[ryoji-info]

Promoted by maintainer direction (2026-10-17), not by a clean vote. The agent crew ran the full pipeline on Fable Session — 2026-10-16, discussion #137 and voted 3/3 to store in every round (round ladder: 6 / 5 / 5, 4 / 6 / 3, 2 / 4 / 4, 2 / 3 / 2 required edits, 🧵 / 🌊 / ⚛️ order; round 4 = the pipeline's hard cap, run as a scope reduction). Twelve fresh, mutually blind assessor seats on claude-opus-5-5 plus two pre-publication check seats on claude-fable-5-1; execution and every rework round in-session on claude-fable-5-1. Every claimed defect was re-executed before acceptance; no required edit was declined in any round (two pre-publication items were declined on the record / on measurement, dispositions). Per the routine the crew opened nothing at the cap; the maintainer directed promotion (the PR #98 / #131 / #136 precedents); the agents:approved-2of3 label records the store majority, not a clean pass, and the note's Status line discloses the direction.

This PR adds notes/2026-10-16-no-density-route-to-a-not-in-the-released-cell.md — the round-4 reply text (part 1, part 2) with the two corrections of its erratum applied (two stale copies of a superseded phase-window band, disclosed in the Status line) and the corrections of its post-cap supplement applied (the round-4 seats' required edits and accepted residues, each re-executed and listed there; 30 exact replacements, listing 6; the round-4 tally discloses the seat loss and relaunch), with the round-4 vote record appended verbatim. Per the 2026-07-26 finding-5 rule it maintains the navigation layer in the same PR: the note's notes/INDEX.md row, updates to the 07-15, 07-17, 07-21, 08-02, 10-08 and 10-13 rows it bears on, dated post-promotion annotations appended to those six notes (line endings preserved per file), and one notes/CITATIONS.md row (Dyakonov & Shur 1996, recalled — the closed form was derived independently and the note does not rest on it). No results.json value is contested and no promoted key moves.

What changed between rounds, and what was withdrawn. Round 2: the −0.20 threshold depletion scoped to the absolute leg (0.80·M_c; G1's pin −0.30); Eq. (5)'s Δn mapped to the tanh scale (×2.197); (C′) regraded qualified; the drain-density observable, the NOT's direct 10–90 % rows (k = 1.31), the measured-curve butterfly, the decay estimators, the AND grids and the ratio leg added. Round 3: the all-'1' burst train disclosed and the isolated-'1' residual (0.29–0.35) added; the AND's 10 % point (0.0018, k = 1.34); D-INV-2 by observable; the friction-family scope of the G1-pin crossing (−0.21 at n^+½, beyond −0.54 at n^−½); the β ladder's fitted order (0.75 → 1.22); the sign of rectification shown to be the source's convective flux against the drain's pressure curvature (23.19α − 11.30P″), not the clamps. Round 4, the cap: withdrawn — the "tristable at 7.5°" statement and every rails figure (interpolant artifacts; on direct two-tone rows the 7.5° pair is monostable), the registered key inv_fieldsum_not_snm (grid- and interpolant-dependent; open in the continuum), the "steady" oscillator readings (replaced by 60-round-trip window means and ranges from long records), the unpinned width construction (pinned to δ_off = −0.20), the universal depletion band (scoped: a held-amplitude input at fixed drain current never reaches 10 dB on the velocity observables) and the fixed-drift friction universal (false for n^+½). Nothing in the verdict moved in any round: no density route to a NOT exists in the released cell, rectification accumulates, the only NOT is interference with a 10–12° window and at most marginal restoration, the field-sum AND separates by 2.6–3.3 dB, the cell as its own pump is pulled, not quenched.

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

🤖 Generated with Claude Code