cmux NEXT model picker: compact like Claude Code's menu (as wide as its rows), one tab per harness pinned by teamleaderleo · Pull Request #30194 · manaflow-ai/cmux

teamleaderleo · 2026-11-01T05:18:04Z

Leo, 2026-11-01: "i also don't like the way this composer has so much space between stuff surely we can get more compact horizontally" / "see claude code has that" / "we have too many harnesses because you didn't put those things into accounts or whatever"

Spec: design/model-picker.md (checklist 1 and new 21, teamleaderleo/leo-workspace PR on lane/design-model-picker-compact).

What changes

Compact width. The symbols-row menu was a fixed 480px (.acpmux-quick-railed, from fix(next): size model picker to available content #26615), and its width probe measured every model of every harness, Legacy lists included. So the keys and the effort sat far from short names like "Opus 5.5".
- Now the menu is as wide as the open tab's rows (fit), as Claude Code's menu is. The probe leaves out Legacy models and adds the star and ⌘N columns.
- The effort column is 88px (a level's name and chevron) instead of 112px.
- While open, the menu keeps its widest width (a ResizeObserver on the popup, made and disconnected with the rail's ref), so a tab switch never narrows it or moves the rail (Fix the combined model picker's harness list and sizing #25796's test still passes).
One tab per harness. Base already folds routes since cmux NEXT model picker: one tab per model family; transports and logins are Account routes #29770 (00:03Z, before Leo's 05:05Z message): every entry that draws the Claude mark (Claude Code, claude-sr2, a work login, the SR route) is one Claude tab, and Codex with its app server and SR route is one OpenAI tab. Each route is a radio in that tab's Account submenu, above the logins, so every route stays reachable. Leo's screenshot (nine tabs; Account showing only "CLI login") matches a build from before cmux NEXT model picker: one tab per model family; transports and logins are Account routes #29770.
- This PR adds a gallery fixture shaped like his rail and a test that pins the fold: routes-one-tab, routes-account, and "one tab per harness: Claude and Codex routes are Account choices, never tabs".

Gallery (`new-tab.quick-model-menu`): 4 versions, `fit` picked

version	width in the fixture
`width-fit` (picked; closest to Claude Code's)	323px
`width-stable` (every tab's rows)	391px
`width-snug` (tighter columns, 360px cap)	360px
`width-wide` (before)	480px

Before (width-wide, dark) and after (width-fit, dark and light):

Leo's rail, eight entries folded into four tabs, and its Account submenu with the routes:

Viewed after image: yes. Visual verdict: SHIP. The fit menu reads like Claude Code's: effort, star and ⌘N sit right after the names, with no empty band, and one mark per harness.

Tests

quick-model-menu-layout: 22 pass, including the two new ones and "harness controls keep their position while each direction follows visible content". model-menu-compact, QuickModelMenu and model-menu-started-chat-fork: 105 pass. vp fmt and vp lint are clean.

Risk

Small: the compact menu's CSS width and its width probe, plus one ResizeObserver whose cleanup is the rail's ref callback (it disconnects when the rail unmounts or the menu closes).

Agent session: agent session

🤖 Generated with Claude Code

… close to the name; gallery widths and Leo's routes rail Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

github-actions · 2026-11-01T06:06:14Z

cmux status _{(one comment, edited in place; each line is updated by the job that owns it)}

Review dca8c99: HQ feed card · gallery before/after runs · checks

App of this head (app.zip, kept 1 day)

gh workflow run reload-build.yml -R manaflow-ai/cmux --ref feat-cmux-next -f tag=pr30194 -f ref=lane/nx-picker-compact-one-tab-per-harness -f platform=macos -f backend_mode=local

Gallery before/after for dca8c99: 16 states rendered, 16 images differ, 1535 more touched states not rendered. Images are in the run's artifact, not on this public PR; the independent review of this head reads them on its own, so the PR needs no hand-made screenshots of these states.

gh run download 41318157754 -R manaflow-ai/cmux -n gallery-advisory-30194-dca8c999acab2e702ed5e7644750507ec4450c98 (then read gallery-review.json)

