# Meeting notes — the brief, the fixture and the scoring

Sent as one message with `meeting.m4a` attached through each tool's own upload (`.mp3` where a tool refuses `.m4a`), on the free tier, in a fresh chat (Claude: incognito). For Otter and Fireflies the recording is uploaded to the app and the brief is what their notes are judged against, since they have no chat to paste it into - their summary, action items and transcript are taken as delivered. If a tool cannot take audio at all, that is its answer; the transcript is not pasted in for it, because turning a recording into notes is the Task.

> Here's the recording of our team call from this morning, about three and a half minutes. Please give me:
> 1. A transcript with the speakers labelled.
> 2. A short summary.
> 3. Every decision we made.
> 4. Every action item, with who owns it and the due date.
> 5. Every number that was mentioned.
> Only what was actually said - do not add anything.

## The fixture

`meeting.m4a` (3 min 30 s, mono AAC; `meeting.mp3` is the same audio): a Q3 planning call for a fictional Lisbon shop, three speakers read by macOS voices from `script.md`, generated by `make_fixture.py`. The script is the answer key for the transcript; `ground-truth.md` lists what the notes must contain: 5 decisions, 1 thing explicitly not decided, 6 action items with owner and date, 10 numbers, 6 planted transcript checks (a name spelled out, a corrected number, a rejected price, two Portuguese words, a clarified fact).

## Scoring

- **Took the audio** - yes or no, and in which format. A tool that cannot take a recording scores nothing else.
- **Transcript** - delivered or not; speakers labelled and how many (3 is right); the 6 planted phrases found, of 6.
- **Decisions** - of 5; the parked Christmas boxes listed as a decision or an action counts as invented.
- **Action items** - of 6, an item counts only with the right owner and the right date; near misses recorded.
- **Numbers** - of 10.
- **Invented** - anything in the notes that was not said.
- Time to answer, whether the free tier refused or truncated, output form.