1 · TTS said Voice readback is off.

2 · Card shown Voice readback is off.

4 · UI did navigated to "settings" 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 2026-08-02T21:17:50.918Z

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"no I meant the other list the camping one" (difficulty —)2.2s

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 2.2s

1 · TTS said Which list should I add "batteries" to?

2 · Card shown Which list should I add "batteries" to?

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 2026-08-02T21:18:15.308Z

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 complete the pending action"actually the whole wheat turkey one" (difficulty —)3.1s

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 3.1s

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.

3 · App data rows written update_sync_failed food_log_entry: Oat milk latte

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 2026-08-02T21:18:44.009Z

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"Add finish the quarterly report to my to-do list." (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 I could not find a to-do matching quarterly report to my.

2 · Card shown I could not find a to-do matching quarterly report to my.

4 · UI did navigated to "todos" 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 2026-08-02T21:18:55.214Z

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 navigate"sorry I meant the food tab" (difficulty —)3.2s

Verdict Expected NAVIGATE — should navigate. FAIL: Did not navigate — misrouted into a food/workout clarification or dead-ended.

Why verdict Did not navigate — misrouted into a food/workout clarification or dead-ended.

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 What would you like me to do with that?

2 · Card shown 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 2026-08-02T21:19:20.573Z

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"I had one cup of cereal." (difficulty —)0.7s

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 0.7s

1 · TTS said Which one did you mean? 1. Cereal, Granola, 2. Cereal, O's, Honey Nut, 3. Cereal, Other, Peanut Butter

2 · Card shown Which one did you mean? 1. Cereal, Granola, 2. Cereal, O's, Honey Nut, 3. Cereal, Other, Peanut Butter

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 2026-08-02T21:19:32.469Z

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"like um grilled four ounces maybe" (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 I need to resolve what portion before I log this meal. What should I use for what portion?

2 · Card shown I need to resolve what portion before I log this meal. What should I use for what portion?

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 2026-08-02T21:20:11.841Z

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"I had a latte." (difficulty —)0.2s

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 0.2s

1 · TTS said What exact food and amount should I use for a latte? 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 latte? 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 2026-08-02T21:20:23.208Z

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"two slices pepperoni from Domino's" (difficulty —)2.7s

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 2.7s

1 · TTS said What exact food and amount should I use for two slices pepperoni from dominos? 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 two slices pepperoni from dominos? 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 2026-08-02T21:21:01.861Z

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"five kilometers in twenty eight minutes" (difficulty —)3.5s

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 3.5s

1 · TTS said What would you like me to do with that?

2 · Card shown 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 2026-08-02T21:21:30.412Z

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"Fairlife chocolate shake one bottle" (difficulty —)0.7s

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 0.7s

1 · TTS said Which one did you mean? 1. Nutrition Plan Chocolate Shake — Fairlife, 2. Chocolate Nutrition Shake — FAIRLIFE, 3. Elite High Complete Protein 42 G Milk Shake, Chocolate — FAIRLIFE

2 · Card shown Which one did you mean? 1. Nutrition Plan Chocolate Shake — Fairlife, 2. Chocolate Nutrition Shake — FAIRLIFE, 3. Elite High Complete Protein 42 G Milk Shake, Chocolate — FAIRLIFE

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 2026-08-02T21:21:57.016Z

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"three sets twelve reps fifty pound dumbbells" (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 Which exercise, weight, and reps should I log?

2 · Card shown Which exercise, weight, and reps should I 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 2026-08-02T21:22:30.686Z

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 complete the pending action"Make a note that my passport expires in November." (difficulty —)0.2s

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 0.2s

1 · TTS said Logged your note in General about my passport expires in November.

2 · Card shown Logged your note in General about my passport expires in November.

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 2026-08-02T21:22:42.028Z

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"errands category due Friday" (difficulty —)2.0s

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 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: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:23:17.573Z

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"thirty minutes on the elliptical" (difficulty —)2.4s

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 2.4s

1 · TTS said Great — I’ll log 30 minutes on the elliptical as cardio. You can check your workout summary in the app after it syncs. Is there anything else you’d like to track right now?

2 · Card shown Great — I’ll log 30 minutes on the elliptical as cardio. You can check your workout summary in the app after it syncs. Is there anything else you’d like to track right now?

