# Evidence — Usage Limits period fix

Verified the changed Limits header and explanation with a genuine empty limits state. No new defect was found in the changed path. Populated live-limit cards remain unverified. An existing Tokens header overlap at 390px was reproduced on both commits.

Before: [bb7783e295](https://git.cafe/pingdotgg/t3code/commit/bb7783e295). After: [459dcae8](https://git.cafe/pingdotgg/t3code/commit/459dcae8b94d3df5641df217635e3ec4e57ecb6f). Only UsagePage.tsx differs. The test agent made no source edits, commits, pushes, PRs, comments or merges. Branch fix/issue-16179 is restored and clean.

## Environment and fixture

Linux fs-31, server-owned Chrome 154.0.0.0 preview, dedicated default-profile tab. Existing lead server: web localhost:7478, backend 127.0.0.1:15518, PID 2237403; worktree /home/ubuntu/wt/issue-16179, isolated T3 home .t3. Both processes were confirmed in this worktree. Lead session 33280 was neither restarted nor stopped.

CSS viewports 1440×900 and 390×844, light/dark through appearance emulation, zoom 100%, DPR 2, root font 16px, computed system sans-serif stack. The preview snapshot service caps wide PNGs at 1280px: wide outputs are 1280×800, phone outputs 780×1688. Responsive web was tested; native mobile is unchanged and excluded. Desktop wraps this web page; native titlebar/insets were not captured.

All environments (one connected environment, fs-31) and all providers selected. Genuine transcript/cache history supplied 471 historical sessions and 27 Past-24h sessions. No live subscription pool was available. No fake provider replies, DOM copy, synthetic data, account setup or database writes were used. Cursor history required CLI login. Historical totals changed as real transcripts were scanned; these estimates were not frozen.

Matched Limits captures use the same 30-day retained preference, 256px expanded wide sidebar, hidden phone sidebar, selection and theme. CSS disables motion/caret and hides only relative sidebar clocks on both commits. Fonts were ready; completed app content was awaited; a first scaled repaint was discarded. Each retained still has two pixel-identical RGBA frames separated by at least 500ms. Clock/data were not changed. The available tools did not expose reduced-motion media emulation; CSS suppression was used. [Frame stability proof](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c710b30696c63d64/frame-stability.json)

[Fixture provenance](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a6a51fc1c61945d6/fixture-provenance.md) · [Exact capture geometry and settling expression](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/d1f7655a90c44f15/matched-final-index.json) · [Capture recipe](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/26453b80035e81ec/capture-recipe.js) · [Diff script](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/5944db92c5084ba9/make-diffs.py)

## Matrix coverage

G = wide light/dark and phone light/dark. “Partial” is not a sign-off of the missing state.

| Row | Result | Evidence / limitation |
|---|---|---|
| 1 | Partial | Before/after in G prove both header families and paragraph placement with empty limits. Genuine populated live-limit pool unavailable; existing cards not verified. |
| 2 | Covered, existing layout caveat | Cost / 7 days has real nonzero history and an enabled period control in G; paragraph absent. See stills below. |
| 3 | Covered, existing layout caveat | Tokens / Past 24h has real nonzero hourly history in G; paragraph absent. At 390px provider breadcrumb overlaps Tokens; baseline header is pixel-identical. |
| 4 | Covered | G recordings show Cost→7 days→Limits→Tokens→Cost, all four options, 90 days, Limits reload, retained 90 days on return. Compact popups shown. |
| 5 | Covered | G keyboard recordings + DOM observations: Usage period count 0 on Limits; refresh follows metric in focus traversal. Day shortcut leaves retained 7 days unchanged. Palette search captures C without changing metric. Focus setup uses DOM focus on the real metric control before actual Tab keys. |
| 6 | Covered | G refresh recordings and attribute observations show genuine disabled→enabled transitions. Limits busy ~372–465ms; historical refresh ~121–390ms. No latency or provider values fabricated. Explanation/period remain stable. |
| 7 | Partial | 390 light deselect/restore recording + still prove paragraph guard. No-config environment branch source-confirmed, uncaptured; existing lead server was not reset. |
| 8 | Covered | 390 dark natural empty limits state: explanation above no-provider message. No provider filtering needed or account data fabricated. |
| 9 | Unverified | No genuine Limits notice, usage link or Cursor enable prompt was supplied by this environment. Existing child path unchanged. |
| 10 | Partial | Boundary + sidebar collapse recordings: one visible metric/refresh family; period 0 on Limits / 1 on history. At sidebar 256px the switch is viewport 1191→1192px (header container 935→936px, including 40px padding). Desktop native chrome not tested. |
| 11 | Covered for local web | Final entry recording shows sidebar, Ctrl+U and palette routes retain Limits preference. Remote/relay source-only: same renderer; no transport/schema/origin changes. No relay provisioned. |

## Before / After — Limits (rows 1 and 8)

These show the genuine empty subscription-limits state, not a populated-account substitute. Wide uses ToggleGroup; phone uses Select.

