A finished turn ends the same way on the Mac, the web and the iPhone. Answer Layout decides how it is laid out: Focused, the default, or Classic, the Summary table it replaced. Choose it in Settings → Appearance; it is shared with your other devices. Focused moves every Summary row somewhere; it never drops one.
Focused, from the top
- The title. The discussion's own name. Your message below it is shown exactly as you wrote it.
- Discussion · 4 earlier replies. One row per reply: the seat, a chip for what the turn did, one line in its words, and the time. Click a row to read it, or Expand Replies.
- Final Answer. The answer the run reached, with how it was reached at the right: Agreed Answer, Not Unanimous, No Agreement, and Revised After Review when the lead answered its reviewers.
- The heading. When the answer opens with one clear result, that first sentence becomes its heading. It is moved, not repeated; an answer that opens with a list, a question or a preamble gets none.
- Run Summary. One card under the answer: the verdict in a pill (No agreement · 0 of 3 agree, amber when it needs a look, red when it may be wrong) and one line of counts from the run's own record, for example 30 of 33 review commands passed · 5 not verified · 12 changed files. The whole card is one button: click it, or View Details, and Inspect opens on its Summary tab for that run. It never opens in place.
- Actions. Continue the Discussion and Decide It Yourself appear only when the room ended without agreement and the run can still go on — the latest run, not stopped by you, not failed or out of credit. Ask Grok again and Add Credit appear when they apply.
- The footer. Where it ran, the checkout and how long it took.
Inspect → Summary
The run's details, in sections. A section appears only when the run recorded something for it:
- Outcome — how it ended, what is still open or blocked, and why nothing was built.
- What Was Checked — the reviewers' commands, the app's own test and build commands, the evidence with each check, and where it was checked.
- Not Verified — each gap once, numbered; commands the sandbox refused fold under Blocked Attempts.
- Where They Disagreed and What Remains — the answer's own sections, moved here from the answer, word for word.
- Changed Files — grouped by what they are for (Code, Screens, Setup, Styles, Docs, Tests); each name still opens the file.
- Errors and Background Work — the background commands (a link to Background Tasks), each failing check's output exactly as stored, and the review's source references.
- Run Details — who led, stand-ins, tools, network, sandbox, environment and usage, with Per-Seat Account, and Project Memory at the foot.
Inspect shows the latest run; step back to an earlier run with the arrows beside Run 2 of 3. Each earlier run's card in the transcript opens its own run.
The seat chips
| Chip | What it means |
|---|---|
| Draft | The lead's first answer |
| Revised | The lead answering its reviewers |
| Agrees | The seat agreed with the answer |
| Disagrees | A reviewer that raised something instead of agreeing |
| Incomplete | The seat started but did not reach a verdict |
| Could Not Answer | The seat never answered — out of credit, timed out, or failed |
The chip comes from the run record, not from the reply's wording.
While the room works
Every reply stays open while the run goes, so there is always something to read while the next seat thinks. When the run ends, the earlier replies fold to their rows, the final answer stays open, and the Run Summary card appears. Channels keep their own short replies.
Classic
Classic shows the whole Summary card as before: every row in one table under the answer, with the change card above it and the timing pill below.