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 2026-08-02T21:23:44.028Z

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 complete the pending action"Dave's killer bread two slices" (difficulty —)0.9s

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 0.9s

1 · TTS said Got it, I logged Organic Bread 21 Whole Grains and Seeds — Dave's Killer Bread.

2 · Card shown I logged Organic Bread 21 Whole Grains and Seeds — Dave's Killer Bread — let me know if that's not right.

3 · App data rows written created food_log_entry: Organic Bread 21 Whole Grains and Seeds ×2 (slice) 220 cal · 10g P · 44g C · 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 2026-08-02T21:24:07.158Z

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"open the workouts tab" (difficulty —)0.1s

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 0.1s

1 · TTS said Opening Workouts.

2 · Card shown Opening Workouts.

4 · UI did navigated to "workouts" 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 2026-08-02T21:24:31.932Z

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 complete the pending action"turn off voice responses" (difficulty —)0.1s

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 0.1s

1 · TTS said Voice readback is off.

2 · Card shown Voice readback is off.

4 · UI did navigated to "settings" 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 2026-08-02T21:24:56.126Z

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"spinach feta wrap from Starbucks" (difficulty —)1.9s

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 1.9s

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 2026-08-02T21:25:21.885Z

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 complete the pending action"leg press three sets fifteen at two seventy" (difficulty —)0.3s

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 0.3s

1 · TTS said Logged all 3 sets of leg press at 270 lb: 15 reps, 15 reps, and 15 reps.

2 · Card shown Logged all 3 sets of leg press at 270 lb: 15 reps, 15 reps, and 15 reps.

3 · App data rows written created workout_set:
created workout_set:
created workout_set:

4 · UI did navigated to "workouts" 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 2026-08-02T21:25:44.443Z

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 exercise list to "Leg Press" at 88% 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 complete the pending action"Add fertilizer to a list." (difficulty —)0.1s

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 0.1s

1 · TTS said Which list should I add fertilizer to?

2 · Card shown Which list should I add fertilizer to?

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 2026-08-02T21:25:55.610Z

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"just the regular one" (difficulty —)1.8s

Verdict Expected CLARIFY_ANSWER — should complete the pending action. UNVERIFIED: Informational non-gating row (excluded from pass/fail tallies).

Why verdict Informational non-gating row (excluded from pass/fail tallies).

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 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 2026-08-02T21:26:30.929Z

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 0.0 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)

UNVERIFIEDunclassifiedshould complete the pending action"a few sets" (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 2026-08-02T21:26:53.160Z

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"the usual list" (difficulty —)2.7s

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 2.7s

1 · TTS said What exact food and amount should I use for the usual list? 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 the usual list? 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 2026-08-02T21:27:18.078Z

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"something about the trip" (difficulty —)2.6s

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 2.6s

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 2026-08-02T21:27:42.882Z

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"later this week" (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 2026-08-02T21:28:05.118Z

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"bodyweight three sets to failure" (difficulty —)3.7s

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 3.7s

1 · TTS said Got it. For bodyweight exercises, you can log each set by saying the exercise name, number of reps, and "to failure" for each set. What exercise are you doing?

2 · Card shown Got it. For bodyweight exercises, you can log each set by saying the exercise name, number of reps, and "to failure" for each set. What exercise are you doing?

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 2026-08-02T21:28:31.006Z

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 complete the pending action"scrambled two large eggs" (difficulty —)0.3s

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 0.3s

1 · TTS said Logged scrambled two large eggs.

2 · Card shown Logged scrambled two large eggs.

3 · App data rows written created food_log_entry: Egg ×1 (2 eggs) 143 cal · 12.6g P · 0.7g C · 9.5g 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 2026-08-02T21:28:53.556Z

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 "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 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"greek salad no olives" (difficulty —)1.7s

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 1.7s

1 · TTS said I heard "greek salad no olives." For the usual list, you can log Greek salad (no olives) with an amount like "one bowl" or "one serving." What amount should I use?

2 · Card shown I heard "greek salad no olives." For the usual list, you can log Greek salad (no olives) with an amount like "one bowl" or "one serving." What amount should I use?

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 2026-08-02T21:29:19.096Z

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"the packing list for the beach trip" (difficulty —)5.1s

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 5.1s

1 · TTS said What would you like me to do with that?

2 · Card shown 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 2026-08-02T21:29:46.368Z

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"personal category call mom Sunday" (difficulty —)2.1s

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 2.1s

1 · TTS said What would you like me to do with that?

