[PsychedelicShayna]

Note

GPT-6.1 Sol on behalf of PsychedelicShayna

What

Keep closed monospace font pickers from classifying every installed family when Appearance opens. When a picker opens, classify fonts in cancellable, approximately 8ms batches and reuse completed measurements across pickers and remounts. Display loading status while classification runs. Saved font choices, setting writes and live previews are unchanged.

Why

Refs #126.

The existing closed-picker render filters the full installed-font catalog through a canvas probe that can perform 32 measurements per family. Already-granted font permission starts enumeration on mount, so opening Appearance enters that synchronous work without opening a picker. Remounting repeats it. The repair removes that work from closed-picker rendering and yields to input between opened-picker batches.

Testing

Installed nightly source 3ca07c37659338f9fde4142baf50d20f09e4c5d8, version 0.0.42-nightly.20261022.39742893669.1. The NeoCode launcher now selects this versioned AppRun; the previous artifact and launcher backup remain available. AppImage SHA256: bde7d6dce6a2928cedf8104ffcf912d8dafb75d59b65f79a052aa65b671bc113.

Proof used the actual installed Electron artifact, normal shell hydration and private state on Xvfb :10. The owner's live window, database and wallet were not used or restarted.

PR #131 is delivered. Same installed-font catalog in both measurements: 554 families, 2,817 faces.

Appearance navigation	Before, 3c5789f	Installed, 3ca07c3
Longest renderer task	7,807 ms	156 ms

This is one measured before/after case, not a general benchmark. The after-proof also typed into the opened monospace picker, closed it during classification, navigated away and back with no long tasks in that interaction window, completed classification, selected DejaVu Sans Mono, retained it across navigation, and restored the default Liberation Mono.

Three focused scheduler tests, web types, targeted lint and formatting passed. The initial bootstrap-root Vite runner failed before collecting tests; the corrected worktree-local runner passed.

Before image · After image · Interaction recording · Before profile · After profile · Installed receipt.

---

• vp check and vpr typecheck pass (CI runs both; locally run only the checks for the touched scope, per AGENTS.md › Verifying). Focused checks above; no full suite.
• Tested locally. Installed private-state proof and focused checks are recorded above.
• Before/after images (video for motion) uploaded to GitHub for UI changes, not committed (if user-facing). Linked above.

Model: openai-codex/gpt-6.1-sol. Harness: NeoPi.

Installed integration: GPT-6 Astra in NeoPi; focused execution by GPT-6 Luna.

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ The exact PR base did not have a successful artifact. Baseline uses the latest successful main measurement shown below.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.5 KiB	13.5 KiB	−26 B (−0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.1 KiB	7.1 KiB	+12 B (+0.2%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.4 KiB	−38 B (−0.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	56.3 KiB	56.2 KiB	−88 B (−0.2%)	66.4 KiB	✅
Codex	Live turn messages	10	8	−2 (−20.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	−17 B (−0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.1 KiB	7.1 KiB	0 B (0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.4 KiB	−17 B (−0.3%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.0 KiB	57.0 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	9	9	0 (0.0%)	21	✅

Baseline: 3c10e1c · PR result: c1fc2fc · Source CI: failure

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 113.9 KiB
• Claude decoded thread snapshot: 114.6 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.