notes-todos-lists-A-324f4ab3-c63-20260802

Supposed to	N	Correct	Wrong	Unverified
▸ LOG — log the entry	75	55 (73%)	19 (25%)	1 (1%)
19 handled wrong — click one to jump to its full detail below "note that the client prefers email over phone" Reported could-not-confirm (write-confirmation false negative). "note that down" WRITE-TRUTH FAIL — WRONG/MISSING content "parking on Main Street" — no saved note row with matching identity (rows: the client prefers email over phone) "jot this down somewhere" WRITE-TRUTH FAIL — WRONG/MISSING content "contractor needs the deposit by Friday" — no saved note row with matching identity (rows: the parking thing on main street) "uh put buy birthday card for mom on my list" Reported could-not-confirm (write-confirmation false negative). "remind me tomorrow" OVER-ASK — asked instead of writing the note/todo/list (no saved row). "new list called weekend projects" OVER-ASK — asked instead of writing the note/todo/list (no saved row). "new pet supplies list please" OVER-ASK — asked instead of writing the note/todo/list (no saved row). "add fix the leaky faucet to weekend projects" Reported could-not-confirm (write-confirmation false negative). "put milk on the list" OVER-ASK — asked instead of writing the note/todo/list (no saved row). "add eggs" Did not write — misroute-into-food; no notes-domain saved row. "add batteries and light bulbs to the hardware list" OVER-ASK — asked instead of writing the note/todo/list (no saved row). "put that on packing" OVER-ASK — asked instead of writing the note/todo/list (no saved row). "add eggs bread and butter to the shopping list" WRITE-TRUTH FAIL — WRONG/MISSING content "eggs" — no saved list_item row with matching identity (rows: butter); WRONG/MISSING content "bread" — no saved list_item row with matching identity (rows: butter) "add milk eggs and bread to my grocery list" WRITE-TRUTH FAIL — WRONG/MISSING content "milk" — no saved list_item row with matching identity (rows: bread); WRONG/MISSING content "eggs" — no saved list_item row with matching identity (rows: bread) "add three avocados and two limes to groceries" WRITE-TRUTH FAIL — misroute-into-food: notes/todo/list intent wrote food_log_entries instead of a notes-domain table. "log that I finished the quarterly report" OVER-ASK — asked instead of writing the note/todo/list (no saved row). "log that standup ran long today" Did not write — misroute-into-food; no notes-domain saved row. "record that I completed the client onboarding call" OVER-ASK — asked instead of writing the note/todo/list (no saved row). "log I shipped the hotfix at two fifteen p m" WRITE-TRUTH FAIL — expected notes/todo/list write missing — local notes-domain delta is 0 for this turn (proven no write, not a capture gap)
▸ UPDATE — update the entry	27	14 (52%)	13 (48%)	0 (0%)
13 handled wrong — click one to jump to its full detail below "change take out trash from seven to eight thirty" OVER-ASK — asked instead of update (no notes-domain mutation). "mark all the morning tasks done" OVER-ASK — asked instead of update (no notes-domain mutation). "check that off" OVER-ASK — asked instead of update (no notes-domain mutation). "uncheck water the plants I still need to do it" OVER-ASK — asked instead of update (no notes-domain mutation). "edit the grocery list change two percent milk to oat milk" Reported could-not-confirm (write-confirmation false negative). "change the dentist reminder to next wednesday" WRITE-TRUTH FAIL — UPDATE touched notes-domain rows but none matched expected content (dentist reminder). "actually make that almond milk not regular milk" Reported could-not-confirm (write-confirmation false negative). "move eggs to the top of my grocery list" Reported could-not-confirm (write-confirmation false negative). "move milk below bread on the shopping list" Reported could-not-confirm (write-confirmation false negative). "move passport to the top of the packing list" Reported could-not-confirm (write-confirmation false negative). "swap the order of eggs and milk on groceries" WRITE-TRUTH FAIL — misroute-into-food: UPDATE on notes/todo/list wrote food_log_entries. "move that to the top" Reported could-not-confirm (write-confirmation false negative). "move milk down" Did not update — misroute-into-food.
▸ QUERY — answer the query	16	10 (63%)	6 (38%)	0 (0%)
6 handled wrong — click one to jump to its full detail below "how many things are on the beach list" Reported could-not-confirm (write-confirmation false negative). "do I have milk on my grocery list" Query not answered — misroute-into-food. "what's left on weekend projects" Query not answered — asked back instead of answering. "show me what I added to groceries today" Reported could-not-confirm (write-confirmation false negative). "what's on that list" Query not answered — asked back instead of answering. "search todos for anything about insurance" Reported could-not-confirm (write-confirmation false negative).
▸ DELETE — delete the entry	8	6 (75%)	2 (25%)	0 (0%)
2 handled wrong — click one to jump to its full detail below "remove eggs from groceries" OVER-ASK — asked instead of delete (no notes-domain mutation). "remove the second item on my grocery list" WRITE-TRUTH FAIL — DELETE touched notes-domain rows but none matched expected content (second grocery-list item).
Total	126	85 (68%)	40	1

PASSshould log the entry"note that the wifi password is sunflower22" (difficulty —)0.7s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: the wifi password is sunflower22.

Why verdict Notes write-truth verified: the wifi password is sunflower22.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.7s

1 · TTS said Logged your note in General about the wifi password is sunflower22.

2 · Card shown Logged your note in General about the wifi password is sunflower22.

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 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 2026-08-02T21:07:36.738Z

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

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 log the entry"jot down that parking is free after six on main street" (difficulty —)0.6s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: parking is free after six on main street.

Why verdict Notes write-truth verified: parking is free after six on main street.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Logged your note in General about parking is free after six on main street.

2 · Card shown Logged your note in General about parking is free after six on main street.

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 2026-08-02T21:07:48.493Z

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 0.0 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 log the entry"remember the spare key is under the blue flower pot" (difficulty —)0.7s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: the spare key is under the blue flower pot.

Why verdict Notes write-truth verified: the spare key is under the blue flower pot.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.7s

1 · TTS said Logged your note in General about the spare key is under the blue flower pot.

2 · Card shown Logged your note in General about the spare key is under the blue flower pot.

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 2026-08-02T21:08:00.256Z

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 log the entry"make a note the contractor said they start on the fifteenth" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: the contractor said they start on the fifteenth.

Why verdict Notes write-truth verified: the contractor said they start on the fifteenth.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 the contractor said they start on the fifteenth.

2 · Card shown Logged your note in General about the contractor said they start on the fifteenth.

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 2026-08-02T21:08:11.597Z

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 0.0 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 log the entry"save a note my blood pressure was one twenty over eighty this morning" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: my blood pressure was one twenty over eighty this morning.

Why verdict Notes write-truth verified: my blood pressure was one twenty over eighty this morning.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 blood pressure was one twenty over eighty this morning.

2 · Card shown Logged your note in General about my blood pressure was one twenty over eighty this morning.

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 2026-08-02T21:08:22.975Z

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 log the entry"note from the meeting budget is due friday" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: budget is due friday, Meeting.

Why verdict Notes write-truth verified: budget is due friday, Meeting.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Meeting about budget is due friday. Say if that category should be different.

2 · Card shown Logged your note in Meeting about budget is due friday. Say if that category should be different.

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 2026-08-02T21:08:34.338Z

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 log the entry"jot that Jake's jersey size is youth large" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: Jake's jersey size is youth large.

Why verdict Notes write-truth verified: Jake's jersey size is youth large.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Jake's jersey size is youth large.

2 · Card shown Logged your note in General about Jake's jersey size is youth large.

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 2026-08-02T21:08:45.702Z

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)

FAILshould log the entry"note that the client prefers email over phone" (difficulty —)0.6s

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

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

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.6s

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 2026-08-02T21:08:57.460Z

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 log the entry"quick note the gym locker code is four seven two nine" (difficulty —)0.3s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: the gym locker code is four seven two nine.

Why verdict Notes write-truth verified: the gym locker code is four seven two nine.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.3s

1 · TTS said Logged your note in General about the gym locker code is four seven two nine.

2 · Card shown Logged your note in General about the gym locker code is four seven two nine.

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 2026-08-02T21:09:08.881Z

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)

FAILshould log the entry"note that down" (difficulty —)0.3s

Verdict Expected LOG — should log the entry. FAIL: WRITE-TRUTH FAIL — WRONG/MISSING content "parking on Main Street" — no saved note row with matching identity (rows: the client prefers email over phone)

Why verdict WRITE-TRUTH FAIL — WRONG/MISSING content "parking on Main Street" — no saved note row with matching identity (rows: the client prefers email over phone)

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.3s

1 · TTS said Logged your note in General about the client prefers email over phone.