States

State	Theme	Status	Diff	Picked by
`agent-pane.composer-pickers/closed`	Apple System Colors Light	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/closed`	Apple System Colors	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/pill-rail`	Apple System Colors Light	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/pill-rail`	Apple System Colors	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/model-menu`	Apple System Colors Light	changed	1.01%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/model-menu`	Apple System Colors	changed	0.989%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/model-more`	Apple System Colors Light	changed	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/model-more`	Apple System Colors	changed	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/model-focus`	Apple System Colors Light	changed	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/model-focus`	Apple System Colors	changed	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/model-press-drag-release`	Apple System Colors Light	changed	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/model-press-drag-release`	Apple System Colors	changed	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/effort-menu`	Apple System Colors Light	changed	3.62%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/effort-menu`	Apple System Colors	changed	3.62%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/effort-focus`	Apple System Colors Light	changed	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-pickers/effort-focus`	Apple System Colors	changed	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/claude-auto`	Apple System Colors Light	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/claude-auto`	Apple System Colors	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/grid-running`	Apple System Colors Light	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/grid-running`	Apple System Colors	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/grid-draft`	Apple System Colors Light	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/grid-draft`	Apple System Colors	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/long-model`	Apple System Colors Light	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/long-model`	Apple System Colors	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/leo-report`	Apple System Colors Light	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/leo-report`	Apple System Colors	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/granted`	Apple System Colors Light	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/granted`	Apple System Colors	unchanged	0%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/fast-off`	Apple System Colors Light	changed	4.12%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/fast-off`	Apple System Colors	changed	3.99%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/fast-on`	Apple System Colors Light	changed	4.12%	own `QuickModelMenu.tsx`
`agent-pane.composer-row/fast-on`	Apple System Colors	changed	3.99%	own `QuickModelMenu.tsx`

_{Pick other states with a Gallery: entry.id, entry.id/variant line in the PR body.}

Unclear: web / react-apps-check failed: no known signature; failed step: Check webview test shards.

CI failed on dca8c999ac (run 41318158195 attempt 1): 1 code, 1 unknown.

Job	Verdict	Why
web / react-apps-base-check (6)	code	a web test failed
web / react-apps-check	unknown	no known signature; failed step: `Check webview test shards`

Failures:

seen on other PRs new-tab-typing.test.ts shell typing and the normal prompt keep Markdown punctuation literal: failed (job)

Matched log lines

web / react-apps-base-check (6): (fail) shell typing and the normal prompt keep Markdown punctuation literal [592.53ms]

Not re-run automatically: web / react-apps-base-check (6), web / react-apps-check are not machine failures.

Written by scripts/ci/classify_failures.py (ci-failure-attribution.yml); signatures are its SIGNATURES table. A machine verdict is the runner's fault, not this PR's; yours means the failing file is one this PR changes, also red on main that main's latest full suite fails the same way, seen on other PRs that it failed on another pull request's run lately.

red: cmux-next checks (god files, concurrency, crash safety, l10n); no finished base run on cefc64e to compare; do not merge until green; see lane-rules

Shipped in:

cmux NEXT nightly 1.0.0-nightly.4132201832701 (621470180f)
cmux NEXT nightly 1.0.0-nightly.4132260480101 (f3ae381789)
cmux NEXT nightly 1.0.0-nightly.4132310303501 (2ed0f1c3a4)
cmux NEXT nightly 1.0.0-nightly.4132418488301 (5f712a27cc)
cmux NEXT nightly 1.0.0-nightly.4132492786501 (3c0c02a5ae)
cmux NEXT nightly 1.0.0-nightly.4132569907101 (641e45405d)

blacksmith-sh · 2026-11-01T07:21:39Z

Found 9 test failures on Blacksmith runners:

Failures

Test	View Logs
`/tmp/tmprkj7udf9/summary-missing.json (live):/summary-missing.json (live):`	View Logs
`/tmp/tmpz0nyirtb/summary-fail.json (live):/summary-fail.json (live):`	View Logs
`[new_workspaces] debug.new_workspace timed 0 of 50 new workspace rows/` `[new_workspaces] debug.new_workspace timed 0 of 50 new workspace rows`	View Logs
`[new_workspaces] the window draws no frames/[new_workspaces] the window draws no frames`	View Logs
`[tabs] debug.tab_switch timed 0 of 40 switches/` `[tabs] debug.tab_switch timed 0 of 40 switches`	View Logs
`[tabs] the window draws no frames/[tabs] the window draws no frames`	View Logs
`test/new-tab-typing.test.ts/new-tab-typing.test`	View Logs
`test/new-tab-typing.test.ts/new-tab-typing.test`	View Logs
`test_the_repository_highlights_validate (main.DigestTests.test_the_repository_highl` `ights_validate)/` `test_the_repository_highlights_validate (main.DigestTests.test_the_repository_highl` `ights_validate)`	View Logs

^{Need help on this PR? Tag @codesmith-bot with what you need.}

… stable to every tab Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

…dth (#25796) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

…er narrows it Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

teamleaderleo · 2026-11-01T14:56:19Z

Review (subagent claude-opus-5-5 medium): SHIP at dca8c99

The full review at 8169d24 was a SHIP. The ResizeObserver lifecycle is correct: it disconnects on unmount, the next open gets a fresh node, and it cannot loop. wide keeps the old behaviour exactly, and specificity is correct. The delta dca8c99 caps the held min-width at min(Wpx, var(--available-width, ...)), which is valid inline, inherits from Base UI's positioner, and keeps the observer's early return. At this head, quick-model-menu-layout passes 22 tests and the model-menu unit suites pass 105; fmt and lint are clean.

