# n215 — The reconciliation, the corrected map, and the verdict: leg 1 FAILS — the frozen wording fires

**Date:** 2026-07-11 · **Resolves n214's INCONCLUSIVE.** Two-backend reconciliation per audit §5; the bug found by inspection, verified live, repaired fail-closed.

## The bug (in the n202 Rust)

A stray normalization loop half-mutated the x-components of diagonal direction vectors ((1,1,0) became (0.707,1,0)) while its comment claimed "keep lattice steps as-is." The cubic configuration had no diagonals — its 0.51 was clean; **every 9- and 13-complex number in n202/n203 was contaminated.** Validation chain: Python matches clean-Rust on the cubic (0.5129 vs 0.5130); corrected Rust matches Python on the 9-derived complex (1.449 vs 1.44). The backends now agree; correction headers are on both notes (rule 5).

## The corrected map — worse, everywhere

| complex | ratio c_g²/c² (was, contaminated) | **corrected** |
|---|---|---|
| cubic | 0.51 | **0.51 — now the family MAXIMUM** |
| 9 uniform / derived / axes-0.25 / fcc | 0.70 / 0.81 / 0.85 / 0.83 | **0.38 / 0.32 / 0.26 / 0.27** |
| 13-complexes | 0.40 / 0.34 | 0.34 / 0.31 |

The "0.85 climb toward isotropy" never existed. **Isotropy makes the deficit worse.** No scanned complex reaches even 52% of the cone. The corrected bargain (for the record; already withdrawn): b\* = 2–3× the induced stiffness on the 9-family — no longer a modest fraction.

## Leg 1, re-adjudicated under the frozen gate

INCONCLUSIVE is resolved (backends agree; grid convergence within the declared clause). The findings: the undressed ratio is ≤ 0.51 across the entire scanned family, and the only independently motivated dressing — the guard, Newton's own back-action — **monotonically lowers it further** (trend robust at three grid resolutions, both backends' machinery). The curve c_g²(s_g)/c\*²(s_g) never approaches the PASS band [0.98, 1.02] from anywhere, for any admissible s_g ≥ 0.

**LEG-1: FAIL.** Per the gate frozen before execution, the consequence is not negotiable and fires verbatim:

> **The radiative sector is dead at every computed order; the GR-equivalence claim is withdrawn.**

FP-gate condition B fires with it. Alternative dressings may be proposed only as new preregistered gates (the no-escape clause) — none is currently on the table.

## What died and what stands

**Dead:** the graviton program — the claim that this substrate's blind shear channel is, or dresses into, GR's radiation sector. With it: the GR-equivalence reading of the gauge-fixed structure (n213's merge cut both ways — one number decided both constructions); predictions whose backing was the radiative recovery (17's "pure tensor, locked to light" loses its framework backing and is re-tagged; 31's covariant-strength language inherits the flag). The triple-consistency knot is moot — there is nothing left for it to normalize.

**Standing, untouched by this death:** the trunk (dark states now derivation-grade, n211; the knee family as monitor-model predictions; the odd/even analog test); the static sector (Coulomb-class force, the constraint-owned pole); Wing B's topology conjectures and Wing C's cosmological wagers (never depended on the graviton); and the honest mathematical results of the radiative campaign itself — the TT projector, the gauge-fixed Dirac structure, the speed maps — which now document *why* this class of models cannot radiate like Einstein: **the monitor pours stiffness into what it watches, and the blind channel it leaves behind is always too soft.** That sentence is the sector's epitaph, and it is a theorem-shaped one.

**Class: reconciliation [A — verified two-backend]; the corrected map [A-model]; the verdict [frozen-gate execution, no discretion exercised].**