[ryoji-info]

Promoted by maintainer direction (2026-10-17), not by a clean vote. The agent crew ran the full pipeline on Fable Session — 2026-10-17, discussion #142 and voted 3/3 to store in every round (round ladder: 4 / 3 / 3, 2 / 2 / 1, 1 / 1 / 1, 2 / 1 / 2 required edits, 🧵 / 🌊 / ⚛️ order). 12 fresh, mutually blind assessor seats plus two pre-publication check seats; at the maintainer's direction every seat and the execution ran on claude-opus-5-5. Every claimed defect was re-executed before it was accepted; none was declined, and no round refuted a measurement.

This PR adds notes/2026-10-17-a-pumped-cell-restores-phase-in-a-narrow-window.md — the round-4 reply as posted, with a Status block, the correction history and the deciding round's verdicts and tally — plus:

• an INDEX row, and row updates on 08-05, 10-13 and 10-16;
• dated post-promotion annotations on those three notes (08-05's DC-drive no-escape scoped; 10-13's Table C row derived; 10-16's inv_rect_sign scoped to carrier-frequency inputs and its "only in-model NOT" scoped to the amplitude encoding);
• a CITATIONS row for Dyakonov & Shur, PRL 1993 (the released solver's docstring source; locator checked against Crossref).

What the note says (in-model, released Galilean cell): a 2f₀ pump makes the cell a phase-sensitive amplifier (8.7–9.3 dB at a 5 % source pump) that restores phase inside the velocity domain only at knee-level swings, and only weakly at the delivered '1'; above onset it is a latching parametric oscillator (its drain onset rising with refinement, continuum open) which, gated in 07-31's slot, amplifies a bit stream's flips ×1.75–3.15 over an unpumped cell but re-levels none.

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

🤖 Generated with Claude Code