Follow-ups:

no browser test narrows the pane while the menu is open (the held-width cap)
the probe assumes a 16px key slot; at a larger interface scale the menu may come out slightly narrower than its rows (cosmetic)

…eed run Base 762fda7 (after #30652 and #30194) adds center-line violations to the folded-run, queued-from-agent, hover-fade and incoming-message states; the base replay cannot excuse them until base has the checker, so they join the baseline rows from gallery-health run 41322226213. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a

…#31124) In the compact fit menu (#30194) a short searched row's middle falls on its effort item, which sits over the row and takes the click, so react-apps-check shard (7) timed out clicking 'Sonnet 5.5'. Agent-Session: https://cmux.com/deeplink/agent?id=aid-25f47763ee5347118457 Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

…30256) * gallery health: gate layout shift on load and every play step (element boxes per frame) WIP: baseline seeded in a following commit. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery: shift scenarios play Codex folds and the debug load_seed history; document the shift gate Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery health: check the design specs' geometry rules from the DOM of every state Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery-matrix: write shift rows from the worst seeding run A state that shifts in some runs got a 0 row when rows were written from the least run, and would then fail unrelated PRs on its flaky runs. Review follow-ups 2 and 3 on #30250. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery: older-history waits on a boolean (typecheck) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery-matrix: seed shift-baseline.json from two gallery-health runs Rows from the worst of #30250's two full gallery-health runs (0ac1691 and a17daa5): 2763 states, 561 with a non-zero row. A state at 0 now fails on any shift; the rest fail past their row by a quarter. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery: a wait step never fails a play; gallery health scores its shift The webview suite found that measuring waitFor as a step made existing plays fail on what they wait for: the header PR button arriving (header-tools), the mention list opening (composer-mention), the composer controls settling (narrow-layouts). A wait now goes unjudged in judgeStep, for shift, anchor moves, long frames and settle. Its shift is still measured and scored against the state's row in gallery health's shift baseline. Test: bun test test/gallery-shift.test.ts. The new case fails on the previous head and passes here. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery: say a wait step still gets the popup-layer check Review follow-up on #30250. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery: smooth zero-row shift limit, copy-proof violation keys, listboxes only in popups Three review follow-ups on #30256, fixed before the geometry baseline is seeded so its keys do not bake them in: - Every shift row's limit now adds LIMIT_SLACK (0.0005), a 0 row's included. A 0 row was ten times stricter than a tiny row, at the level where frame timing alone scores. - Violation keys use the row id, data-id, data-value, value or aria-label before an element's text, so a copy-only PR does not rename a listed violation. Plain generated ids are never used. - A listbox counts as a menu only inside a popover or dialog, so pane-wide list panels do not trip menu-fit. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery-matrix: reseed shift rows from three runs, seed geometry rows Shift rows are rebuilt from the worst of three full gallery-health runs: the two on #30250 and this PR's run at 92c5bb5 on base 593b25f. That gives 2843 states, 608 of them non-zero. Base moved after the first seed, and one flaky state, model-picker#quick-search-field-opencode (0.0024 to 0.0081 across renders), came out as this PR's own red. Geometry rows come from that run: 2841 states, 345 with listed violations. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * Gallery geometry baseline: add the violations base gained since the seed run Base 762fda7 (after #30652 and #30194) adds center-line violations to the folded-run, queued-from-agent, hover-fade and incoming-message states; the base replay cannot excuse them until base has the checker, so they join the baseline rows from gallery-health run 41322226213. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * Shift baseline: motion.sync-list#inserts-from-second-client row 0.0059 The insert from a second client animates the list rows into place; three seed runs scored 0, gallery-health run 41327725064 scored 0.0059 on the Replay click. A real insert moves rows, so its row is that run's score. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * Gallery baselines: seed shift rows from the worst of five full runs Full renders (this PR changes scripts/gallery-matrix) kept turning up states that shift in some runs and in every attempt of that run (guest-page.waiting#decided, artifact-panel#svg, settings-menus#customized), so three seed runs were not enough. Shift rows are rebuilt from scratch as the worst score of the three seed runs plus gallery-health runs 41322226213 and 41335560293; geometry rows gain the 3 keys the last run saw. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