2 · Card shown Logged your note in General about the client prefers email over phone.

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 2026-08-02T21:09:20.261Z

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)

FAILshould log the entry"jot this down somewhere" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. FAIL: WRITE-TRUTH FAIL — WRONG/MISSING content "contractor needs the deposit by Friday" — no saved note row with matching identity (rows: the parking thing on main street)

Why verdict WRITE-TRUTH FAIL — WRONG/MISSING content "contractor needs the deposit by Friday" — no saved note row with matching identity (rows: the parking thing on main street)

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 the parking thing on main street.

2 · Card shown Logged your note in General about the parking thing on main street.

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 2026-08-02T21:09:44.638Z

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 update the entry"append also grab oat milk to my latest note" (difficulty —)0.2s

Verdict Expected UPDATE — should update the entry. PASS: Notes UPDATE write-truth verified (changed:personal_notes×1).

Why verdict Notes UPDATE write-truth verified (changed:personal_notes×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Updated your note in General about the parking thing on main street.

2 · Card shown Updated your note in General about the parking thing on main street.

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 2026-08-02T21:10:19.958Z

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 0.0 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 log the entry"remind me to call the dentist tomorrow" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: call the dentist.

Why verdict Notes write-truth verified: call the dentist.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: call the dentist. Priority 3, due Aug 3. Tell me if the title or due date should be different.

2 · Card shown Added to-do: call the dentist. Priority 3, due Aug 3. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: call the dentist

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 2026-08-02T21:10:31.344Z

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 log the entry"add pick up dry cleaning to my to-do list" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: pick up dry cleaning.

Why verdict Notes write-truth verified: pick up dry cleaning.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: pick up dry cleaning. Priority 3. Tell me if the title or due date should be different.

2 · Card shown Added to-do: pick up dry cleaning. Priority 3. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: pick up dry cleaning

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 2026-08-02T21:10:42.741Z

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 log the entry"put return the library books on my to-dos" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: return the library books.

Why verdict Notes write-truth verified: return the library books.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: return the library books. Priority 3. Tell me if the title or due date should be different.

2 · Card shown Added to-do: return the library books. Priority 3. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: return the library books

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 2026-08-02T21:10:54.058Z

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)

FAILshould log the entry"uh put buy birthday card for mom on my list" (difficulty —)0.1s

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

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

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

1 · TTS said I'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 2026-08-02T21:11:05.234Z

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 log the entry"remind me to text Sarah about dinner plans" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: text Sarah about dinner plans.

Why verdict Notes write-truth verified: text Sarah about dinner plans.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: text Sarah about dinner plans. Priority 3. Tell me if the title or due date should be different.

2 · Card shown Added to-do: text Sarah about dinner plans. Priority 3. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: text Sarah about dinner plans

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 2026-08-02T21:11:16.611Z

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 log the entry"put schedule oil change on my car to-do list" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: schedule oil change.

Why verdict Notes write-truth verified: schedule oil change.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: schedule oil change. Priority 3. Tell me if the title or due date should be different.

2 · Card shown Added to-do: schedule oil change. Priority 3. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: schedule oil change

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 2026-08-02T21:11:27.925Z

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 log the entry"add call the insurance company to my morning list" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: call the insurance company.

Why verdict Notes write-truth verified: call the insurance company.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: call the insurance company. Priority 3. Tell me if the title or due date should be different.

2 · Card shown Added to-do: call the insurance company. Priority 3. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: call the insurance company

4 · UI did navigated to "todos" screen

5 · Seed/context data

Seeded starting state:

Seeded value

{
  "table": "todo_categories",
  "record": {
    "id": "seed-todo-cat-morning",
    "name": "Morning",
    "created_at": {
      "$relativeDays": -14
    }
  }
}

snapshot captured 2026-08-02T21:11:53.691Z

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 log the entry"new todo finish the slide deck by thursday" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: finish the slide deck.

Why verdict Notes write-truth verified: finish the slide deck.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: finish the slide deck. Priority 3, due Aug 6. Tell me if the title or due date should be different.

2 · Card shown Added to-do: finish the slide deck. Priority 3, due Aug 6. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: finish the slide deck

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 2026-08-02T21:12:06.663Z

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 log the entry"remind me friday morning to submit the expense report" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: submit the expense report.

Why verdict Notes write-truth verified: submit the expense report.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: submit the expense report. Priority 3, due Aug 7. Tell me if the title or due date should be different.

2 · Card shown Added to-do: submit the expense report. Priority 3, due Aug 7. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: submit the expense report

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 2026-08-02T21:12:27.737Z

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 log the entry"add a to-do to renew my passport in the category Admin" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: renew my passport.

Why verdict Notes write-truth verified: renew my passport.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: renew my passport. Priority 3.

2 · Card shown Added to-do: renew my passport. Priority 3.

3 · App data rows written created todo_item: renew my passport

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 2026-08-02T21:12:39.106Z

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)

FAILshould log the entry"remind me tomorrow" (difficulty —)0.1s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of writing the note/todo/list (no saved row).

Why verdict OVER-ASK — asked instead of writing the note/todo/list (no saved row).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

1 · TTS said Want me to add that as a to-do or a list item?

2 · Card shown Want me to add that as a to-do or a list item?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:12:50.284Z

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 log the entry"add it to todos" (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: submit my timesheet.

Why verdict Notes write-truth verified: submit my timesheet.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 2026-08-02T21:13:15.741Z

7 · Lookup method

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

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

PASSshould log the entry"make sure I text Sarah back and also add a to-do to buy milk" (difficulty —)0.3s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: buy milk, text Sarah back.

Why verdict Notes write-truth verified: buy milk, text Sarah back.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added 2 to-dos: text Sarah back; buy milk.

2 · Card shown Added 2 to-dos: text Sarah back; buy milk.

3 · App data rows written created todo_item: text Sarah back
created todo_item: buy milk

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 2026-08-02T21:13:38.406Z

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 0.0 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 log the entry"remind me when I get home to defrost the chicken" (difficulty —)2.8s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: when I get home to defrost the chicken.

Why verdict Notes write-truth verified: when I get home to defrost the chicken.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: when I get home to defrost the chicken. Priority 3. Tell me if the title or due date should be different.

2 · Card shown Added to-do: when I get home to defrost the chicken. Priority 3. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: when I get home to defrost the chicken

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 2026-08-02T21:13:52.367Z

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

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

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

PASSshould log the entry"remind me later" (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: mail the package this afternoon.

Why verdict Notes write-truth verified: mail the package this afternoon.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 2026-08-02T21:14:03.530Z

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 log the entry"start a grocery list" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: Grocery List.

Why verdict Notes write-truth verified: Grocery List.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Created list: Grocery list.

2 · Card shown Created list: Grocery list.

3 · App data rows written created general_list: Grocery List

4 · UI did navigated to "lists" 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 2026-08-02T21:14:26.100Z

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 log the entry"create a packing list for the trip" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: Packing List For The Trip.

Why verdict Notes write-truth verified: Packing List For The Trip.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Created list: Packing list for the trip.

2 · Card shown Created list: Packing list for the trip.

3 · App data rows written created general_list: Packing List For The Trip

4 · UI did navigated to "lists" 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 2026-08-02T21:14:37.421Z

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)

FAILshould log the entry"new list called weekend projects" (difficulty —)2.8s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of writing the note/todo/list (no saved row).

Why verdict OVER-ASK — asked instead of writing the note/todo/list (no saved row).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 2.8s

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 2026-08-02T21:14:51.349Z

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 log the entry"make a shopping list" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: Shopping List.

Why verdict Notes write-truth verified: Shopping List.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Created list: Shopping list.

2 · Card shown Created list: Shopping list.

3 · App data rows written created general_list: Shopping List

4 · UI did navigated to "lists" 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 2026-08-02T21:15:02.724Z

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 log the entry"start a beach list for saturday" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: Beach List For Saturday.

Why verdict Notes write-truth verified: Beach List For Saturday.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Created list: Beach list for saturday.

2 · Card shown Created list: Beach list for saturday.

3 · App data rows written created general_list: Beach List For Saturday

4 · UI did navigated to "lists" 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 2026-08-02T21:15:14.018Z

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 0.0 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 log the entry"make a new list for that" (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: Camping Supplies List.

Why verdict Notes write-truth verified: Camping Supplies List.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

1 · TTS said What should I call the new list?

2 · Card shown What should I call the new 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 2026-08-02T21:15:25.190Z

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 log the entry"create a hardware store list" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: Hardware Store List.

Why verdict Notes write-truth verified: Hardware Store List.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Created list: Hardware store list.

2 · Card shown Created list: Hardware store list.