2 · Card shown 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 2026-08-02T21:30:10.683Z

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 decline"never mind" (difficulty —)0.1s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 2026-08-02T21:30:33.343Z

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 decline"no cancel that" (difficulty —)0.0s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 2026-08-02T21:30:44.510Z

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 decline"forget it" (difficulty —)0.0s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 2026-08-02T21:30:56.383Z

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 decline"actually skip it" (difficulty —)0.0s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 2026-08-02T21:31:06.841Z

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 decline"nah leave it" (difficulty —)0.0s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 2026-08-02T21:31:29.304Z

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 decline"scratch that whole thing" (difficulty —)3.5s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 3.5s

1 · TTS said Understood. I cleared that from our conversation. What would you like to do instead?

2 · Card shown Understood. I cleared that from our conversation. What would you like to do 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 2026-08-02T21:31:44.014Z

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 decline"wait no stop" (difficulty —)0.1s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 setting would you like to change?

2 · Card shown Which setting would you like to change?

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 2026-08-02T21:31:55.221Z

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 decline"I changed my mind don't save anything" (difficulty —)1.5s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 Understood. I cleared that from our conversation. What would you like to do instead?

2 · Card shown Understood. I cleared that from our conversation. What would you like to do 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 2026-08-02T21:32:07.900Z

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 decline"cancel whatever you were doing" (difficulty —)1.5s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 Understood. I cleared that from our conversation. What would you like to do instead?

2 · Card shown Understood. I cleared that from our conversation. What would you like to do 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 2026-08-02T21:32:20.590Z

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 decline"don't bother" (difficulty —)0.0s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 2026-08-02T21:32:31.745Z

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 decline"leave it alone" (difficulty —)0.0s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 2026-08-02T21:32:42.909Z

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 decline"not anymore forget the whole thing" (difficulty —)1.5s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 Understood. I cleared that from our conversation. What would you like to do instead?

2 · Card shown Understood. I cleared that from our conversation. What would you like to do 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 2026-08-02T21:32:55.582Z

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 decline"uh no don't do that" (difficulty —)2.0s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 Understood. I cleared that from our conversation. What would you like to do instead?

2 · Card shown Understood. I cleared that from our conversation. What would you like to do 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 2026-08-02T21:33:08.715Z

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 decline"yeah no never mind about that" (difficulty —)2.3s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 Understood. I cleared that from our conversation. What would you like to do instead?

2 · Card shown Understood. I cleared that from our conversation. What would you like to do 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 2026-08-02T21:33:22.297Z

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 decline"skip it I'm good" (difficulty —)1.4s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 Understood. Let me know if you need anything else.

2 · Card shown Understood. Let me know if you need anything else.

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 2026-08-02T21:33:34.871Z

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 decline"don't add that after all" (difficulty —)1.3s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 Understood. I cleared that from our conversation. What would you like to do instead?

2 · Card shown Understood. I cleared that from our conversation. What would you like to do 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 2026-08-02T21:33:47.331Z

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 decline"no thanks I'm done" (difficulty —)1.3s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 Understood. Have a great day.

2 · Card shown Understood. Have a great day.

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 2026-08-02T21:33:59.735Z

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 decline"hold on cancel" (difficulty —)0.1s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 2026-08-02T21:34:10.942Z

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 decline"actually don't" (difficulty —)0.0s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 2026-08-02T21:34:22.270Z

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 decline"never mind about the sandwich" (difficulty —)1.4s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 Understood. Let me know if you need anything else.

2 · Card shown Understood. Let me know if you need anything else.

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 2026-08-02T21:34:59.366Z

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 decline"no skip the list thing" (difficulty —)0.7s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 Understood. What would you like to do instead?

2 · Card shown Understood. What would you like to do 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 2026-08-02T21:35:44.801Z

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 decline"forget the note" (difficulty —)0.1s

Verdict Expected DECLINE — should decline. 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'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or a to-do.

2 · Card shown I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or 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 2026-08-02T21:36:07.059Z

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 decline"don't create that todo" (difficulty —)0.1s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 Okay — I won't create that to-do.

2 · Card shown Okay — I won't create that to-do.

4 · UI did navigated to "todos" 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 2026-08-02T21:36:29.383Z

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 decline"cancel the food entry" (difficulty —)2.4s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 2.4s

1 · TTS said What would you like me to do with that?