…ays) (#30265) * gallery health: gate layout shift on load and every play step (element boxes per frame) WIP: baseline seeded in a following commit. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery: shift scenarios play Codex folds and the debug load_seed history; document the shift gate Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery health: check the design specs' geometry rules from the DOM of every state Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery: real-shaped recorded Claude and Codex chats (anonymized replays) Leo, 2026-11-01: "how about like the apparatuses for improving ui verification? let's examine what we can do better for our workflows". anonymize-recording.ts turns an acpmux session recording into a fixture with none of its words (same lengths, line breaks and punctuation; fake ids, fixed epoch). agent-pane.recorded-chats replays the #30176 Claude turn and a real Codex turn steered by agent inbox messages through the pane's own client, so gallery health's shift and geometry checks see real-shaped chats. Tests: bun test test/gallery-anonymize-recording.test.ts (2 pass). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery-matrix: write shift rows from the worst seeding run A state that shifts in some runs got a 0 row when rows were written from the least run, and would then fail unrelated PRs on its flaky runs. Review follow-ups 2 and 3 on #30250. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery: older-history waits on a boolean (typecheck) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery-matrix: seed shift-baseline.json from two gallery-health runs Rows from the worst of #30250's two full gallery-health runs (0ac1691 and a17daa5): 2763 states, 561 with a non-zero row. A state at 0 now fails on any shift; the rest fail past their row by a quarter. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery: a wait step never fails a play; gallery health scores its shift The webview suite found that measuring waitFor as a step made existing plays fail on what they wait for: the header PR button arriving (header-tools), the mention list opening (composer-mention), the composer controls settling (narrow-layouts). A wait now goes unjudged in judgeStep, for shift, anchor moves, long frames and settle. Its shift is still measured and scored against the state's row in gallery health's shift baseline. Test: bun test test/gallery-shift.test.ts. The new case fails on the previous head and passes here. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery: say a wait step still gets the popup-layer check Review follow-up on #30250. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery: smooth zero-row shift limit, copy-proof violation keys, listboxes only in popups Three review follow-ups on #30256, fixed before the geometry baseline is seeded so its keys do not bake them in: - Every shift row's limit now adds LIMIT_SLACK (0.0005), a 0 row's included. A 0 row was ten times stricter than a tiny row, at the level where frame timing alone scores. - Violation keys use the row id, data-id, data-value, value or aria-label before an element's text, so a copy-only PR does not rename a listed violation. Plain generated ids are never used. - A listbox counts as a menu only inside a popover or dialog, so pane-wide list panels do not trip menu-fit. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * gallery-matrix: reseed shift rows from three runs, seed geometry rows Shift rows are rebuilt from the worst of three full gallery-health runs: the two on #30250 and this PR's run at 92c5bb5 on base 593b25f. That gives 2843 states, 608 of them non-zero. Base moved after the first seed, and one flaky state, model-picker#quick-search-field-opencode (0.0024 to 0.0081 across renders), came out as this PR's own red. Geometry rows come from that run: 2841 states, 345 with listed violations. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * Gallery geometry baseline: add the violations base gained since the seed run Base 762fda7 (after #30652 and #30194) adds center-line violations to the folded-run, queued-from-agent, hover-fade and incoming-message states; the base replay cannot excuse them until base has the checker, so they join the baseline rows from gallery-health run 41322226213. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * Shift baseline: motion.sync-list#inserts-from-second-client row 0.0059 The insert from a second client animates the list rows into place; three seed runs scored 0, gallery-health run 41327725064 scored 0.0059 on the Replay click. A real insert moves rows, so its row is that run's score. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * Gallery baselines: seed shift rows from the worst of five full runs Full renders (this PR changes scripts/gallery-matrix) kept turning up states that shift in some runs and in every attempt of that run (guest-page.waiting#decided, artifact-panel#svg, settings-menus#customized), so three seed runs were not enough. Shift rows are rebuilt from scratch as the worst score of the three seed runs plus gallery-health runs 41322226213 and 41335560293; geometry rows gain the 3 keys the last run saw. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a * anonymize-recording: keep each event's kind in its typed output The scrubbed rest is typed through Omit of an index-signature type, which drops the required kind, so the merge with base failed the webviews type check (TS2741). The kind is a protocol word the scrubber keeps; set it from the event. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-e1bbe8a486fbacd68f4a --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

cmux NEXT model picker: as wide as the rows it shows, effort and keys…

b764a98

… close to the name; gallery widths and Leo's routes rail Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

teamleaderleo temporarily deployed to ci-routing-agent November 1, 2026 05:18 — with GitHub Actions Inactive

github-actions Bot added the risk:normal Risk tier (docs/ci/risk-tiers.md) label Nov 1, 2026

teamleaderleo temporarily deployed to ci-routing-agent November 1, 2026 05:21 — with GitHub Actions Inactive

