# n210 — The missing executables written, the knee gates re-registered and passed

**Date:** 2026-07-11 · Closing the last two entries of the defect manifest (audit gates 7.5, 33, 51).

## n159 executable (`n159_photon_audit.py`) — 9/9, fail-closed

Reproduces the note's static kernel facts exactly: the symbol turns signed at q\* (dev 10⁻¹⁴); dark kernel = transverse plane, dim 2, residual 4×10⁻¹⁷ on-axis; longitudinal watched at the quoted 0.299 ≈ 0.3. Scope printed in the script itself: static constraint-kernel facts only — no photon claimed (audit §51 carried into the code).

## n175 executable (`n175_propagator.py`) — 6/6, fail-closed, with a rule-5 re-quote

All three positivity legs hold: KL density ≥ 0 (manifest, verified on-grid), dΠ_E/dΩ² < 0 both channels, dΠ_E/dq² < 0 both channels — no ghost, no tachyon at the free bubble, as claimed. **Rule 5:** the note's unexecutable magnitudes ({+}: −0.0051..−0.0039) are replaced by the executable's (−0.0003, both channels — a different normalization/grid than whatever produced the note's numbers, which cannot now be reconstructed). The signs and positivity were the load-bearing content; the magnitudes are now whatever the reproducible script says.

## n48b re-registration (`n48b_reregistered.py`) — 4/4 as declared, plus the adjudication table

The four gates, declared in ω in the script header before the run: **P1** static zero (ratio 8×10⁻²⁸) · **P2** onset slope 1.9845 ∈ 2.00±0.05 · **P3** knee half-saturation 0.5009 · **P4** γ-scaling exponent −1.9956 ∈ −2.00±0.05 (the audit-corrected power). The original n48's failed-as-registered record stands untouched in the prereg file; this is a new registration, passed on its first and only run.

The audit's deeper point is answered with a number instead of a concession: the same protocol run on the two declared alternative monitors gives the **discriminability table** — a two-pole monitor deviates from the best-fit Lorentzian by 8.4% (max, over the sweep), an exponential-memory (non-Markovian) monitor by ~104%. So: *a real frequency sweep at ~5% shape precision adjudicates the Markovian single-pole postulate against the two-pole alternative; any sweep at all kills or confirms it against the memory-kernel class.* The knee's shape stops being merely "the inserted transfer function" and becomes a measurable choice among named monitor models.

## Manifest state

Repair queue: **empty.** Repaired/written: n113, n123, n136, n159, n175, n48b — six entries, dated. The constructive program (the three decisive constructions + sector gates) is all that remains.