2 · Card shown 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 2026-08-02T21:36:54.044Z

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 decline"nah don't save the workout" (difficulty —)1.9s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 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 2026-08-02T21:37:18.134Z

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 decline"don't change my settings" (difficulty —)0.0s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 Which setting would you like to change?

2 · Card shown Which setting would you like to change?

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 2026-08-02T21:38:02.752Z

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 decline"forget going there" (difficulty —)0.1s

Verdict Expected DECLINE — should decline. PASS: Declined / no unwanted write.

Why verdict Declined / no unwanted write.

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 2026-08-02T21:38:27.353Z

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 handle appropriately"okay so anyway she said she'd call back tomorrow" (difficulty —)1.8s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Got it — sounds like you’re expecting a call tomorrow. If you want to set a reminder or add a to-do for that, just let me know.

2 · Card shown Got it — sounds like you’re expecting a call tomorrow. If you want to set a reminder or add a to-do for that, just let me know.

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 2026-08-02T21:38:40.293Z

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 handle appropriately"yeah I know but like that's not even the point" (difficulty —)2.6s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 2.6s

1 · TTS said Got it — I’m here to listen. What’s on your mind?

2 · Card shown Got it — I’m here to listen. What’s on your mind?

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 2026-08-02T21:38:54.061Z

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 handle appropriately"so then he goes and tells me the whole story" (difficulty —)1.7s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Got it — I’m here to listen. What’s on your mind?

2 · Card shown Got it — I’m here to listen. What’s on your mind?

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 2026-08-02T21:39:06.940Z

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 handle appropriately"mm-hmm right right I hear you" (difficulty —)1.5s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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’m here if you want to talk or need help with something specific.

2 · Card shown I’m here if you want to talk or need help with something specific.

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 2026-08-02T21:39:19.582Z

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 handle appropriately"honestly I don't even know what to tell her" (difficulty —)1.8s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 It sounds like you're in a tough spot with someone. Want to talk it through, or would it help to write it down first as a note?

2 · Card shown It sounds like you're in a tough spot with someone. Want to talk it through, or would it help to write it down first as a note?

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 2026-08-02T21:39:32.515Z

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 handle appropriately"wait what did you just say about the meeting" (difficulty —)0.0s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 It sounds like you're in a tough spot with someone. Want to talk it through, or would it help to write it down first as a note?

2 · Card shown It sounds like you're in a tough spot with someone. Want to talk it through, or would it help to write it down first as a note?

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 2026-08-02T21:39:43.686Z

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 handle appropriately"no yeah totally I was thinking the same thing" (difficulty —)0.8s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Got it. If you ever want to turn that thought into an action — like adding it to your notes or starting a list — just tell me.

2 · Card shown Got it. If you ever want to turn that thought into an action — like adding it to your notes or starting a list — just tell me.

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 2026-08-02T21:39:55.650Z

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 handle appropriately"and then we ended up just staying in" (difficulty —)2.3s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 — sounds like a low-key evening. If you want to log that as a note or reflect on how you felt about staying in, I’m here to help.

2 · Card shown Got it — sounds like a low-key evening. If you want to log that as a note or reflect on how you felt about staying in, I’m here to help.

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 2026-08-02T21:40:10.557Z

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 handle appropriately"I mean it's fine it's whatever" (difficulty —)1.9s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Alright — I hear you. If you want to log it as a note or move on to something else, just say the word.

2 · Card shown Alright — I hear you. If you want to log it as a note or move on to something else, just say the word.

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 2026-08-02T21:40:22.996Z

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 handle appropriately"so basically the whole thing got pushed to Monday" (difficulty —)2.2s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 2.2s

1 · TTS said Got it — so the timeline shifted. If you want to set a reminder or add a to-do for Monday, just let me know.

2 · Card shown Got it — so the timeline shifted. If you want to set a reminder or add a to-do for Monday, just let me know.

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 2026-08-02T21:40:36.374Z

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 handle appropriately"babe what time are we leaving" (difficulty —)2.3s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 — if you want to set a reminder or add a to-do for your departure time, just say “add a to-do” or “set a reminder” and I’ll help you do that.

2 · Card shown Got it — if you want to set a reminder or add a to-do for your departure time, just say “add a to-do” or “set a reminder” and I’ll help you do 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 2026-08-02T21:41:23.862Z

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 handle appropriately"mom I'll be down in five minutes" (difficulty —)1.9s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Got it — if you want to set a reminder or add a to-do for that, just let me know.

