# Evidence supplement — thresholded V1 differences

Issue [#24221](https://git.cafe/pingdotgg/t3code/issues/24221). This supplement adds DiffT16 for the four selected V1 pairs in the [original evidence index](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/edc6a551febb0526/index.md). The original index and all existing artifacts are retained unchanged.

Base runtime: [d53873b6a4](https://git.cafe/pingdotgg/t3code/commit/d53873b6a4). After runtime: [09c6299d0d](https://git.cafe/pingdotgg/t3code/commit/09c6299d0d). Final test-only commit: [83d50c1a8a](https://git.cafe/pingdotgg/t3code/commit/83d50c1a8a). Runtime-code equivalence is the lead's confirmation recorded in the original index; this supplement does not independently revalidate it.

Environment (from original index): captured 2026-10-24 UTC with product-native preview, own tabs. Lead-owned before web 8449/server 16489 and after web 6637/server 14677; isolated fresh worktree data, synthetic empty persisted thread and harmless unsent draft, no provider run. Migration 68/69 conflict prevented copying real seed. CSS viewports 1440×900 and 390×844, DPR 2. Wide snapshots normalize to 1280×800; narrow rasters are 780×1688. Follow-up innerWidth measurements confirmed 1440 on both tabs; historical V1 endpoints lack per-frame viewport measurements. Fonts settled, animation/transition/caret disabled, explicit light/dark; clock was not frozen.

Diff0 amplified reuses the existing zero-threshold absolute RGB difference amplified 8×. New DiffT16 is a binary mask computed directly from the same original RGB endpoints, without resizing, alignment, cropping, or amplification: white iff max(|Rbefore−Rafter|, |Gbefore−Gafter|, |Bbefore−Bafter|) > 16 on 8-bit channels (0–255); otherwise black. A difference of exactly 16 is excluded. These are separate outputs, not a relabeling of Diff0.

| V1 CSS layout / appearance | Before | After | Diff0 amplified 8× | DiffT16 mask | DiffT16 changed pixels |
|---|---|---|---|---|---|
| 1440-light | ![Before V1 1440-light](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/46141e95fd51d508/V1-before-1440light.png) | ![After V1 1440-light](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/7f1461327f04529d/V1-after-1440light.png) | ![Diff0 amplified 8× V1 1440-light](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b7684e0a9ac284ee/diff-amplified8-1440-light.png) | ![diffT16-1440-light.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e28b9213290ac6e3/diffT16-1440-light.png) | 669 / 1,024,000 (0.0653%) |
| 1440-dark | ![Before V1 1440-dark](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/836d572898f22094/V1-before-1440-dark-0.png) | ![After V1 1440-dark](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/85a875667c769ff1/V1-after-1440-dark-0.png) | ![Diff0 amplified 8× V1 1440-dark](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/2f5bcb550b685ddf/diff-amplified8-1440-dark.png) | ![diffT16-1440-dark.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ee40d35bbae835ac/diffT16-1440-dark.png) | 665 / 1,024,000 (0.0649%) |
| 390-light | ![Before V1 390-light](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e0ad439388ac512e/V1-before-390-light-final-0.png) | ![After V1 390-light](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c8a28538dcfa2408/V1-head-390-light-verified-0.png) | ![Diff0 amplified 8× V1 390-light](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/d7f872ca88948cbe/diff-amplified8-390-light.png) | ![diffT16-390-light.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/5918233940df6d8d/diffT16-390-light.png) | 4,255 / 1,316,640 (0.3232%) |
| 390-dark | ![Before V1 390-dark](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a708d1c1d6d8ee74/V1-before-390-dark-final-0.png) | ![After V1 390-dark](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f2f5836251b74b64/V1-head-390-dark-verified-0.png) | ![Diff0 amplified 8× V1 390-dark](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/9ce9a6fd8fe957b6/diff-amplified8-390-dark.png) | ![diffT16-390-dark.png](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ed465bb5ddd5ff36/diffT16-390-dark.png) | 4,210 / 1,316,640 (0.3198%) |

Visual inspection: opened all four side-by-side inspection sheets containing Before, After, Diff0 amplified, and DiffT16. The wide masks isolate the Send-button region. Narrow Diff0 also shows the previously reported header sidebar-button hover/focus difference; DiffT16 filters that header difference and retains the Send-button region. Pixel counts describe image differences, not additional behavioral proof.

## Limitations and FRICTION

- Native iOS/Android V1/V2/V3/I1 remains unverified. EAS authentication/capability was healthy, but there was no current-source dev client/Metro; **lead did not schedule native build**. This does not claim the user waived native verification.
- Actual Electron timeout/recovery smoke remains missing.
- C1 full command execution/destination isolation/serial revocation partition remains unproven by this capture operator. Full negative UI partition also remains incomplete; browser-local synthetic HTTP 401 is not server credential revocation.
- Clock freezing remains unmet. Historical per-frame viewport telemetry remains absent.
- V3 narrow-light exact repeat stability remains unmet; no before aggregate comparison or native aggregate-reader verification was added.
- Original capture friction remains: stale wait/remount was unreliable; diagnostic registry refresh triggered real HTTP typed timeout. Watch reloads produced rejected 503/502/reconnect attempts; those are not timeout proof. Earlier rejected narrow/menu/fetch-error attempts remain retained and excluded from selected proof.
- This completion only generated local evidence artifacts outside the repository and uploaded through ghdrop. No source, browser, server, or forge state was changed; no commit or merge was made. No proof gap above was resolved by adding threshold masks.