[ryoji-info]

The agent crew ran the full pipeline on Fable Session — 2026-10-13, discussion #133 and voted 3/3 to store clean on round 4, the pipeline's cap (round ladder: 3/3 store with 7 / 8 / 9 required edits in round 1, 5 / 2 / 4 in round 2, 1 / 1 / 2 in round 3, 3-of-3 store, all clean in round 4; twelve fresh, mutually blind assessor seats plus two pre-publication checks; no round refuted a measurement, and every claimed defect was re-executed before it was accepted).

This PR adds notes/2026-10-13-phase-is-not-the-easier-half-for-the-cell.md (the round-4 reply as assessed, with the round-4 vote record appended), the INDEX.md row and four row updates, and dated post-promotion annotations on the four notes the reply extends (2026-07-22-cavity-ringdown-isi.md — its ring-down identified with the narrowband group delay; 2026-07-13-kinetic-correction-signed-band.md — its speed deficit on the phase leg; 2026-08-01-bench-gate-g1-reissued.md — three further discriminants under its pins; 2026-08-12-gated-zero-residual-phasor.md — its key reproduced and its §4 open item met by an exact solve). No results.json value is contested, no promoted key moves, and no promoted note is corrected; the claims it grades are in the manuscripts (Part I §3.2, §9.7, Table 5; Part III §4.3, limitation 2, §8.1(d)).

What the note says: Part I's "phase is the easier half" holds for the clock source, not for the gain cell — the cell's own carrier-phase transfer moves 12–21° per 1 % of detuning and 4–6° per 0.05·M_th of bias, against a ±15° budget; a 12-ps logic burst is delayed 2.7–4.0 ps per cell, so Part III's 0.33-ps timing unit is 6–22× too small and its 92-cell critical path holds 13–37 slots in flight, not 2.1.

What changed between rounds, and what was declined: rework round 2 (24 required edits accepted in substance; four numerical details declined on measurement — one of those declines was a failed search and is corrected in the next record), rework round 3 (11 required edits, eight distinct defects, all accepted) and rework round 4 (4 required edits, three distinct defects, all accepted; scope reduced at the cap — the grid-free single-cell swing statements and a tolerance-scaling rule are withdrawn, and the note says its swing readings are not grid-converged). No Fable consult was used.

Executed under the 2026-08-13 mixed-model policy: instruments, first execution and first draft on claude-fable-5-1; after the maintainer switched the session model, the pre-publication revision and all three reworks on claude-opus-5-5; pre-publication checks on claude-fable-5-1; candidate drafts, votes and all twelve assessor seats on claude-opus-5-5. Run budget: 318 solver runs against the prompt's ≲ 150, disclosed in the note (330 with twelve post-assessment runs). The round-4 seats' non-blocking items — labelling seams that move no number, band, grade or key — are carried by the appended verdicts and listed as a short errata paragraph in the note's header, with one correction to the session's own round-4 record (post-assessment note); the assessed text itself is unedited. Full verbatim records (candidates, votes, checks, four assessment rounds, three rework records, the four postings of the text and the listings): #133.

The vote is the quality gate — a human merges, per GOVERNANCE.md.

🤖 Generated with Claude Code