Font picker in the command palette and heavier text weight

kayg commented

2026-10-08 07:20:44 +00:00

Owner

Owner request (verbatim): "maple mono should be used with a higher font weight? it looks too thin right now? allow for the whole font to be changed please with cmd + k -> let's add all the popular monospace fonts like we discussed before."

Work (branch font-picker, base dev 5e483ed, worktree .worktrees/font-picker), as separate commits:

Per-font weight registry; compare 400/450/500 for Maple Mono NF.
Font group in the command palette with live preview; nine OFL fonts (D87), lazy loaded, per-font transfer budget.
Exact cell metrics and cursor proof for every font at DPR 1 and 2; redraw p50 per font.
Terminal ligatures stay off for every font (D175).

Decisions: D87, D160, D171, D174, D175. Job started; findings and gate output follow in comments.

Owner request (verbatim): "maple mono should be used with a higher font weight? it looks too thin right now? allow for the whole font to be changed please with cmd + k -> let's add all the popular monospace fonts like we discussed before." Work (branch font-picker, base dev 5e483ed, worktree .worktrees/font-picker), as separate commits: 1. Per-font weight registry; compare 400/450/500 for Maple Mono NF. 2. Font group in the command palette with live preview; nine OFL fonts (D87), lazy loaded, per-font transfer budget. 3. Exact cell metrics and cursor proof for every font at DPR 1 and 2; redraw p50 per font. 4. Terminal ligatures stay off for every font (D175). Decisions: D87, D160, D171, D174, D175. Job started; findings and gate output follow in comments.

kayg commented

2026-10-08 07:20:57 +00:00

Author

Owner

Started. Branch font-picker, base SHA 5e483ed (dev), worktree .worktrees/font-picker. Host load at start: 2.26 3.48 8.02.

Started. Branch font-picker, base SHA 5e483ed (dev), worktree .worktrees/font-picker. Host load at start: 2.26 3.48 8.02.

kayg commented

2026-10-08 08:26:26 +00:00

Author

Owner

Finished. Branch font-picker, head 4c5c08a16f, base 5e483ed. Host load during final runs 2.3-3.0 on 8 cores.

Commits:

4c5c08a terminal: place each glyph on its cell for Monaspace Neon; the cursor proof sees line-end drift (#90, D87)
459f10a ui: let the browser slant Fira Code for italics (#90)
0ce502a test(tracer): cursor, ligature and redraw proofs for every font (#90, D87, D175)
f2b2a27 ui: Font group in the command palette with live preview (#90, D87)
3a3a03b fonts: ship the eight other curated families as latin woff2 subsets (#90, D87)
b42be2f ui: per-font text weights; Maple Mono NF as one variable face at body weight 500 (#90, D175)

Finding (Monaspace Neon drift): Chrome without subpixel glyph positioning rounds each glyph advance (0.62 em = 8.68 px) to 8 or 9, so a 56-glyph line ended at column 60.9 and coloured runs lost their last letter; fixed with per-glyph placement for that family. The cursor proof fails without the fix (line ends 24 px short at DPR 1, 47 px at DPR 2).

scripts/verify fast (tail, verbatim):

==> Tracer compiler, precompression and bundle transfer budget
bun install v1.4.2 (744846f84)

Checked 38 installs across 69 packages (no changes) [2.00ms]
$ bun x tsc --project tsconfig.json
bun test v1.4.2 (744846f84)

 20 pass
 0 fail
 376 expect() calls
Ran 20 tests across 6 files. [105.00ms]
$ bun run build.ts
built main.js, chunk-gkkj8c3x.js, chunk-21a2ehxw.js, backdrop.worker.js, faulty-terminal.worker.js, chunk-7bkjrwp6.js, chunk-ne2mxgwd.js, chunk-k8ea9tv3.js, chunk-axbkkqp1.js, chunk-bzvsykg5.js
faulty-terminal.worker.js: identity=1021 gzip=534 brotli=457 bytes
backdrop.worker.js: identity=241 gzip=209 brotli=158 bytes
main.js: identity=84023 gzip=31253 brotli=27230 bytes
JavaScript transfer: identity=849773 gzip=259399/307200 brotli=213792/256000 bytes
Font transfer: monaspace-neon=84340/88064 jetbrains-mono=70384/75776 source-code-pro=46484/49152 geist-mono=41364/45056 ibm-plex-mono=41344/45056 cascadia-code=85008/88064 maple-mono-nf=99876/106496 (critical 46432/51200) symbols=223748/235520 fira-code=36756/40960 commit-mono=71812/75776 bytes