3 · App data rows written created general_list: Hardware Store List

4 · UI did navigated to "lists" 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 2026-08-02T21:15:47.737Z

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)

FAILshould log the entry"new pet supplies list please" (difficulty —)1.4s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of writing the note/todo/list (no saved row).

Why verdict OVER-ASK — asked instead of writing the note/todo/list (no saved row).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 1.4s

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 2026-08-02T21:16:00.490Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.0 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 log the entry"add milk to my grocery list" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: milk.

Why verdict Notes write-truth verified: milk.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added milk to Grocery List.

2 · Card shown Added milk to Grocery List.

3 · App data rows written created general_list_item:

4 · UI did navigated to "lists" 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 2026-08-02T21:16:11.601Z

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 0.0 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 log the entry"add greek yogurt to the shopping list" (difficulty —)0.3s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: greek yogurt.

Why verdict Notes write-truth verified: greek yogurt.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added greek yogurt to Shopping List.

2 · Card shown Added greek yogurt to Shopping List.

3 · App data rows written created general_list_item:

4 · UI did navigated to "lists" 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 2026-08-02T21:16:22.985Z

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 "Plain Greek yogurt" 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 log the entry"add cat litter to my pet supplies list" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: Pet Supplies, cat litter.

Why verdict Notes write-truth verified: Pet Supplies, cat litter.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Created list: Pet supplies. Added cat litter.

2 · Card shown Created list: Pet supplies. Added cat litter.

3 · App data rows written created general_list: Pet Supplies
created general_list_item:

4 · UI did navigated to "lists" 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 2026-08-02T21:16:34.428Z

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 log the entry"add charger and passport to my packing list" (difficulty —)0.4s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: passport, charger.

Why verdict Notes write-truth verified: passport, charger.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added 2 items to Packing List For The Trip: charger and passport.

2 · Card shown Added 2 items to Packing List For The Trip: charger and passport.

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

4 · UI did navigated to "lists" 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 2026-08-02T21:16:45.931Z

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)

FAILshould log the entry"add fix the leaky faucet to weekend projects" (difficulty —)0.1s

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

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

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

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

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

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:16:57.096Z

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 0.0 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 log the entry"add sunscreen and snacks to the beach list" (difficulty —)0.3s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: snacks, sunscreen.

Why verdict Notes write-truth verified: snacks, sunscreen.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added 2 items to Beach List For Saturday: sunscreen and snacks.

2 · Card shown Added 2 items to Beach List For Saturday: sunscreen and snacks.

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

4 · UI did navigated to "lists" 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 2026-08-02T21:17:20.019Z

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 log the entry"add two bottles of sparkling water to groceries" (difficulty —)2.9s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: Groceries, two bottles of sparkling water.

Why verdict Notes write-truth verified: Groceries, two bottles of sparkling water.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Created list: Groceries. Added two bottles of sparkling water.

2 · Card shown Created list: Groceries. Added two bottles of sparkling water.

3 · App data rows written created general_list: Groceries
created general_list_item:

4 · UI did navigated to "lists" 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 2026-08-02T21:17:35.237Z

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

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

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

FAILshould log the entry"put milk on the list" (difficulty —)0.1s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of writing the note/todo/list (no saved row).

Why verdict OVER-ASK — asked instead of writing the note/todo/list (no saved row).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

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

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

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seeded value

{
  "table": "general_lists",
  "record": {
    "id": "seed-list-shopping",
    "name": "Shopping",
    "description": null,
    "list_type": "shopping",
    "archived": false,
    "sort_mode": "manual",
    "sort_direction": "asc",
    "aliases": [],
    "created_at": {
      "$relativeDays": -14
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null
  }
}

snapshot captured 2026-08-02T21:17:45.210Z

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 0.0 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

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

FAILshould log the entry"add eggs" (difficulty —)0.1s

Verdict Expected LOG — should log the entry. FAIL: Did not write — misroute-into-food; no notes-domain saved row.

Why verdict Did not write — misroute-into-food; no notes-domain saved row.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

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

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

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:18:12.118Z

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 0.0 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 log the entry"add eggs to my list" (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: eggs.

Why verdict Notes write-truth verified: eggs.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 eggs to?

2 · Card shown Which list should I add eggs 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 2026-08-02T21:18:37.021Z

6 · Why this food

The lookup compared against planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
Matched 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 0.0 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 log the entry"add olive oil and paper towels to groceries" (difficulty —)0.3s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: paper towels, olive oil.

Why verdict Notes write-truth verified: paper towels, olive oil.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.3s

1 · TTS said Added 2 items to Groceries: olive oil and paper towels.

2 · Card shown Added 2 items to Groceries: olive oil and paper towels.

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

4 · UI did navigated to "lists" 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 2026-08-02T21:18:59.704Z

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 "Olive oil" 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)

FAILshould log the entry"add batteries and light bulbs to the hardware list" (difficulty —)0.1s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of writing the note/todo/list (no saved row).

Why verdict OVER-ASK — asked instead of writing the note/todo/list (no saved row).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

1 · TTS said Did you mean 2 items: batteries and light bulbs?

2 · Card shown Did you mean 2 items: batteries and light bulbs?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:19:10.949Z

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)

FAILshould log the entry"put that on packing" (difficulty —)0.1s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of writing the note/todo/list (no saved row).

Why verdict OVER-ASK — asked instead of writing the note/todo/list (no saved row).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

1 · TTS said What should I add to Packing List For The Trip?

2 · Card shown What should I add to Packing List For The Trip?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:19:22.209Z

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 log the entry"add 1.5 cups of rice to my meal prep list" (difficulty —)2.4s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: Meal Prep, 1.5 cups of rice.

Why verdict Notes write-truth verified: Meal Prep, 1.5 cups of rice.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Created list: Meal prep. Added 1.5 cups of rice.

2 · Card shown Created list: Meal prep. Added 1.5 cups of rice.

3 · App data rows written created general_list: Meal Prep
created general_list_item:

4 · UI did navigated to "lists" 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 2026-08-02T21:19:46.888Z

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

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

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

PASSshould log the entry"add the usual stuff to shopping" (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: bread, eggs, milk.

Why verdict Notes write-truth verified: bread, eggs, milk.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

1 · TTS said What should I add to Shopping List?

2 · Card shown What should I add to Shopping List?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seeded value

{
  "table": "general_lists",
  "record": {
    "id": "seed-list-shopping-usual",
    "name": "Shopping",
    "description": null,
    "list_type": "shopping",
    "archived": false,
    "sort_mode": "manual",
    "sort_direction": "asc",
    "aliases": [],
    "created_at": {
      "$relativeDays": -14
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null
  }
}

Seeded value

{
  "table": "general_list_items",
  "record": {
    "id": "seed-item-usual-milk",
    "list_id": "seed-list-shopping-usual",
    "position": 0,
    "notes": null,
    "created_at": {
      "$relativeDays": -3
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null,
    "title": "milk",
    "state": "active",
    "priority": null,
    "tags": [],
    "url": null,
    "due_date": null,
    "review_date": null,
    "source_utterance": null
  }
}

snapshot captured 2026-08-02T21:19:58.101Z

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)

FAILshould log the entry"add eggs bread and butter to the shopping list" (difficulty —)0.4s

Verdict Expected LOG — should log the entry. FAIL: WRITE-TRUTH FAIL — WRONG/MISSING content "eggs" — no saved list_item row with matching identity (rows: butter); WRONG/MISSING content "bread" — no saved list_item row with matching identity (rows: butter)

Why verdict WRITE-TRUTH FAIL — WRONG/MISSING content "eggs" — no saved list_item row with matching identity (rows: butter); WRONG/MISSING content "bread" — no saved list_item row with matching identity (rows: butter)

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Saved 1 of 3 items to Shopping List: butter. Already on Shopping List: eggs and bread.

2 · Card shown Saved 1 of 3 items to Shopping List: butter. Already on Shopping List: eggs and bread.

3 · App data rows written created general_list_item:

4 · UI did navigated to "lists" 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 2026-08-02T21:20:21.089Z

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 log the entry"put bananas and uh the other thing on groceries" (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: peanut butter, bananas.

Why verdict Notes write-truth verified: peanut butter, bananas.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Did you mean 2 items: bananas and uh the other thing?