2 · Card shown Got it — if you want to set a reminder or add a to-do for that, just let me know.

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 2026-08-02T21:41:36.833Z

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 handle appropriately"dude did you see that play" (difficulty —)2.0s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 didn’t catch it — what play are you talking about?

2 · Card shown I didn’t catch it — what play are you talking about?

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 2026-08-02T21:41:50.002Z

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 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 handle appropriately"can someone turn down the TV" (difficulty —)1.7s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 I couldn't find Someone Turn Down The TV 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 Someone Turn Down The TV 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 2026-08-02T21:42:03.006Z

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 handle appropriately"hey are you almost ready" (difficulty —)2.2s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 2.2s

1 · TTS said I don’t see a departure time set. Want to set a reminder or add a to-do for when you need to leave?

2 · Card shown I don’t see a departure time set. Want to set a reminder or add a to-do for when you need to leave?

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 2026-08-02T21:42:16.327Z

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 handle appropriately"tell Jake I'll text him back later" (difficulty —)3.0s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Got it. If you want to set a reminder to text Jake later, just say "set a reminder" and I'll help.

2 · Card shown Got it. If you want to set a reminder to text Jake later, just say "set a reminder" and I'll help.

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 2026-08-02T21:42:30.435Z

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 handle appropriately"and the winning team takes home the championship" (difficulty —)3.4s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 3.4s

1 · TTS said Got it — sounds like a big moment. If you want to log it as a note or set a reminder to celebrate later, just say the word.

2 · Card shown Got it — sounds like a big moment. If you want to log it as a note or set a reminder to celebrate later, just say the word.

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 2026-08-02T21:42:44.880Z

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 handle appropriately"breaking news tonight in the capital" (difficulty —)1.8s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 heard that — if you want to save a note about it or set a reminder to look into it later, just let me know.

2 · Card shown I heard that — if you want to save a note about it or set a reminder to look into it later, just let me know.

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 2026-08-02T21:42:57.802Z

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 handle appropriately"your local forecast shows rain tomorrow" (difficulty —)1.5s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Got it — if you want to set a reminder to bring an umbrella or pack rain gear, just say “set a reminder” and I’ll help you add that.

2 · Card shown Got it — if you want to set a reminder to bring an umbrella or pack rain gear, just say “set a reminder” and I’ll help you add 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 2026-08-02T21:43:10.364Z

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 handle appropriately"sponsored by progressive insurance" (difficulty —)1.9s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Got it — if you’d like to save a note about Progressive Insurance or set a reminder to review your policy later, just say “add a note” or “set a reminder” and I’ll help.

2 · Card shown Got it — if you’d like to save a note about Progressive Insurance or set a reminder to review your policy later, just say “add a note” or “set a reminder” and I’ll help.

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 2026-08-02T21:43:23.525Z

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 handle appropriately"tune in next week for the season finale" (difficulty —)2.0s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Got it — sounds like a show you're following. If you want to set a reminder for when it airs, just say "set a reminder" and I’ll help.

2 · Card shown Got it — sounds like a show you're following. If you want to set a reminder for when it airs, just say "set a reminder" and I’ll help.

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 2026-08-02T21:43:36.855Z

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 handle appropriately"the Dow Jones closed up two hundred points" (difficulty —)1.8s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Got it — if you want to add a note about the market or set a reminder to check your portfolio later, just say “add a note” or “set a reminder” and I’ll help.

2 · Card shown Got it — if you want to add a note about the market or set a reminder to check your portfolio later, just say “add a note” or “set a reminder” and I’ll help.

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 2026-08-02T21:43:49.744Z

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 handle appropriately"where did I put my keys" (difficulty —)2.1s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 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 2026-08-02T21:44:02.955Z

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 handle appropriately"okay think think what was I doing" (difficulty —)3.5s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 3.5s

1 · TTS said What would you like me to do with that?

2 · Card shown 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 2026-08-02T21:44:17.572Z

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 handle appropriately"man I'm tired today" (difficulty —)3.1s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 3.1s

1 · TTS said I hear you. If you’d like to log how you’re feeling or set a reminder to rest later, just say the word.

2 · Card shown I hear you. If you’d like to log how you’re feeling or set a reminder to rest later, just say the word.