### 1440px light

| Before | After |
|---|---|
| ![Before: Limits, 1440px light](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ea7b1a47740121f8/before-1440-light-f1.png) | ![After: Limits, 1440px light](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a6658ff86b54ad09/after-1440-light-f1.png) |

[Threshold 0, RGB difference amplified 8×](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f6261cc4fdb8ca9e/diff-1440-light-threshold0.png) · [Threshold 16/255, magenta changed-pixel mask](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b5b1474650eb8830/diff-1440-light-threshold16.png)

### 1440px dark

| Before | After |
|---|---|
| ![Before: Limits, 1440px dark](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/1b4d1817d8a80873/before-1440-dark-f1.png) | ![After: Limits, 1440px dark](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/7bcae094653e667f/after-1440-dark-f1.png) |

[Threshold 0, RGB difference amplified 8×](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/631cca595234e2b5/diff-1440-dark-threshold0.png) · [Threshold 16/255, magenta changed-pixel mask](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/3751a84f43ed3ef1/diff-1440-dark-threshold16.png)

### 390px light

| Before | After |
|---|---|
| ![Before: Limits, 390px light](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/7222db9cada1d55b/before-390-light-f1.png) | ![After: Limits, 390px light](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/4e39bbf3aaf9907f/after-390-light-f1.png) |

[Threshold 0, RGB difference amplified 8×](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/1151e4ef6d0b0eb3/diff-390-light-threshold0.png) · [Threshold 16/255, magenta changed-pixel mask](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0d160fae3b311c40/diff-390-light-threshold16.png)

### 390px dark

| Before | After |
|---|---|
| ![Before: Limits, 390px dark](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/cd1b6598cdfcac12/before-390-dark-f1.png) | ![After: Limits, 390px dark](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/f8ac872160dae2ea/after-390-dark-f1.png) |

[Threshold 0, RGB difference amplified 8×](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/ad6fdd1e723b35e8/diff-390-dark-threshold0.png) · [Threshold 16/255, magenta changed-pixel mask](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/7aa3142089b55543/diff-390-dark-threshold16.png)

Diffs composite transparency consistently onto the theme background, then compare max absolute RGB channel difference. Threshold 0 includes every nonzero pixel; threshold 16 marks pixels with any channel difference >16/255. No regions are masked from these diffs. [Changed-pixel counts](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/990c2d5aa32f3c2e/diff-stats.json)

## Historical regression stills (rows 2–3)

| View | Cost / 7 days | Tokens / Past 24h |
|---|---|---|
| 1440px light | ![Cost 7 days, 1440px light](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/4b2ce26f9b137430/after-cost7-1440-light-f1.png) | ![Tokens Past 24h, 1440px light](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b65c3605020a3b38/after-tokens1-1440-light-f1.png) |
| 1440px dark | ![Cost 7 days, 1440px dark](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/d7f6aaeeb9101622/after-cost7-1440-dark-f1.png) | ![Tokens Past 24h, 1440px dark](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/da6dc4c9c504062a/after-tokens1-1440-dark-f1.png) |
| 390px light | ![Cost 7 days, 390px light](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/6d88f8198a4644fb/after-cost7-390-light-f1.png) | ![Tokens Past 24h, 390px light](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/18151dcce4d20e5e/after-tokens1-390-light-f1.png) |
| 390px dark | ![Cost 7 days, 390px dark](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/d78330f0bf69ce50/after-cost7-390-dark-f1.png) | ![Tokens Past 24h, 390px dark](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/3fcfb9aa26e454b2/after-tokens1-390-dark-f1.png) |

