Skip to content

Optional second round: --revise shows every lane the locked round-1 answers - #65

Merged
CryptoJones merged 3 commits into
mainfrom
feature/revise-round
Sep 1, 2026
Merged

Optional second round: --revise shows every lane the locked round-1 answers#65
CryptoJones merged 3 commits into
mainfrom
feature/revise-round

Conversation

@CryptoJones

Copy link
Copy Markdown
Owner

Closes #64.

CJ's ask: a way for the roundtable lanes to have access to each other's answers — with round 1 staying isolated and the second round strictly optional.

What this does

  • --revise TRANSCRIPT reruns the panel with each lane shown the locked prior answers. Its own is marked YOURS; peers are anonymised as PANELIST A/B/C (attribution invites exactly the authority-weighting the instructions forbid). Lanes must open HOLD or REVISE and move only for a reason they can state.
  • Accepts latest, latest:N, and comma-separated paths, and merges them — --each writes one transcript per lane, so "the previous round" of a 12-lane run is 12 files. Refuses to mix transcripts whose briefs differ (that is only ever an accident). An answer never loses to a later failure of the same lane.
  • latest resolves to concrete paths in the parent before --each fans out — a child resolving it for itself would pick up a sibling's round-2 transcript, the exact self-echo this feature must not create.
  • The report banners the round: "Agreement here is persuasion, not independent convergence", and tallies HOLD vs moved. The transcript records round, parent paths, and the alias map, so anonymity is auditable. Round 3+ works by pointing --revise at a round-2 transcript.
  • Round-1 behaviour is untouched; calibration is skipped on revision rounds (per-lane packets differ in length).

Verification

  • 132 tests pass (10 new: transcript merge, brief-mismatch refusal, alias skipping failed lanes, packet anonymity, NEW path for absent lanes, verdict parsing, and an end-to-end --revise against the stub server asserting the packet reaches the wire and the banner reaches the report).
  • ./roundtable --list — zero network calls, correct routes.
  • README, skill, and BACKLOG updated; README's stale test count (25 → 132) fixed in passing.

🤖 Generated with Claude Code

https://claude.ai/code/session_01VmauK5UCYGRYyXoQ36FQ3S

Proudly Made in Nebraska. Go Big Red! 🌽 https://xkcd.com/2347/

…nswers (#64)

Round 1 stays blind by construction — the packet a lane receives is built from a finished run's transcript, so nothing can leak between live lanes. --revise accepts latest, latest:N, and comma-separated paths, merging the per-lane transcripts --each writes (an answer never loses to a later failure of the same lane) and refusing to mix differing briefs. Each lane sees its own answer marked YOURS and peers anonymised as PANELIST letters — attribution invites the authority-weighting the instructions forbid — and must open HOLD or REVISE with a stated reason. The report banners the round as persuasion rather than independent convergence and tallies who moved; the transcript records round, parents, and the alias map so anonymity is auditable. Calibration is skipped on revision rounds because per-lane packets differ in length.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Claude-Session: https://claude.ai/code/session_01VmauK5UCYGRYyXoQ36FQ3S
@coderabbitai

coderabbitai Bot commented Sep 1, 2026

Copy link
Copy Markdown

Important

  • 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Team

Run ID: eda539a8-211e-4899-b1dc-73ac9bacc9de


Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

…st live use

Retrying one failed lane of a revision round with latest:N swept in the siblings' round-2 transcripts. The brief guard cannot catch that — a revision round records the same brief — and the merge would hand the lane its peers' round-2 answers labelled as the round it is trying to redo. Transcripts spanning different rounds now refuse, naming the rounds; retry with the files of one round listed explicitly.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Claude-Session: https://claude.ai/code/session_01VmauK5UCYGRYyXoQ36FQ3S
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Claude-Session: https://claude.ai/code/session_01VmauK5UCYGRYyXoQ36FQ3S
@CryptoJones
CryptoJones merged commit 78194cc into main Sep 1, 2026
4 checks passed
@CryptoJones
CryptoJones deleted the feature/revise-round branch September 1, 2026 18:31
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Optional second round: lanes see each other's locked answers (--revise)

1 participant