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 2026-08-02T21:44:31.806Z

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 handle appropriately"gotta remember to email Sarah" (difficulty —)1.9s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Got it. If you want to set a reminder to email Sarah, just say "set a reminder" and I'll help you add it.

2 · Card shown Got it. If you want to set a reminder to email Sarah, just say "set a reminder" and I'll help you add it.

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 2026-08-02T21:44:44.862Z

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 handle appropriately"why is it so hot in here" (difficulty —)2.6s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 2.6s

1 · TTS said I don’t control your environment, but I can help you log how you’re feeling or set a reminder to check the thermostat later if you’d like.

2 · Card shown I don’t control your environment, but I can help you log how you’re feeling or set a reminder to check the thermostat later if you’d like.

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 2026-08-02T21:44:58.595Z

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 handle appropriately"I should probably stretch later" (difficulty —)1.8s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Got it. If you’d like to set a reminder to stretch later, just say “set a reminder” and I’ll help you add it.

2 · Card shown Got it. If you’d like to set a reminder to stretch later, just say “set a reminder” and I’ll help you add it.

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 2026-08-02T21:45:11.597Z

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 handle appropriately"so I was going to" (difficulty —)0.1s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or a to-do.

2 · Card shown I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or 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 2026-08-02T21:45:35.137Z

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 handle appropriately"and then I think" (difficulty —)2.2s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 2.2s

1 · TTS said I’m still here. What would you like to add or do next?

2 · Card shown I’m still here. What would you like to add or do next?

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 2026-08-02T21:45:48.504Z

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 handle appropriately"if we could just" (difficulty —)3.0s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 I didn’t catch the end of that — could you finish your thought?

2 · Card shown I didn’t catch the end of that — could you finish your thought?

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 2026-08-02T21:46:02.705Z

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 handle appropriately"the thing about it is" (difficulty —)1.7s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 I’m listening — go ahead and finish your thought.

2 · Card shown I’m listening — go ahead and finish your thought.

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 2026-08-02T21:46:15.558Z

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 handle appropriately"I was gonna say" (difficulty —)0.0s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or a to-do.

2 · Card shown I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or 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 2026-08-02T21:46:26.720Z

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 handle appropriately"the um the the with the" (difficulty —)1.6s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 1.6s

1 · TTS said I’m still listening — go ahead and finish your thought.

2 · Card shown I’m still listening — go ahead and finish your thought.

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 2026-08-02T21:46:39.506Z

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 0.0 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 handle appropriately"so like uh yeah mm" (difficulty —)1.7s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 I’m here. Take your time — what’s on your mind?

2 · Card shown I’m here. Take your time — what’s on your mind?

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 2026-08-02T21:46:52.356Z

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 handle appropriately"it was the the one with the" (difficulty —)0.7s

Verdict Expected OTHER — should handle appropriately. FAIL: OTHER ambient_non_command — wrote workout on a non-workout request

Why verdict OTHER ambient_non_command — wrote workout on a non-workout request

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 Updated that workout set.

2 · Card shown Updated that workout set.

3 · App data rows written updated workout_set:

4 · UI did navigated to "workouts" 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 2026-08-02T21:47:08.161Z

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 0.0 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 handle appropriately"uh huh yeah so um" (difficulty —)0.6s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 I’m here. Take your time — what’s on your mind?

2 · Card shown I’m here. Take your time — what’s on your mind?

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 2026-08-02T21:47:19.762Z

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 handle appropriately"the thingy with the stuff" (difficulty —)0.8s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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’m not sure what you’re referring to — could you tell me more about the “thingy” or the “stuff” you want to log or do?

2 · Card shown I’m not sure what you’re referring to — could you tell me more about the “thingy” or the “stuff” you want to log or 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 2026-08-02T21:47:31.718Z

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 workout equipment list to "the" 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 handle appropriately"hmm" (difficulty —)0.1s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 2026-08-02T21:47:42.902Z

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 0.0 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 handle appropriately"..." (difficulty —)1.5s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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'm here. Take your time — what’s on your mind?

2 · Card shown I'm here. Take your time — what’s on your mind?

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 2026-08-02T21:47:55.537Z

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 0.0 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 handle appropriately"anyway where was I" (difficulty —)1.9s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 You were in the middle of something. I'm here to help you get back on track. Would you like to continue logging your workout, or was it something else?

2 · Card shown You were in the middle of something. I'm here to help you get back on track. Would you like to continue logging your workout, or was it something else?

