[teamleaderleo]

HOLD: with its gates fixed (098aec3) the spare is adopted on every open, and the adopted page regresses: sidebar opens show an empty transcript after the cover leaves; reopens hold the cover ~15 s and never paint. Not to land until the adopted page draws its chat and a same-host bench shows it faster.

Leo, 2026-11-01: "the loading for everything is just slow end to end / loading into a chat and some of its recent messages, loading the various ui elements of the metadata, etc. it's not cached well enough nor prerendered/prefetched well enough". Earlier, 2026-10-29: "as in i want ours to load fast too :(".

Finding 1 of the chat-open plan (manaflow-ai/cmuxterm-hq#5914): every tab not yet shown opens a cold page.

This PR carries #27759 (its 10 commits, unchanged) and fixes the blocking finding of its last review. #27759's lane has been idle since 04:32Z. #27759 stays open for its owner to close.

What changes

From #27759. Opening an existing chat of this Mac uses the prewarmed agent page (NewTabSparePool) instead of building a new one. That covers a sidebar click, Open in Tab, chat.openTab and a subagent's pane. The page is already loaded, mounted and connected, so it switches to the chat in place (chat.select), drawing the host's kept tail when there is one, and then attaches.

• The pane keeps its cover, with the page laid out outside the window, until the page reports the chat drawn.
• A chat from another machine, a Chief home chat, a seeded tab and a tab with a kept or relaunch wake state each still get their own page.

See #27759's body for the full design and its three earlier review rounds.

New here (the BLOCK at cffa960, P1). Before, a spare counted as ready once it had a handshake and a drawn New Tab frame. Both come before its acpmux client connects. An adoption that arrived then queued until connect() finished, and that includes the spare's whole first chat list. A cold page attaches its chat before the list (askEarly / attachedBeforeList). So under load, adopting could be slower than the cold path it replaces.

• The page now reports its client state (pane.clientReady {ready}): true once connect() finishes and its actions answer, false when the connection is lost. Each new handshake resets it (a reload or a reconnect).
• NewTabSparePool.servesChat requires it. A spare that is still connecting stays in the pool, and the tab opens cold as before.
• In Low Power Mode or at serious thermal state, no chat takes the spare. Each take would mean building another page. The tab opens cold, and the spare stays for New Tab.
• When an existing chat takes the spare, its replacement is built once input has been quiet for 750 ms (Refill.whenQuiet), not on the next main-actor turn. A new page's load on the main thread would otherwise land inside that chat's open and the next sidebar clicks. New Tab and new workspace still refill at once (Refill.now).

Spare cost

• Count. Still one spare per app (NewTabSpareSlot, a single slot). This PR adds no page.
• Memory. Unchanged: one parked page, which is one Chromium renderer.
• Idle CPU. Unchanged. perf-lead row F3 records ~2% idle CPU for Home's three renderers together. The spare is now taken more often, so pages are built more often. That build is the existing pool.makeSpare, and it happens only after quiet input on the chat path.
• Snapshot. The parked page's per-resize snapshot (gpu-audit Add bump-version script and improve reload script #7) is already a fast-mode JPEG on Chromium (PageWebView.snapshotParams). This PR changes neither when it is taken nor how.
• Pressure. A memory pressure warning or critical level still drops the spare. Low Power Mode and serious thermal state keep chats off it (above).

Performance

Measured with scripts/measure/chat-open.sh (chat-open-bench.py, the app's own debug.chat_open_trace from the click to the first frame of chat content), on a fleet display Mac. Base and head run in the same harness.

metric	before	after	delta	harness or run link
(pending)

Timers and observers

• No new timers or observers.
• The deferred refill uses the pool's existing warmTimer and idle input monitor, which stop once the spare is made (warmNow removes the monitor).
• The hold on an adopted page ends on its paint report, or on the loading state's existing give-up check.