2 · Card shown Did you mean 2 items: bananas and uh the other thing?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:20:32.315Z

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 "Banana" 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)

FAILshould log the entry"add milk eggs and bread to my grocery list" (difficulty —)0.6s

Verdict Expected LOG — should log the entry. FAIL: WRITE-TRUTH FAIL — WRONG/MISSING content "milk" — no saved list_item row with matching identity (rows: bread); WRONG/MISSING content "eggs" — no saved list_item row with matching identity (rows: bread)

Why verdict WRITE-TRUTH FAIL — WRONG/MISSING content "milk" — no saved list_item row with matching identity (rows: bread); WRONG/MISSING content "eggs" — no saved list_item row with matching identity (rows: bread)

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Saved 1 of 3 items to Grocery List: bread. Already on Grocery List: milk and eggs.

2 · Card shown Saved 1 of 3 items to Grocery List: bread. Already on Grocery List: milk and eggs.

3 · App data rows written created general_list_item:

4 · UI did navigated to "lists" 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 2026-08-02T21:20:55.493Z

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 0.0 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 log the entry"add tomatoes onions and cilantro to the taco night list" (difficulty —)0.5s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: Taco Night, cilantro, onions, tomatoes.

Why verdict Notes write-truth verified: Taco Night, cilantro, onions, tomatoes.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.5s

1 · TTS said Created list: Taco night. Added 3 items: tomatoes, onions, and cilantro.

2 · Card shown Created list: Taco night. Added 3 items: tomatoes, onions, and cilantro.

3 · App data rows written created general_list: Taco Night
created general_list_item:
created general_list_item:
created general_list_item:

4 · UI did navigated to "lists" 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 2026-08-02T21:21:07.103Z

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 log the entry"throw apples oranges and bananas on the fruit list" (difficulty —)0.3s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: Fruit, bananas, oranges, apples.

Why verdict Notes write-truth verified: Fruit, bananas, oranges, apples.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Created list: Fruit. Added 3 items: apples, oranges, and bananas.

2 · Card shown Created list: Fruit. Added 3 items: apples, oranges, and bananas.

3 · App data rows written created general_list: Fruit
created general_list_item:
created general_list_item:
created general_list_item:

4 · UI did navigated to "lists" 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 2026-08-02T21:21:18.559Z

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 "Banana" 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 log the entry"add shampoo conditioner and body wash to the drugstore list" (difficulty —)0.3s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: Drugstore, body wash, conditioner, shampoo.

Why verdict Notes write-truth verified: Drugstore, body wash, conditioner, shampoo.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Created list: Drugstore. Added 3 items: shampoo, conditioner, and body wash.

2 · Card shown Created list: Drugstore. Added 3 items: shampoo, conditioner, and body wash.

3 · App data rows written created general_list: Drugstore
created general_list_item:
created general_list_item:
created general_list_item:

4 · UI did navigated to "lists" 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 2026-08-02T21:21:30.075Z

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 log the entry"add chicken thighs broccoli and brown rice to meal prep" (difficulty —)0.4s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: brown rice, broccoli, chicken thighs.

Why verdict Notes write-truth verified: brown rice, broccoli, chicken thighs.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added 3 items to Meal Prep: chicken thighs, broccoli, and brown rice.

2 · Card shown Added 3 items to Meal Prep: chicken thighs, broccoli, and brown rice.

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

4 · UI did navigated to "lists" 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 2026-08-02T21:21:41.662Z

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 log the entry"put tape measure screws and anchors on the hardware list" (difficulty —)0.4s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: anchors, screws, tape measure.

Why verdict Notes write-truth verified: anchors, screws, tape measure.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added 3 items to Hardware Store List: tape measure, screws, and anchors.

2 · Card shown Added 3 items to Hardware Store List: tape measure, screws, and anchors.

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

4 · UI did navigated to "lists" 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 2026-08-02T21:21:53.269Z

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)

FAILshould log the entry"add three avocados and two limes to groceries" (difficulty —)0.5s

Verdict Expected LOG — should log the entry. FAIL: WRITE-TRUTH FAIL — misroute-into-food: notes/todo/list intent wrote food_log_entries instead of a notes-domain table.

Why verdict WRITE-TRUTH FAIL — misroute-into-food: notes/todo/list intent wrote food_log_entries instead of a notes-domain table.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.5s

1 · TTS said Logged three avocados and Lime.

2 · Card shown Logged three avocados and Lime.

3 · App data rows written created food_log_entry: Avocado ×1 (3 avocados) 720 cal · 9g P · 38.3g C · 66.1g F
created food_log_entry: Lime ×2 (2 limes) 80 cal · 1.8g P · 28.2g C · 0.6g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:22:04.934Z

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

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

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

PASSshould log the entry"add dish soap sponges and trash bags to the household list" (difficulty —)0.4s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: Household, trash bags, sponges, dish soap.

Why verdict Notes write-truth verified: Household, trash bags, sponges, dish soap.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Created list: Household. Added 3 items: dish soap, sponges, and trash bags.

2 · Card shown Created list: Household. Added 3 items: dish soap, sponges, and trash bags.

3 · App data rows written created general_list: Household
created general_list_item:
created general_list_item:
created general_list_item:

4 · UI did navigated to "lists" 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 2026-08-02T21:22:16.565Z

7 · Lookup method

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

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

PASSshould log the entry"remind me to take out the trash tonight at seven" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: take out the trash tonight at seven.

Why verdict Notes write-truth verified: take out the trash tonight at seven.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: take out the trash tonight at seven. Priority 3. Tell me if the title or due date should be different.

2 · Card shown Added to-do: take out the trash tonight at seven. Priority 3. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: take out the trash tonight at seven

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 2026-08-02T21:22:39.624Z

7 · Lookup method

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

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

PASSshould log the entry"remind me in two hours to check the oven" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: in two hours to check the oven.

Why verdict Notes write-truth verified: in two hours to check the oven.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: in two hours to check the oven. Priority 3. Tell me if the title or due date should be different.

2 · Card shown Added to-do: in two hours to check the oven. Priority 3. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: in two hours to check the oven

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 2026-08-02T21:22:51.031Z

7 · Lookup method

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

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

PASSshould log the entry"remind me next monday to renew my car registration" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: renew my car registration.

Why verdict Notes write-truth verified: renew my car registration.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: renew my car registration. Priority 3, due Aug 3. Tell me if the title or due date should be different.

2 · Card shown Added to-do: renew my car registration. Priority 3, due Aug 3. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: renew my car registration

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 2026-08-02T21:23:02.341Z

7 · Lookup method

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

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

PASSshould log the entry"new to-do water the plants every sunday" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: water the plants.

Why verdict Notes write-truth verified: water the plants.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: water the plants. Priority 3, due Aug 2. Tell me if the title or due date should be different.

2 · Card shown Added to-do: water the plants. Priority 3, due Aug 2. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: water the plants

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 2026-08-02T21:23:13.653Z

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 "Water" at 90% confidence.

7 · Lookup method

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

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

PASSshould log the entry"remind me every morning at eight to take my vitamins" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: every morning at eight to take my vitamins.

Why verdict Notes write-truth verified: every morning at eight to take my vitamins.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: every morning at eight to take my vitamins. Priority 3. Tell me if the title or due date should be different.

2 · Card shown Added to-do: every morning at eight to take my vitamins. Priority 3. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: every morning at eight to take my vitamins

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 2026-08-02T21:23:24.996Z

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 log the entry"set a reminder every monday to take out the recycling" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: take out the recycling every monday.

Why verdict Notes write-truth verified: take out the recycling every monday.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: take out the recycling. Priority 3, due Aug 3.

2 · Card shown Added to-do: take out the recycling. Priority 3, due Aug 3.

3 · App data rows written created todo_item: take out the recycling

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 2026-08-02T21:23:36.369Z

7 · Lookup method

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

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

PASSshould log the entry"remind me about that thing tomorrow" (difficulty —)0.1s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: dentist appointment.

Why verdict Notes write-truth verified: dentist appointment.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 2026-08-02T21:23:47.587Z

7 · Lookup method

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

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

PASSshould log the entry"new reminder for the appointment" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: the appointment.

Why verdict Notes write-truth verified: the appointment.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: the appointment. Priority 3. Tell me if the title or due date should be different.

2 · Card shown Added to-do: the appointment. Priority 3. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: the appointment

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 2026-08-02T21:24:10.188Z

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 log the entry"uh same reminder as last week but for friday" (difficulty —)1.5s

Verdict Expected LOG — should log the entry. 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 1.5s

1 · TTS said I can help you set a reminder for Friday, but I'll need a bit more detail. Could you tell me what the reminder should be about?