==> Rust stream oracle (for the differential test)

==> TypeScript protocol tests
bun test v1.4.2 (744846f84)

src/stream.test.ts:
stream differential: 134 streams, 11439 schedules, zero divergences

 7 pass
 0 fail
 34390 expect() calls
Ran 7 tests across 2 files. [2.32s]
EXIT 0

Note: renderer-regression 'predictive echo keeps 50 MB output throughput within five percent' failed at -6.0% (budget -5) in my run at 300 ms RTT; it compares prediction on and off with the same font and I did not run it on the base, so treat it as unconfirmed noise.

Finished. Branch font-picker, head 4c5c08a16f82c9cc668d7a02b4c3c6669901b752, base 5e483ed. Host load during final runs 2.3-3.0 on 8 cores. Commits: - 4c5c08a terminal: place each glyph on its cell for Monaspace Neon; the cursor proof sees line-end drift (#90, D87) - 459f10a ui: let the browser slant Fira Code for italics (#90) - 0ce502a test(tracer): cursor, ligature and redraw proofs for every font (#90, D87, D175) - f2b2a27 ui: Font group in the command palette with live preview (#90, D87) - 3a3a03b fonts: ship the eight other curated families as latin woff2 subsets (#90, D87) - b42be2f ui: per-font text weights; Maple Mono NF as one variable face at body weight 500 (#90, D175) Finding (Monaspace Neon drift): Chrome without subpixel glyph positioning rounds each glyph advance (0.62 em = 8.68 px) to 8 or 9, so a 56-glyph line ended at column 60.9 and coloured runs lost their last letter; fixed with per-glyph placement for that family. The cursor proof fails without the fix (line ends 24 px short at DPR 1, 47 px at DPR 2). scripts/verify fast (tail, verbatim): ``` ==> Tracer compiler, precompression and bundle transfer budget bun install v1.4.2 (744846f84) Checked 38 installs across 69 packages (no changes) [2.00ms] $ bun x tsc --project tsconfig.json bun test v1.4.2 (744846f84) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [105.00ms] $ bun run build.ts built main.js, chunk-gkkj8c3x.js, chunk-21a2ehxw.js, backdrop.worker.js, faulty-terminal.worker.js, chunk-7bkjrwp6.js, chunk-ne2mxgwd.js, chunk-k8ea9tv3.js, chunk-axbkkqp1.js, chunk-bzvsykg5.js faulty-terminal.worker.js: identity=1021 gzip=534 brotli=457 bytes backdrop.worker.js: identity=241 gzip=209 brotli=158 bytes main.js: identity=84023 gzip=31253 brotli=27230 bytes JavaScript transfer: identity=849773 gzip=259399/307200 brotli=213792/256000 bytes Font transfer: monaspace-neon=84340/88064 jetbrains-mono=70384/75776 source-code-pro=46484/49152 geist-mono=41364/45056 ibm-plex-mono=41344/45056 cascadia-code=85008/88064 maple-mono-nf=99876/106496 (critical 46432/51200) symbols=223748/235520 fira-code=36756/40960 commit-mono=71812/75776 bytes ==> Rust stream oracle (for the differential test) ==> TypeScript protocol tests bun test v1.4.2 (744846f84) src/stream.test.ts: stream differential: 134 streams, 11439 schedules, zero divergences 7 pass 0 fail 34390 expect() calls Ran 7 tests across 2 files. [2.32s] EXIT 0 ``` Note: renderer-regression 'predictive echo keeps 50 MB output throughput within five percent' failed at -6.0% (budget -5) in my run at 300 ms RTT; it compares prediction on and off with the same font and I did not run it on the base, so treat it as unconfirmed noise.