Tests

• SpareSessionAdoptionTests:
  • thePageReportsItsClientReady: the request decodes; the state resets on a handshake; a ready report does not count as the person touching the page.
  • whichSpareServesAChat: the readiness, Low Power Mode and thermal decision.
  • The connected-spare helper now reports ready.
• cmux NEXT: opening an existing chat adopts the prewarmed, connected page #27759's Swift and web tests are carried unchanged.

Risk

• Medium. This changes the open path for every existing local chat (see cmux NEXT: opening an existing chat adopts the prewarmed, connected page #27759).
• The cost of a wrong readiness answer is bounded. A spare that is not ready means a cold open, which is today's behavior.

Agent session: agent session

🤖 Generated with Claude Code

---

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

---

Summary by cubic

Opening an existing chat now adopts the app's prewarmed, connected New Tab page instead of building a cold one. The page switches to the chat in place, keeping the reader's position and any subagent turn. Carries #27759's adoption path (sidebar click, Open in Tab, chat.openTab, subagent pane).

• The page reports its client state (pane.clientReady); a spare still connecting stays in the pool and the tab opens cold.
• The composer's audio report and the header's, project notice's and New Tab project list's mount reads are boot traffic now, so the spare is no longer refused as touched and stays adoptable.
• No chat takes the spare in Low Power Mode or at serious thermal state, since each take means building another page.
• A spare taken by a chat is replaced only after input is quiet (750 ms); New Tab and new workspace still refill immediately.
• A chat the daemon lacks is shown as missing, never replaced by another chat.
• Chats from another machine, Chief home chats, seeded tabs, and tabs with kept or relaunch wake states keep their own page.
• The chat-open trace records whether the spare was taken (spare_adopted) or why not (spare_empty, spare_unready, spare_refused).

^{Written for commit 098aec3. Summary will update on new commits.}

[cursor]

Bugbot needs on-demand usage enabled

Bugbot uses usage-based billing for this team and requires on-demand usage to be enabled.

A team admin can enable on-demand usage in the Cursor dashboard.

[github-actions]

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

Review 098aec3: 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=pr31033 -f ref=lane/ux-warm-agent-page -f platform=macos -f backend_mode=local

Gallery before/after for 098aec3: 16 states rendered, 0 images differ, 2899 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 41354707322 -R manaflow-ai/cmux -n gallery-advisory-31033-098aec3a4019b5e9c104572a5b496355b3feac77 (then read gallery-review.json)

States

State	Theme	Status	Diff	Picked by
agent-pane.chat-width/standard	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.chat-width/standard	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.chat-width/wide	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.chat-width/wide	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.chat-width/wide-blocks	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.chat-width/wide-blocks	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.composer-edge/short-edge	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.composer-edge/short-edge	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.composer-edge/wide-edge	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.composer-edge/wide-edge	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.composer-edge/no-edge	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.composer-edge/no-edge	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.composer-floats/scrolled-to-end	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.composer-floats/scrolled-to-end	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.composer-floats/composer-grown	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.composer-floats/composer-grown	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.composer-rest-gap/gap-40	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.composer-rest-gap/gap-40	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.composer-rest-gap/gap-56	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.composer-rest-gap/gap-56	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.composer-rest-gap/gap-72	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.composer-rest-gap/gap-72	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.continuation-spacing/shipped	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.continuation-spacing/shipped	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.continuation-spacing/two-thirds	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.continuation-spacing/two-thirds	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.continuation-spacing/half-tucked	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.continuation-spacing/half-tucked	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.continuation-spacing/before	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.continuation-spacing/before	Apple System Colors	unchanged	0%	own App.tsx
agent-pane.copy-controls/output	Apple System Colors Light	unchanged	0%	own App.tsx
agent-pane.copy-controls/output	Apple System Colors	unchanged	0%	own App.tsx

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

Passes: CI passes on 098aec3a40.

CI passes on 098aec3a40 (run 41354707389 attempt 1).