2 · Card shown I can help you set a reminder for Friday, but I'll need a bit more detail. Could you tell me what the reminder should be about?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seeded value

{
  "table": "todo_categories",
  "record": {
    "id": "seed-todo-cat-personal",
    "name": "Personal",
    "created_at": {
      "$relativeDays": -14
    }
  }
}

Seeded value

{
  "table": "todo_items",
  "record": {
    "id": "seed-todo-last-week-reminder",
    "category_id": "seed-todo-cat-personal",
    "category_name": "Personal",
    "name": "Weekly reminder",
    "description": null,
    "priority": 3,
    "scheduled_date": null,
    "due_date": null,
    "completed_at": null,
    "created_at": {
      "$relativeDays": -7
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null
  }
}

snapshot captured 2026-08-02T21:24:44.963Z

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 log the entry"remind me to pay the electric bill on the first of every month" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: pay the electric bill on the first of every month.

Why verdict Notes write-truth verified: pay the electric bill on the first of every month.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Added to-do: pay the electric bill on the first of every month. Priority 3. Tell me if the title or due date should be different.

2 · Card shown Added to-do: pay the electric bill on the first of every month. Priority 3. Tell me if the title or due date should be different.

3 · App data rows written created todo_item: pay the electric bill on the first of every month

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 2026-08-02T21:25:07.385Z

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 update the entry"set the expense report todo due date to friday at noon" (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. PASS: Notes UPDATE write-truth verified (changed:todo_items×1).

Why verdict Notes UPDATE write-truth verified (changed:todo_items×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Updated to-do: Expense report.

2 · Card shown Updated to-do: Expense report.

4 · UI did navigated to "todos" screen

5 · Seed/context data

Seeded starting state:

Seeded value

{
  "table": "todo_categories",
  "record": {
    "id": "seed-todo-cat-work",
    "name": "Work",
    "created_at": {
      "$relativeDays": -14
    }
  }
}

Seeded value

{
  "table": "todo_items",
  "record": {
    "id": "seed-todo-expense",
    "category_id": "seed-todo-cat-work",
    "category_name": "Work",
    "name": "Expense report",
    "description": null,
    "priority": 3,
    "scheduled_date": null,
    "due_date": null,
    "completed_at": null,
    "created_at": {
      "$relativeDays": -7
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null
  }
}

snapshot captured 2026-08-02T21:25:18.698Z

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 "Dates" at 90% confidence.

7 · Lookup method

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

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

FAILshould update the entry"change take out trash from seven to eight thirty" (difficulty —)2.0s

Verdict Expected UPDATE — should update the entry. FAIL: OVER-ASK — asked instead of update (no notes-domain mutation).

Why verdict OVER-ASK — asked instead of update (no notes-domain mutation).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 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 2026-08-02T21:25:31.793Z

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 update the entry"mark call the dentist done" (difficulty —)0.2s

Verdict Expected UPDATE — should update the entry. PASS: Notes UPDATE write-truth verified (changed:todo_items×1).

Why verdict Notes UPDATE write-truth verified (changed:todo_items×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Marked done: call the dentist.

2 · Card shown Marked done: call the dentist.

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 2026-08-02T21:25:43.163Z

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 update the entry"check off call the dentist" (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. PASS: UPDATE — no notes-domain write (as expected).

Why verdict UPDATE — no notes-domain write (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 call the dentist is already done — want me to uncheck it?

2 · Card shown call the dentist is already done — want me to uncheck it?

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 2026-08-02T21:26:08.763Z

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 update the entry"mark pick up dry cleaning as complete" (difficulty —)0.2s

Verdict Expected UPDATE — should update the entry. PASS: Notes UPDATE write-truth verified (changed:todo_items×1).

Why verdict Notes UPDATE write-truth verified (changed:todo_items×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Marked done: pick up dry cleaning.

2 · Card shown Marked done: pick up dry cleaning.

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 2026-08-02T21:26:20.072Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 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 update the entry"I finished the buy milk task" (difficulty —)0.2s

Verdict Expected UPDATE — should update the entry. PASS: Notes UPDATE write-truth verified (changed:todo_items×1).

Why verdict Notes UPDATE write-truth verified (changed:todo_items×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Marked done: buy milk.

2 · Card shown Marked done: buy milk.

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 2026-08-02T21:26:31.338Z

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 0.1 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 update the entry"mark the laundry done" (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. PASS: Notes UPDATE write-truth verified (changed:todo_items×1).

Why verdict Notes UPDATE write-truth verified (changed:todo_items×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Marked done: laundry.

2 · Card shown Marked done: laundry.

4 · UI did navigated to "todos" screen

5 · Seed/context data

Seeded starting state:

Seeded value

{
  "table": "todo_categories",
  "record": {
    "id": "seed-todo-cat-personal",
    "name": "Personal",
    "created_at": {
      "$relativeDays": -14
    }
  }
}

Seeded value

{
  "table": "todo_items",
  "record": {
    "id": "seed-todo-laundry",
    "category_id": "seed-todo-cat-personal",
    "category_name": "Personal",
    "name": "laundry",
    "description": null,
    "priority": 3,
    "scheduled_date": null,
    "due_date": null,
    "completed_at": null,
    "created_at": {
      "$relativeDays": -7
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null
  }
}

snapshot captured 2026-08-02T21:26:42.604Z

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 update the entry"mark defrost chicken done" (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. PASS: Notes UPDATE write-truth verified (changed:todo_items×1).

Why verdict Notes UPDATE write-truth verified (changed:todo_items×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Marked done: when I get home to defrost the chicken.

2 · Card shown Marked done: when I get home to defrost the chicken.

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 2026-08-02T21:26:53.942Z

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 "Chicken breast" at 90% confidence.

7 · Lookup method

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

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

FAILshould update the entry"mark all the morning tasks done" (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. FAIL: OVER-ASK — asked instead of update (no notes-domain mutation).

Why verdict OVER-ASK — asked instead of update (no notes-domain mutation).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

1 · TTS said Which to-do should I mark done: every morning at eight to take my vitamins; call the insurance company?

2 · Card shown Which to-do should I mark done: every morning at eight to take my vitamins; call the insurance company?

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 2026-08-02T21:27:05.300Z

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 update the entry"mark that done" (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. PASS: UPDATE — no notes-domain write (as expected).

Why verdict UPDATE — no notes-domain write (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 to-do should I mark done: pay the electric bill on the first of every month; the appointment; dentist appointment?

2 · Card shown Which to-do should I mark done: pay the electric bill on the first of every month; the appointment; dentist appointment?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:27:16.477Z

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 update the entry"check that off" (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. FAIL: OVER-ASK — asked instead of update (no notes-domain mutation).

Why verdict OVER-ASK — asked instead of update (no notes-domain mutation).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

1 · TTS said Which to-do should I mark done: Expense report; pay the electric bill on the first of every month; the appointment?

2 · Card shown Which to-do should I mark done: Expense report; pay the electric bill on the first of every month; the appointment?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:27:50.173Z

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 update the entry"uncheck water the plants I still need to do it" (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. FAIL: OVER-ASK — asked instead of update (no notes-domain mutation).

Why verdict OVER-ASK — asked instead of update (no notes-domain mutation).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

1 · TTS said I could not find a to-do matching water the plants. Did you mean: call the dentist; pick up dry cleaning; buy milk?

2 · Card shown I could not find a to-do matching water the plants. Did you mean: call the dentist; pick up dry cleaning; buy milk?

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 2026-08-02T21:28:12.438Z

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 "Water" 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 delete the entry"remove milk from my grocery list" (difficulty —)0.1s

Verdict Expected DELETE — should delete the entry. PASS: Notes DELETE write-truth verified (removed:general_list_items×1).

Why verdict Notes DELETE write-truth verified (removed:general_list_items×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Removed milk from the list. Say if that (Grocery List) is the wrong list.

2 · Card shown Removed milk from the list. Say if that (Grocery List) is the wrong list.

4 · UI did navigated to "lists" 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 2026-08-02T21:28:23.658Z

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 0.1 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 delete the entry"remove eggs from groceries" (difficulty —)0.1s

Verdict Expected DELETE — should delete the entry. FAIL: OVER-ASK — asked instead of delete (no notes-domain mutation).

Why verdict OVER-ASK — asked instead of delete (no notes-domain mutation).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

1 · TTS said I could not find a list item matching eggs on Groceries.

2 · Card shown I could not find a list item matching eggs on Groceries.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:28:34.914Z

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)

PASSshould delete the entry"take eggs off the list" (difficulty —)2.7s

Verdict Expected DELETE — should delete the entry. PASS: Notes DELETE write-truth verified (removed:general_list_items×1).

Why verdict Notes DELETE write-truth verified (removed:general_list_items×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Removed eggs from the list. Say if that is the wrong list.

2 · Card shown Removed eggs from the list. Say if that is the wrong list.