## Interaction recordings (rows 4–6)

| View | Metric / period / reload | Keyboard / real refresh |
|---|---|---|
| 1440px light | [Recording](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/5c52a9bb0f648b71/interactions-1440-light.mp4) | [Recording](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/60748cd7e345a343/focus-refresh-1440-light.mp4) |
| 1440px dark | [Recording](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c4631a5d8ef77a5f/interactions-1440-dark.mp4) | [Recording](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/9ce6c4567d2760c3/focus-refresh-1440-dark.mp4) |
| 390px light | [Recording](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a0205171e4dfd343/interactions-390-light.mp4) | [Recording](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/01b8916dc88ddecf/focus-refresh-390-light.mp4) |
| 390px dark | [Recording](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/29f242b8a4c6dbd8/interactions-390-dark.mp4) | [Recording](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/eb491ec9cc676f27/focus-refresh-390-dark.mp4) |

[Timestamped focus / refresh / boundary observations](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/79a4edd66f0bc611/runtime-observations.json) · [Timestamped main interaction observations](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/0a56e10e11108b23/raw-index.json)

[Deselect and restore, with settled endpoints](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/e9b9f97173970bc6/deselect-final.mp4) · [Deselected 390 light still](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a51c660c41807d7b/deselected-390-light.png) · [Responsive and sidebar recording](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/15b769892060070b/boundary.mp4) · [1191 / 1192 / 1193px boundary](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/8b1ab60b825af624/boundary-exact.mp4) · [Sidebar / keyboard / palette and shortcut suppression](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/c3967da0c345712c/entry-final.mp4)

Recordings were opened through 2fps sequential frame contact sheets and inspected against the timestamped observations. This is frame-sequence inspection, not a claim of full-speed audio/video playback. Stills and both diff images for every pair were opened and inspected. Raw recordings remain available above.

## Existing adjacent defect

At 390px, Tokens / Past 24h overlaps the provider breadcrumb. Both baseline and head have provider button x179.73–219.73 and metric selector x186.23–257.66 CSS px. The header crop has zero before/after changed pixels. The full before/after screenshots are also identical for this extra pair.

| Before | After |
|---|---|
| ![Before: existing Tokens breadcrumb overlap](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b2239a45ab07e019/adjacent-before-tokens390dark-f1.png) | ![After: same existing Tokens breadcrumb overlap](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/97c2a89a23393bfc/adjacent-after-tokens390dark-f1.png) |

[Extra pair threshold 0 / 8×](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/d2bb60eccce5ec0e/adjacent-diff-threshold0.png) · [Extra pair threshold 16](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/a9de8dfa5f6eb8dd/adjacent-diff-threshold16.png)

## Timeline / FRICTION

- 11:21–11:23 UTC: opened own tab; preview host disconnected once, then recovered. Default profile was not paired to this origin; supported isolated-home pairing CLI + UI recovered authentication. No tokens uploaded.
- 11:24–11:26: four G interaction recordings and historical stills; real history confirmed. 11:28–11:29: boundary recordings. 11:30–11:31: focus and disabled/recovered refresh observations. 11:33–11:34: final matched static captures and baseline Tokens comparison. 11:35: final entry and deselect/restore recordings with held endpoints.
- Initial light frames caught theme/icon/scaled repaint; the first snapshot is now discarded. A phone-width reload initialized the sidebar at 208px and initially polluted a wide comparison; final pairs reload at 1440 and independently record matching 256px sidebar geometry. Superseded material is labeled rejected in the archive.
- An exploratory Escape after a shortcut unmounted a popup left Usage; subsequent keystrokes entered an unsent local draft. That draft was discarded; no turn was sent. This sequence is rejected evidence, and successful boundary capture was repeated from a clean route. Do not infer a new popup regression from this run.
- Early clips stopped before a final screencast frame painted. Final entry/deselect recordings hold endpoints ≥1 second. Original clips are retained as exploratory material.
- Ongoing real transcript scanning changed historical amounts across configurations; static Limits comparisons use an unchanged empty state. No claims about invoice accuracy or data aggregation are made.
- No configured-empty environment, genuine live-limit pool, Limits notice/link/Cursor prompt was available; native desktop was not tested. Those gaps remain explicit; no credentials, keychain access or accounts were provisioned.
- The only durable browser changes are a fresh auth session and the selected Usage preferences (restored to Limits/30 days). Environment selection is restored. Appearance/viewport overrides belong to this dedicated tab. No source or fixture changes were committed. Lead server and worktree remain available.

[All retained captures, superseded material, logs, scripts and inspection sheets (ZIP)](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/51247bda091ddadf/issue-16179-all-artifacts.zip) · [Original matrix](https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/4d7fb94483d4c40f/matrix.md)

Verification by gpt-6.1-sol through the Codex harness.