Compact looks drop the rail's fixed 480px; fit sizes to the open tab,…

803fdb5

… stable to every tab Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

teamleaderleo temporarily deployed to ci-routing-agent November 1, 2026 13:31 — with GitHub Actions Inactive

Width test: compare the compact band with the fixed menu's

8747acd

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

teamleaderleo temporarily deployed to ci-routing-agent November 1, 2026 13:43 — with GitHub Actions Inactive

Default to the stable compact width: a tab switch keeps the menu's wi…

d351ad2

…dth (#25796) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

teamleaderleo temporarily deployed to ci-routing-agent November 1, 2026 13:50 — with GitHub Actions Inactive

teamleaderleo temporarily deployed to artifacts November 1, 2026 13:56 — with GitHub Actions Inactive

teamleaderleo temporarily deployed to sentry-dsyms November 1, 2026 13:56 — with GitHub Actions Inactive

teamleaderleo and others added 2 commits November 1, 2026 13:56

A compact menu keeps its widest width while open, so a tab switch nev…

1009be9

…er narrows it Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

Hold the compact width with a ResizeObserver on the rail's popup

b5fa1a2

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

teamleaderleo temporarily deployed to ci-routing-agent November 1, 2026 14:08 — with GitHub Actions Inactive

vp fmt

5cb52bf

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

teamleaderleo temporarily deployed to ci-routing-agent November 1, 2026 14:21 — with GitHub Actions Inactive

Default to fit: as wide as the open tab's rows, held while open

8169d24

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

teamleaderleo temporarily deployed to ci-routing-agent November 1, 2026 14:41 — with GitHub Actions Inactive

teamleaderleo changed the title ~~cmux NEXT model picker: compact width (effort and ⌘N close to the name), one tab per harness~~ cmux NEXT model picker: compact like Claude Code's menu (as wide as its rows), one tab per harness pinned Nov 1, 2026

teamleaderleo marked this pull request as ready for review November 1, 2026 14:49

The held width never passes the pane's available width

dca8c99

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-3cf6bed67d0bb92a21bd

teamleaderleo temporarily deployed to ci-routing-agent November 1, 2026 14:50 — with GitHub Actions Inactive

teamleaderleo merged commit eb08adc into feat-cmux-next Nov 1, 2026
164 of 174 checks passed

teamleaderleo deleted the lane/nx-picker-compact-one-tab-per-harness branch November 1, 2026 15:24

This was referenced Nov 1, 2026

cmux NEXT model picker: no browser test that the held menu width stays inside a narrowed pane #31062

Closed

cmux NEXT model picker: width probe assumes a 16px ⌘N slot at every interface scale #31063

Closed

Navigation Menu

Repository navigation

cmux NEXT model picker: compact like Claude Code's menu (as wide as its rows), one tab per harness pinned - #30194

cmux NEXT model picker: compact like Claude Code's menu (as wide as its rows), one tab per harness pinned#30194
teamleaderleo merged 9 commits into
feat-cmux-nextfrom
lane/nx-picker-compact-one-tab-per-harness

teamleaderleo commented Nov 1, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Nov 1, 2026 •

edited

Loading

Uh oh!

blacksmith-sh Bot commented Nov 1, 2026 •

edited

Loading

Uh oh!

teamleaderleo commented Nov 1, 2026

Uh oh!

Uh oh!

This branch was previously deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

teamleaderleo commented Nov 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What changes

Gallery (new-tab.quick-model-menu): 4 versions, fit picked

Tests

Risk

Uh oh!

github-actions Bot commented Nov 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

blacksmith-sh Bot commented Nov 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Failures

Uh oh!

teamleaderleo commented Nov 1, 2026

Uh oh!

Uh oh!

This branch was previously deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

teamleaderleo commented Nov 1, 2026 •

edited

Loading

Gallery (`new-tab.quick-model-menu`): 4 versions, `fit` picked

github-actions Bot commented Nov 1, 2026 •

edited

Loading

blacksmith-sh Bot commented Nov 1, 2026 •

edited

Loading