4 · UI did navigated to "lists" screen

5 · Seed/context data

Seeded starting state:

Seeded value

{
  "table": "general_lists",
  "record": {
    "id": "seed-list-grocery-eggs",
    "name": "Grocery",
    "description": null,
    "list_type": "shopping",
    "archived": false,
    "sort_mode": "manual",
    "sort_direction": "asc",
    "aliases": [],
    "created_at": {
      "$relativeDays": -14
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null
  }
}

Seeded value

{
  "table": "general_list_items",
  "record": {
    "id": "seed-item-eggs",
    "list_id": "seed-list-grocery-eggs",
    "position": 0,
    "notes": null,
    "created_at": {
      "$relativeDays": -3
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null,
    "title": "eggs",
    "state": "active",
    "priority": null,
    "tags": [],
    "url": null,
    "due_date": null,
    "review_date": null,
    "source_utterance": null
  }
}

snapshot captured 2026-08-02T21:28:48.831Z

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 delete the entry"remove passport from the packing list" (difficulty —)0.1s

Verdict Expected DELETE — should delete the entry. PASS: Notes DELETE write-truth verified (removed:general_list_items×1).

Why verdict Notes DELETE write-truth verified (removed:general_list_items×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Removed passport from the list. Say if that (Packing List For The Trip) is the wrong list.

2 · Card shown Removed passport from the list. Say if that (Packing List For The Trip) is the wrong list.

4 · UI did navigated to "lists" 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 2026-08-02T21:29:00.328Z

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 delete the entry"remove cat litter from pet supplies" (difficulty —)0.1s

Verdict Expected DELETE — should delete the entry. PASS: Notes DELETE write-truth verified (removed:general_list_items×1).

Why verdict Notes DELETE write-truth verified (removed:general_list_items×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Removed cat litter from the list. Say if that (Pet Supplies) is the wrong list.

2 · Card shown Removed cat litter from the list. Say if that (Pet Supplies) is the wrong list.

4 · UI did navigated to "lists" 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 2026-08-02T21:29:11.285Z

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 delete the entry"remove that from the list" (difficulty —)0.1s

Verdict Expected DELETE — should delete the entry. PASS: Notes DELETE write-truth verified (removed:general_list_items×1).

Why verdict Notes DELETE write-truth verified (removed:general_list_items×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 item should I take off?

2 · Card shown Which list item should I take off?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state:

Seeded value

{
  "table": "general_lists",
  "record": {
    "id": "seed-list-default",
    "name": "Shopping",
    "description": null,
    "list_type": "shopping",
    "archived": false,
    "sort_mode": "manual",
    "sort_direction": "asc",
    "aliases": [],
    "created_at": {
      "$relativeDays": -14
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null
  }
}

Seeded value

{
  "table": "general_list_items",
  "record": {
    "id": "seed-item-recent",
    "list_id": "seed-list-default",
    "position": 0,
    "notes": null,
    "created_at": {
      "$relativeDays": -3
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null,
    "title": "butter",
    "state": "active",
    "priority": null,
    "tags": [],
    "url": null,
    "due_date": null,
    "review_date": null,
    "source_utterance": null
  }
}

snapshot captured 2026-08-02T21:29:22.457Z

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 delete the entry"remove bread and butter from the shopping list" (difficulty —)0.1s

Verdict Expected DELETE — should delete the entry. PASS: Notes DELETE write-truth verified (removed:general_list_items×2).

Why verdict Notes DELETE write-truth verified (removed:general_list_items×2).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Removed 2 items: bread and butter. Say if that (Shopping List) is the wrong list.

2 · Card shown Removed 2 items: bread and butter. Say if that (Shopping List) is the wrong list.

4 · UI did navigated to "lists" 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 2026-08-02T21:29:44.757Z

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 delete the entry"remove the second item on my grocery list" (difficulty —)0.1s

Verdict Expected DELETE — should delete the entry. FAIL: WRITE-TRUTH FAIL — DELETE touched notes-domain rows but none matched expected content (second grocery-list item).

Why verdict WRITE-TRUTH FAIL — DELETE touched notes-domain rows but none matched expected content (second grocery-list item).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Grocery no longer includes eggs.

2 · Card shown Grocery no longer includes eggs.

4 · UI did navigated to "lists" screen

5 · Seed/context data

Seeded starting state:

Seeded value

{
  "table": "general_lists",
  "record": {
    "id": "seed-list-grocery-2",
    "name": "Grocery",
    "description": null,
    "list_type": "shopping",
    "archived": false,
    "sort_mode": "manual",
    "sort_direction": "asc",
    "aliases": [],
    "created_at": {
      "$relativeDays": -14
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null
  }
}

Seeded value

{
  "table": "general_list_items",
  "record": {
    "id": "seed-item-g1",
    "list_id": "seed-list-grocery-2",
    "position": 0,
    "notes": null,
    "created_at": {
      "$relativeDays": -3
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null,
    "title": "milk",
    "state": "active",
    "priority": null,
    "tags": [],
    "url": null,
    "due_date": null,
    "review_date": null,
    "source_utterance": null
  }
}

Seeded value

{
  "table": "general_list_items",
  "record": {
    "id": "seed-item-g2",
    "list_id": "seed-list-grocery-2",
    "position": 1,
    "notes": null,
    "created_at": {
      "$relativeDays": -3
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null,
    "title": "eggs",
    "state": "active",
    "priority": null,
    "tags": [],
    "url": null,
    "due_date": null,
    "review_date": null,
    "source_utterance": null
  }
}

Seeded value

{
  "table": "general_list_items",
  "record": {
    "id": "seed-item-g3",
    "list_id": "seed-list-grocery-2",
    "position": 2,
    "notes": null,
    "created_at": {
      "$relativeDays": -3
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null,
    "title": "bread",
    "state": "active",
    "priority": null,
    "tags": [],
    "url": null,
    "due_date": null,
    "review_date": null,
    "source_utterance": null
  }
}

snapshot captured 2026-08-02T21:29:55.968Z

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 update the entry"edit the grocery list change two percent milk to oat milk" (difficulty —)0.1s

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

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

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

1 · TTS said I'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 2026-08-02T21:30:07.178Z

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 "2% 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 0.1 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 update the entry"change call insurance to call the insurance company about the claim" (difficulty —)0.2s

Verdict Expected UPDATE — should update the entry. PASS: Notes UPDATE write-truth verified (changed:todo_items×1).

Why verdict Notes UPDATE write-truth verified (changed:todo_items×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Updated to-do: call the insurance company.

2 · Card shown Updated to-do: call the insurance company.

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 2026-08-02T21:30:18.487Z

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 update the entry"change the dentist reminder to next wednesday" (difficulty —)0.2s

Verdict Expected UPDATE — should update the entry. FAIL: WRITE-TRUTH FAIL — UPDATE touched notes-domain rows but none matched expected content (dentist reminder).

Why verdict WRITE-TRUTH FAIL — UPDATE touched notes-domain rows but none matched expected content (dentist reminder).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Updated to-do: dentist appointment.

2 · Card shown Updated to-do: dentist appointment.

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 2026-08-02T21:30:29.810Z

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 update the entry"update the note about the contractor start date to the twentieth" (difficulty —)0.2s

Verdict Expected UPDATE — should update the entry. PASS: Notes UPDATE write-truth verified (changed:personal_notes×4).

Why verdict Notes UPDATE write-truth verified (changed:personal_notes×4).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.2s

1 · TTS said I could not confirm that was saved, so I did not mark it as done.

2 · Card shown I could not confirm that the change was saved, so I did not mark it as done.

4 · UI did navigated to "notes" screen

5 · Seed/context data

Seeded starting state:

Seeded value

{
  "table": "personal_note_categories",
  "record": {
    "id": "seed-note-cat-general",
    "name": "General",
    "created_at": {
      "$relativeDays": -14
    }
  }
}

Seeded value

{
  "table": "personal_notes",
  "record": {
    "id": "seed-note-contractor",
    "category_id": "seed-note-cat-general",
    "category_name": "General",
    "note_text": "Contractor start date is the fifteenth",
    "pinned_to_widget": false,
    "created_at": {
      "$relativeDays": -7
    },
    "deleted_at": null,
    "deleted_by_request_id": null
  }
}

snapshot captured 2026-08-02T21:30:41.158Z

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 "Dates" 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 update the entry"edit that note" (difficulty —)0.1s

Verdict Expected UPDATE — should update the entry. PASS: Notes UPDATE write-truth verified (changed:personal_notes×1).

Why verdict Notes UPDATE write-truth verified (changed:personal_notes×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 note should I edit: the parking thing on main street also grab oat milk; the client prefers email over phone; the gym locker code is four seven two nine?

2 · Card shown Which note should I edit: the parking thing on main street also grab oat milk; the client prefers email over phone; the gym locker code is four seven two nine?