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 2026-08-02T21:48:08.600Z

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 handle appropriately"oh that's interesting" (difficulty —)0.0s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 2026-08-02T21:48:19.761Z

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 handle appropriately"yeah no I get it" (difficulty —)2.1s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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. If you’re ready to log something or set a goal, just say the word.

2 · Card shown Got it. If you’re ready to log something or set a goal, just say the word.

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 2026-08-02T21:48:33.002Z

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 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 handle appropriately"I should really log my food more often" (difficulty —)2.8s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 What exact food and amount should I use for i should really log my food more often? 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 i should really log my food more often? 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 2026-08-02T21:48:46.885Z

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 handle appropriately"my brother started a new workout program" (difficulty —)2.5s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 That sounds like a great step for him. If you'd like to log a workout yourself, just say what you did, like "log 3 sets of squats."

2 · Card shown That sounds like a great step for him. If you'd like to log a workout yourself, just say what you did, like "log 3 sets of squats."

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 2026-08-02T21:49:00.917Z

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 workout equipment list to "the" 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 handle appropriately"she keeps a list of everything she buys" (difficulty —)3.2s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Got it. If you’d like to keep your own lists in the app — like a shopping list or a running to-do — just tell me what to add, like “add eggs to my shopping list.”

2 · Card shown Got it. If you’d like to keep your own lists in the app — like a shopping list or a running to-do — just tell me what to add, like “add eggs to my shopping list.”

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 2026-08-02T21:49:14.900Z

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 handle appropriately"I wonder how many calories are in that" (difficulty —)0.1s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 2026-08-02T21:49:26.055Z

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 handle appropriately"my trainer says I need more protein" (difficulty —)3.2s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Good advice. Protein supports muscle repair and recovery. If you want to track your protein, you can log your meals by saying something like “log 2 eggs for breakfast.”

2 · Card shown Good advice. Protein supports muscle repair and recovery. If you want to track your protein, you can log your meals by saying something like “log 2 eggs for breakfast.”

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 2026-08-02T21:49:40.796Z

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 handle appropriately"I always forget to write stuff down" (difficulty —)2.4s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 2.4s

1 · TTS said I get it — memory can be tricky. That’s why voice logging can help. If you want to remember something now, just tell me what to log — like “log a note about my headache” or “add ‘call mom’ to my to-do list.”

2 · Card shown I get it — memory can be tricky. That’s why voice logging can help. If you want to remember something now, just tell me what to log — like “log a note about my headache” or “add ‘call mom’ to my to-do list.”

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 2026-08-02T21:50:05.851Z

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 handle appropriately"we made a grocery list on paper" (difficulty —)0.1s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or a to-do.

2 · Card shown I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or 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 2026-08-02T21:50:17.041Z

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 handle appropriately"he logged into the app yesterday" (difficulty —)2.7s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 he logged into the app yesterday? 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 he logged into the app yesterday? 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 2026-08-02T21:50:30.851Z

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 handle appropriately"I need to remember to call the dentist" (difficulty —)0.0s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or a to-do.

2 · Card shown I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or 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 2026-08-02T21:50:42.021Z

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 handle appropriately"they added squats to the class routine" (difficulty —)0.1s

Verdict Expected OTHER — should handle appropriately. FAIL: OTHER ambient_non_command — misrouted ambient speech into food/workout log

Why verdict OTHER ambient_non_command — misrouted ambient speech into food/workout log

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 exercise, weight, and reps should I log?

2 · Card shown Which exercise, weight, and reps should I 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 2026-08-02T21:50:53.277Z

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 exercise list to "squats" at 88% 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 handle appropriately"my wife tracks her meals in a notebook" (difficulty —)2.7s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 That’s a thoughtful approach. If you’d like to keep your own logs in this app too, just tell me what you want to track — like “log chicken and rice for lunch” or “log 3 sets of push-ups.”

2 · Card shown That’s a thoughtful approach. If you’d like to keep your own logs in this app too, just tell me what you want to track — like “log chicken and rice for lunch” or “log 3 sets of push-ups.”

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 2026-08-02T21:51:07.111Z

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 handle appropriately"I read an article about todo apps" (difficulty —)0.1s

Verdict Expected OTHER — should handle appropriately. FAIL: OTHER ambient_non_command — performed a command action on non-command speech

Why verdict OTHER ambient_non_command — performed a command action on non-command speech

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 You do not have any open to-dos.

