# PASS2 supplement audit — issue 24221

**Supplement correct. Overall proof verdict remains INSUFFICIENT.** Fresh read-only supplement check; no full-audit rerun and no new implementation verdict.

Reviewed the [supplement](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/73d14a46d38b107a/evidence-supplement.md) against the [original PASS2 audit](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/43ed7e5ad3f10d5e/audit.md). Individually opened and visually inspected all four local DiffT16 PNGs with view_image. Independently decoded the selected original Before/After RGB PNGs with Pillow, converted to signed integers before subtraction, and computed white iff max(abs(RGBbefore−RGBafter)) > 16; black otherwise. No resize, alignment, crop, or amplification.

| Pair | Original Before → After filenames | Raster | White pixels | Percent | Exact RGB mask match |
|---|---|---|---|---|---|
| 1440-light | V1-before-1440light.png → V1-after-1440light.png | 1280×800 | 669 | 0.0653% | Yes |
| 1440-dark | V1-before-1440-dark-0.png → V1-after-1440-dark-0.png | 1280×800 | 665 | 0.0649% | Yes |
| 390-light | V1-before-390-light-final-0.png → V1-head-390-light-verified-0.png | 780×1688 | 4,255 | 0.3232% | Yes |
| 390-dark | V1-before-390-dark-final-0.png → V1-head-390-dark-verified-0.png | 780×1688 | 4,210 | 0.3198% | Yes |

All masks contain only RGB 0/255. Exactly-16 differences are excluded (respectively 0, 2, 16, 5 pixels). Independent raw >0 counts remain 811/824/8710/8787, consistent with the original audit. Visual inspection and computed bounding boxes place every retained pixel in the Send-button region: inclusive bounds respectively (923,739)–(951,767), (924,739)–(951,767), (657,1558)–(730,1632), (658,1558)–(729,1631). The narrow header differences are filtered at this threshold. These masks are correctly distinguished from the existing amplified Diff0; thresholding does not prove absence of smaller differences or add behavioral proof.

No misleading change claim found. Runtime identity is correctly disclosed as base [d53873b6a4](https://git.cafe/pingdotgg/t3code/commit/d53873b6a4), captured after [09c6299d0d](https://git.cafe/pingdotgg/t3code/commit/09c6299d0d), final test-only head [83d50c1a8a](https://git.cafe/pingdotgg/t3code/commit/83d50c1a8a). This supplement audit inherits the original audit's runtime-code-equivalence finding; it does not independently rerun source provenance checks.

The original accepted web six-second timeout/prior-Success base-disabled/head-enabled evidence, imperative rename, and aggregate evidence retain their prior scope. All matrix rows remain partial/incomplete: native and actual Electron absent; full C1 execution/destination/revocation partition and negative UI fixtures missing; clock freezing and historical per-frame viewport telemetry missing; V3 narrow-light 127-pixel repeat instability; V3 recovery log-supported without settled visual ending; empty history cannot prove history preservation; no base aggregate comparison/native aggregate reader verification. Threshold masks close none of these gaps.

## FRICTION and boundaries

No new tool friction: local images and originals were readable and pixel recomputation succeeded. Public artifact bytes were not downloaded or revalidated in this supplement-only pass. Existing capture/build/telemetry friction and all original exclusions remain. No browser, server, source edit, tests, forge action, PR, push, public comment, or merge. Only this outside-repository report was written and uploaded through ghdrop.

Model/harness: gpt-6.1-sol / Codex in T3 Code.