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:30:52.342Z

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)

PASSshould update the entry"no I said bread not butter" (difficulty —)0.2s

Verdict Expected UPDATE — should update the entry. PASS: Notes UPDATE write-truth verified (changed:general_list_items×1).

Why verdict Notes UPDATE write-truth verified (changed:general_list_items×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Updated the list item on Shopping: bread.

2 · Card shown Updated the list item on Shopping: bread.

4 · UI did navigated to "lists" 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 2026-08-02T21:31:26.124Z

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 update the entry"actually make that almond milk not regular milk" (difficulty —)0.0s

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

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

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.0s

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 2026-08-02T21:31:37.316Z

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 "Unsweetened almond 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 0.1 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 update the entry"add to the note about the wifi password it changes every month" (difficulty —)0.3s

Verdict Expected UPDATE — should update the entry. PASS: Notes UPDATE write-truth verified (changed:personal_notes×1).

Why verdict Notes UPDATE write-truth verified (changed:personal_notes×1).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.3s

1 · TTS said Updated your note in General about the wifi password is sunflower22.

2 · Card shown Updated your note in General about the wifi password is sunflower22.

4 · UI did navigated to "notes" screen

5 · Seed/context data

Seeded starting state:

Seeded value

{
  "table": "personal_note_categories",
  "record": {
    "id": "seed-note-cat-home",
    "name": "Home",
    "created_at": {
      "$relativeDays": -14
    }
  }
}

Seeded value

{
  "table": "personal_notes",
  "record": {
    "id": "seed-note-wifi",
    "category_id": "seed-note-cat-home",
    "category_name": "Home",
    "note_text": "Wifi password is blueharbor42",
    "pinned_to_widget": false,
    "created_at": {
      "$relativeDays": -7
    },
    "deleted_at": null,
    "deleted_by_request_id": null
  }
}

snapshot captured 2026-08-02T21:31:48.736Z

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 update the entry"move eggs to the top of my grocery list" (difficulty —)2.7s

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

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

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 2.7s

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 2026-08-02T21:32:02.568Z

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

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

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

FAILshould update the entry"move milk below bread on the shopping list" (difficulty —)0.0s

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

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

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.0s

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 2026-08-02T21:32:13.784Z

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 update the entry"move passport to the top of the packing list" (difficulty —)0.0s

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

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

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.0s

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 2026-08-02T21:32:36.599Z

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 update the entry"swap the order of eggs and milk on groceries" (difficulty —)0.6s

Verdict Expected UPDATE — should update the entry. FAIL: WRITE-TRUTH FAIL — misroute-into-food: UPDATE on notes/todo/list wrote food_log_entries.

Why verdict WRITE-TRUTH FAIL — misroute-into-food: UPDATE on notes/todo/list wrote food_log_entries.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Logged Egg and Milk. Assumed catalog default servings where you did not say an amount — tell me if that is not right.

2 · Card shown Logged Egg and Milk. Assumed catalog default servings where you did not say an amount — tell me if that is not right.

3 · App data rows written created food_log_entry: Egg ×1 (50 g) 72 cal · 6.3g P · 0.4g C · 4.8g F
created food_log_entry: Milk ×1 (100 g) 50 cal · 3.3g P · 4.8g C · 2g F

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:32:48.372Z

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

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

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

FAILshould update the entry"move that to the top" (difficulty —)1.7s

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

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

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 1.7s

1 · TTS said I'm not sure which item you'd like me to move to the top. Could you please specify the list or to-do item you're referring to?

2 · Card shown I'm not sure which item you'd like me to move to the top. Could you please specify the list or to-do item you're referring 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 2026-08-02T21:33:01.200Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 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 update the entry"move milk down" (difficulty —)5.2s

Verdict Expected UPDATE — should update the entry. FAIL: Did not update — misroute-into-food.

Why verdict Did not update — misroute-into-food.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 5.2s

1 · TTS said What fat level or type of milk should I use? I did not log it yet because milk fat level can change the nutrition.

2 · Card shown What fat level or type of milk should I use? I did not log it yet because milk fat level 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 2026-08-02T21:33:31.219Z

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 answer the query"what's on my grocery list" (difficulty —)0.1s

Verdict Expected QUERY — should answer the query. PASS: Query answered (notes/list read path).

Why verdict Query answered (notes/list read path).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Grocery List: bread.

2 · Card shown Grocery List: bread.

4 · UI did navigated to "lists" 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 2026-08-02T21:33:54.609Z

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 answer the query"what's on my to-do list" (difficulty —)2.2s

Verdict Expected QUERY — should answer the query. PASS: Query answered (notes/list read path).

Why verdict Query answered (notes/list read path).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Your to-do list currently has: Grocery List: bread.

2 · Card shown Your to-do list currently has: Grocery List: bread.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:34:07.948Z

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 answer the query"what do I still need from the store" (difficulty —)2.0s

Verdict Expected QUERY — should answer the query. PASS: Query answered (notes/list read path).

Why verdict Query answered (notes/list read path).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 From your lists, you still need to pick up bread from the store.

2 · Card shown From your lists, you still need to pick up bread from the store.

4 · UI did no screen events captured

5 · Seed/context data

Seeded starting state: none listed in the corpus row.

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

snapshot captured 2026-08-02T21:34:21.045Z

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 answer the query"read me my packing list" (difficulty —)0.1s

Verdict Expected QUERY — should answer the query. PASS: Query answered (notes/list read path).

Why verdict Query answered (notes/list read path).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

1 · TTS said I could not find a list matching me my packing.

2 · Card shown I could not find a list matching me my packing.

4 · UI did navigated to "lists" 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 2026-08-02T21:34:32.293Z

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 0.1 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 answer the query"what are my most important to-dos" (difficulty —)0.1s

Verdict Expected QUERY — should answer the query. PASS: Query answered (notes/list read path).

Why verdict Query answered (notes/list read path).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Most important to-dos: water the plants (P3, due Aug 1); renew my car registration (P3, due Aug 2); take out the recycling (P3, due Aug 2); dentist appointment (P3, due Aug 4); finish the slide deck (P3, due Aug 5).

2 · Card shown Most important to-dos: water the plants (P3, due Aug 1); renew my car registration (P3, due Aug 2); take out the recycling (P3, due Aug 2); dentist appointment (P3, due Aug 4); finish the slide deck (P3, due Aug 5).

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 2026-08-02T21:34:43.483Z

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 answer the query"how many things are on the beach list" (difficulty —)0.1s

Verdict Expected QUERY — should answer the query. 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 2026-08-02T21:34:54.766Z

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 answer the query"do I have milk on my grocery list" (difficulty —)2.6s

Verdict Expected QUERY — should answer the query. FAIL: Query not answered — misroute-into-food.

Why verdict Query not answered — misroute-into-food.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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'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 2026-08-02T21:35:08.543Z

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 answer the query"what's left on weekend projects" (difficulty —)3.5s

Verdict Expected QUERY — should answer the query. FAIL: Query not answered — asked back instead of answering.

Why verdict Query not answered — asked back instead of answering.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 I’m not seeing a list named “weekend projects” in your to-dos. Would you like to create one, or check another list?

2 · Card shown I’m not seeing a list named “weekend projects” in your to-dos. Would you like to create one, or check another 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 2026-08-02T21:35:39.560Z

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 answer the query"show me what I added to groceries today" (difficulty —)0.0s

Verdict Expected QUERY — should answer the query. 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 0.0s

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 2026-08-02T21:35:50.589Z

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 answer the query"what's on that list" (difficulty —)0.1s

Verdict Expected QUERY — should answer the query. FAIL: Query not answered — asked back instead of answering.

Why verdict Query not answered — asked back instead of answering.

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

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

Timing 0.1s

1 · TTS said I could not find a list matching that.

2 · Card shown I could not find a list matching that.

4 · UI did navigated to "lists" 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 2026-08-02T21:36:01.775Z

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 answer the query"find my note about the car" (difficulty —)0.1s

Verdict Expected QUERY — should answer the query. PASS: Query answered (notes/list read path).

Why verdict Query answered (notes/list read path).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 saved note matching that phrase.

2 · Card shown I could not find a saved note matching that phrase.

4 · UI did navigated to "notes" screen

5 · Seed/context data

Seeded starting state:

Seeded value

{
  "table": "personal_note_categories",
  "record": {
    "id": "seed-note-cat-car",
    "name": "Auto",
    "created_at": {
      "$relativeDays": -14
    }
  }
}

Seeded value

{
  "table": "personal_notes",
  "record": {
    "id": "seed-note-car",
    "category_id": "seed-note-cat-car",
    "category_name": "Auto",
    "note_text": "Car registration due in August; VIN on dash",
    "pinned_to_widget": false,
    "created_at": {
      "$relativeDays": -7
    },
    "deleted_at": null,
    "deleted_by_request_id": null
  }
}

snapshot captured 2026-08-02T21:36:24.041Z

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 answer the query"search my notes for wifi password" (difficulty —)0.1s

Verdict Expected QUERY — should answer the query. PASS: Query answered (notes/list read path).