2 · Card shown You do not have any open to-dos.

4 · UI did navigated to "todos" 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 2026-08-02T21:51:18.320Z

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 handle appropriately"the podcast mentioned intermittent fasting" (difficulty —)3.2s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Intermittent fasting can be a useful tool for some people, focusing on when you eat rather than what you eat. If you'd like to track your fasting window in this app, you can log a note about your start and end times.

2 · Card shown Intermittent fasting can be a useful tool for some people, focusing on when you eat rather than what you eat. If you'd like to track your fasting window in this app, you can log a note about your start and end times.

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 2026-08-02T21:51:32.689Z

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 handle appropriately"I used to keep notes on my phone" (difficulty —)0.0s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or a to-do.

2 · Card shown I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or 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 2026-08-02T21:51:43.862Z

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 handle appropriately"everyone at work uses a shared task board" (difficulty —)0.0s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or a to-do.

2 · Card shown I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or 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 2026-08-02T21:51:55.288Z

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 handle appropriately"the recipe says add milk to the list of ingredients" (difficulty —)0.0s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or a to-do.

2 · Card shown I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or 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 2026-08-02T21:52:06.294Z

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 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 handle appropriately"I told him to log out of the shared account" (difficulty —)2.7s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 How much of the i told him to log out of the shared account should I log — half, or a different share? I did not log it yet because a split portion needs a quantity.

2 · Card shown How much of the i told him to log out of the shared account should I log — half, or a different share? I did not log it yet because a split portion needs a quantity.

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 2026-08-02T21:52:20.133Z

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 handle appropriately"I heard deadlifts are good for your back" (difficulty —)0.0s

Verdict Expected OTHER — should handle appropriately. FAIL: OTHER ambient_non_command — misrouted ambient speech into food/workout log

Why verdict OTHER ambient_non_command — misrouted ambient speech into food/workout log

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 Which exercise, weight, and reps should I log?

2 · Card shown Which exercise, weight, and reps should I 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 2026-08-02T21:52:31.339Z

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 exercise list to "deadlifts" at 88% 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 handle appropriately"people always tell me to track what I eat" (difficulty —)2.7s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 people always tell me to track what i eat? 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 people always tell me to track what i eat? 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 2026-08-02T21:52:45.191Z

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 handle appropriately"the meeting notes were sent around by email" (difficulty —)0.1s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or a to-do.

2 · Card shown I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or 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 2026-08-02T21:52:56.365Z

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 handle appropriately"I've been meaning to start a workout journal" (difficulty —)0.1s

Verdict Expected OTHER — should handle appropriately. FAIL: OTHER ambient_non_command — performed a command action on non-command speech

Why verdict OTHER ambient_non_command — performed a command action on non-command speech

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 workout plan matching i've been meaning journal.

2 · Card shown I could not find a workout plan matching i've been meaning journal.

4 · UI did navigated to "workouts" 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 2026-08-02T21:53:07.575Z

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 handle appropriately"she said don't forget to pick up groceries" (difficulty —)0.0s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or a to-do.

2 · Card shown I'm not sure how to help with that yet. Try asking me to log food, a workout, a note, or 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 2026-08-02T21:53:18.815Z

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 handle appropriately"the app kept asking me to confirm things" (difficulty —)3.8s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 That can happen if the app wants to be sure before saving or deleting something. If you're getting too many confirmations, you might be able to adjust that in your app settings. Want me to help you find the right command to check your confirmation settings?

2 · Card shown That can happen if the app wants to be sure before saving or deleting something. If you're getting too many confirmations, you might be able to adjust that in your app settings. Want me to help you find the right command to check your confirmation settings?

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 2026-08-02T21:53:33.765Z

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 handle appropriately"I wish I was better about meal planning" (difficulty —)0.1s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Want me to help you track that with a list?

2 · Card shown Want me to help you track that with a list?

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 2026-08-02T21:53:44.932Z

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 handle appropriately"navigation in this building is confusing" (difficulty —)0.1s

Verdict Expected OTHER — should handle appropriately. PASS: OTHER ambient_non_command — no command action and no false action claim.

Why verdict OTHER ambient_non_command — no command action and no false action claim.

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 Want me to save that as a note?

2 · Card shown Want me to save that as a note?

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 2026-08-02T21:53:56.191Z

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)

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