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); NEW: red check: Web bundle key inputs is not red on the base 93037ef; do not merge until green; see lane-rules

Default changes

This pull request changes 1 default at 098aec3a4019 (what everyone gets without configuring anything):

• tunable-source.CmuxNextApp/Control/DebugChatOpenTrace.swift:json: "299ee38ef9feaa638b8ac37661619814dfa3ba4031363d516e84840cfa1bc7c5" -> "efa437604b054a487f4dd740989919b4712aa37f6f84e0cb039f7047bfce81f7"

Signed off on this commit by review (independent Codex gpt-6.1-sol, high, via the review router; full diff from merge base 93037ef1de to this head).

Merge receipt: merged as 4453d0b036bb with these defaults changed, signed off by review (independent Codex gpt-6.1-sol, high, via the review router; full diff from merge base 93037ef1de to this head).

Regenerate the snapshot with python3 scripts/ci/defaults_guard.py snapshot.

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.4135956244901 (f024646dc3)
• cmux NEXT nightly 1.0.0-nightly.4136073207401 (9a8633089c)
• cmux NEXT nightly 1.0.0-nightly.4136174016701 (6a6302bf13)

[blacksmith-sh]

Found 7 test failures on Blacksmith runners:

Failures

Test	View Logs
/tmp/tmpfsoxuauc/summary-fail.json (live):/summary-fail.json (live):	View Logs
/tmp/tmpia4znb4u/summary-missing.json (live):/summary-missing.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
imported web bundle inputs are not routed:/imported web bundle inputs are not routed:	View Logs

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

[teamleaderleo]

Review (codex, review-31033-warm-a): SHIP at ebd97da

No blocking findings. Follow-ups: capture the adoption transition (cold and cached opens); fill in the Performance table (in progress, chat-open bench before/after on a fleet display Mac).

[teamleaderleo]

Review (next-review patch-id; carried): SHIP at 0196dec

The independent SHIP at ebd97da carries to this head: the PR's own change (git diff $(git merge-base feat-cmux-next HEAD) HEAD) has the same git patch-id --verbatim (e449bf85430ecff4) at both heads, and no merge between them was resolved by hand. Nothing new to review. A newer BLOCK, or any change to the PR's own diff, ends the carry.

[teamleaderleo]

Review (subagent claude-opus-5-5 medium): SHIP at 0196dec

Base-merge delta from ebd97da (Codex SHIP there). The incoming base touches none of the PR's files (checked: empty intersection of changed paths). The only deletions are the two Cloud repository test files reverted by #31023, which nothing here uses. No blocking findings.

[teamleaderleo]

Review (codex gpt-6-astra high, next-review): SHIP at 2f283ca

SHIP at 2f283ca

Visual verdict: unchanged by this diff

1. [P2][follow-up] direct.ts:2074-2082 treats every rejected adoption as “missing” when the background refreshSessions() promise exists, even if that refresh failed or pagination was incomplete (it returns false). This can show a transient daemon/list failure as a permanently missing chat. Use the refresh result before calling adoptedMissing.

Follow-ups:

• [P2][follow-up] Preserve failed/disconnected state when the session-list reread does not complete; only report notListedAtOpen after a successful complete list.

[cmux-bot]

cmux-bot approves 2f283ca: an independent review passed this exact head (next-review receipt by teamleaderleo). A push needs a new review; a later BLOCK at this head dismisses this approval.

A newer verdict blocks 25938ec (no independent SHIP at the head).

[teamleaderleo]

Review (next-review patch-id; carried): SHIP at 25938ec

The independent SHIP at 2f283ca carries to this head: the PR's own change (git diff $(git merge-base feat-cmux-next HEAD) HEAD) has the same git patch-id --verbatim (869cb9aa34f6de5e) at both heads, and no merge between them was resolved by hand. Nothing new to review. A newer BLOCK, or any change to the PR's own diff, ends the carry.

[teamleaderleo]

Review (independent Codex gpt-6.1-sol, high, via the review router; full diff from merge base 93037ef to this head): SHIP at 098aec3.

[P2] [follow-up] webviews/src/agent-session/acpmux/direct.ts:2081 - This checks whether the relisting Promise exists instead of whether it succeeded, so simultaneous transient list and attach failures falsely mark the adopted chat as missing. Gate adoptedMissing on the awaited result being true.