Why verdict Query answered (notes/list read path).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Notes. Latest matching note: the wifi password is sunflower22 it changes every month

2 · Card shown Opening Notes. Latest matching note: the wifi password is sunflower22 it changes every month

4 · UI did navigated to "notes" screen

5 · Seed/context data

Seeded starting state:

Seeded value

{
  "table": "personal_note_categories",
  "record": {
    "id": "seed-note-cat-wifi-search",
    "name": "Home",
    "created_at": {
      "$relativeDays": -14
    }
  }
}

Seeded value

{
  "table": "personal_notes",
  "record": {
    "id": "seed-note-wifi-search",
    "category_id": "seed-note-cat-wifi-search",
    "category_name": "Home",
    "note_text": "Wifi password is blueharbor42",
    "pinned_to_widget": false,
    "created_at": {
      "$relativeDays": -7
    },
    "deleted_at": null,
    "deleted_by_request_id": null
  }
}

snapshot captured 2026-08-02T21:36:35.360Z

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 answer the query"do I have a note about the contractor" (difficulty —)0.1s

Verdict Expected QUERY — should answer the query. PASS: Query answered (notes/list read path).

Why verdict Query answered (notes/list read path).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Notes. Latest matching note: the contractor said they start on the fifteenth

2 · Card shown Opening Notes. Latest matching note: the contractor said they start on the fifteenth

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 2026-08-02T21:36:46.583Z

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 answer the query"find notes mentioning blood pressure" (difficulty —)0.1s

Verdict Expected QUERY — should answer the query. PASS: Query answered (notes/list read path).

Why verdict Query answered (notes/list read path).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

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

Report/grader issue

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

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Notes. Latest matching note: my blood pressure was one twenty over eighty this morning

2 · Card shown Opening Notes. Latest matching note: my blood pressure was one twenty over eighty this morning

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 2026-08-02T21:36:57.832Z

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 answer the query"look up my note about Jake's jersey size" (difficulty —)0.1s

Verdict Expected QUERY — should answer the query. PASS: Query answered (notes/list read path).

Why verdict Query answered (notes/list read path).

My override

Things I think are wrong

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

Comprehension

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

Execution

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

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

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

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Notes. Latest matching note: Jake's jersey size is youth large

2 · Card shown Opening Notes. Latest matching note: Jake's jersey size is youth large

4 · UI did navigated to "notes" screen

5 · Seed/context data

Seeded starting state:

Seeded value

{
  "table": "personal_note_categories",
  "record": {
    "id": "seed-note-cat-kids",
    "name": "Kids",
    "created_at": {
      "$relativeDays": -14
    }
  }
}

Seeded value

{
  "table": "personal_notes",
  "record": {
    "id": "seed-note-jake-jersey",
    "category_id": "seed-note-cat-kids",
    "category_name": "Kids",
    "note_text": "Jake's jersey size is youth large",
    "pinned_to_widget": false,
    "created_at": {
      "$relativeDays": -7
    },
    "deleted_at": null,
    "deleted_by_request_id": null
  }
}

snapshot captured 2026-08-02T21:37:09.042Z

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 answer the query"search todos for anything about insurance" (difficulty —)0.1s

Verdict Expected QUERY — should answer the query. 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:

Seeded value

{
  "table": "todo_categories",
  "record": {
    "id": "seed-todo-cat-admin",
    "name": "Admin",
    "created_at": {
      "$relativeDays": -14
    }
  }
}

Seeded value

{
  "table": "todo_items",
  "record": {
    "id": "seed-todo-insurance",
    "category_id": "seed-todo-cat-admin",
    "category_name": "Admin",
    "name": "Renew car insurance",
    "description": null,
    "priority": 3,
    "scheduled_date": null,
    "due_date": null,
    "completed_at": null,
    "created_at": {
      "$relativeDays": -7
    },
    "updated_at": {
      "$relativeDays": -1
    },
    "deleted_at": null,
    "deleted_by_request_id": null
  }
}

snapshot captured 2026-08-02T21:37:20.334Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"log that I finished the quarterly report" (difficulty —)0.1s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of writing the note/todo/list (no saved row).

Why verdict OVER-ASK — asked instead of writing the note/todo/list (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I could not find a to-do matching quarterly report. Did you mean: dentist appointment; call the insurance company about the claim; Expense report?

2 · Card shown I could not find a to-do matching quarterly report. Did you mean: dentist appointment; call the insurance company about the claim; Expense report?

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 2026-08-02T21:37:42.933Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"note for work I sent the revised contract to legal" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: I sent the revised contract to legal.

Why verdict Notes write-truth verified: I sent the revised contract to legal.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Work about I sent the revised contract to legal. Say if that category should be different.

2 · Card shown Logged your note in Work about I sent the revised contract to legal. Say if that category should be different.

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 2026-08-02T21:37:54.293Z

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 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"log that standup ran long today" (difficulty —)2.3s

Verdict Expected LOG — should log the entry. FAIL: Did not write — misroute-into-food; no notes-domain saved row.

Why verdict Did not write — misroute-into-food; no notes-domain saved row.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 2.3s

1 · TTS said What exact food and amount should I use for that standup ran long today? 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 that standup ran long today? 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 2026-08-02T21:38:07.702Z

6 · Why this food No food-decision explanation was captured for this path.

7 · Lookup method No lookup method was captured for this path.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"record that I completed the client onboarding call" (difficulty —)0.1s

Verdict Expected LOG — should log the entry. FAIL: OVER-ASK — asked instead of writing the note/todo/list (no saved row).

Why verdict OVER-ASK — asked instead of writing the note/todo/list (no saved row).

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.1s

1 · TTS said I could not find a to-do matching client onboarding call. Did you mean: dentist appointment; call the insurance company about the claim; Expense report?

2 · Card shown I could not find a to-do matching client onboarding call. Did you mean: dentist appointment; call the insurance company about the claim; Expense report?

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 2026-08-02T21:38:18.897Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

PASSshould log the entry"jot for work budget review moved to next tuesday" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. PASS: Notes write-truth verified: budget review moved to next tuesday.

Why verdict Notes write-truth verified: budget review moved to next tuesday.

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple 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 Work about budget review moved to next tuesday. Say if that category should be different.

2 · Card shown Logged your note in Work about budget review moved to next tuesday. Say if that category should be different.

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 2026-08-02T21:38:30.259Z

7 · Lookup method

The report captured the app food lookup that ran before or alongside the final decision.
It searched planned foods, your saved foods and aliases, your past food logs, the shared nutrition database, the common-food list, the exercise list, and the workout equipment list.
The lookup evidence took 0.1 seconds to collect.
This lookup is supporting evidence in the report; by itself it does not prove that food was saved.

8 · Candidates → decider no candidate list captured (this path did not run DB resolution — e.g. provisional branded log)

FAILshould log the entry"log I shipped the hotfix at two fifteen p m" (difficulty —)0.2s

Verdict Expected LOG — should log the entry. FAIL: WRITE-TRUTH FAIL — expected notes/todo/list write missing — local notes-domain delta is 0 for this turn (proven no write, not a capture gap)

Why verdict WRITE-TRUTH FAIL — expected notes/todo/list write missing — local notes-domain delta is 0 for this turn (proven no write, not a capture gap)

My override

Things I think are wrong

Use this when the report passed something that should have failed.

Comprehension

Picked wrong food Wrong quantity Should have asked (over-logged) Shouldn't have asked (over-asked) Wrong action entirely

Execution

Didn't save to DB Wrong data saved No voice/TTS Card missing/wrong No chat

Cosmetic

Wording weird

Other thing wrong (describe below)

Things the report marked wrong but I think are right

Use this when the report failed something that should have passed.

Actually right

Food/action was right Quantity/serving was right Asking was right Not asking was right Answer was right No app data row was expected App data row was right Voice/card/chat was acceptable

Report/grader issue

Report expected the wrong result Evidence was enough to judge Failure reason was not valid

Other thing actually right (describe below)

Pick every reason you think applies. Multiple boxes are allowed because one utterance can be wrong in more than one way. Use the note below for "Other" or extra detail.

Timing 0.2s

1 · TTS said Logged shipped hotfix: 2 lb for 15 reps.

2 · Card shown Logged shipped hotfix: 2 lb for 15 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 2026-08-02T21:38:41.543Z

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