fix(chat): keep latest command live between messages

[maria-rcks]

Restores the single live activity row with a Lucide brain icon and keeps the latest successful command animated as Running until the next assistant message replaces it. Failed and declined tools retain terminal behavior.

Before

The completed command dropped into its past state while the turn was still working.

After

Thinking starts in the shared activity row with the brain icon.

Once a command appears, it replaces Thinking on the same row and remains animated as Running until the next assistant text.

The second after screenshot was captured after the server persisted tool.completed at 21:35:40.436Z and before the next assistant message at 21:35:44.125Z.

Verification

• 88 focused timeline tests
• scoped lint
• web typecheck
• real web client at 1920x1080 against an isolated T3 environment

Created by gpt-5.6-sol through the Codex harness.

Note

Unify thinking and latest tool activity into shared live-activity-row in MessagesTimeline

• Replaces the separate thinking-indicator-row with a single shared LIVE_ACTIVITY_ROW_ID ('live-activity-row') that covers both initial thinking and the latest tool activity, so the latest command stays live between messages.
• deriveMessagesTimelineRows now treats the activity as live when a tool entry is running or the latest visible tool indicates success (workEntryIndicatesToolSuccess), setting the active flag and using the shared id; otherwise it falls back to a per-group id with active false.
• Renders the thinking row with an explicit brain icon by adding 'brain' to WorkEntryIconName and mapping workToneIcon tone 'thinking' from 'bot' to 'brain'.
• Moves thinking rows into the same bottom-padding group as work rows in MessagesTimeline.tsx, slightly increasing their vertical spacing.
• Behavioral Change: completed successful tools now keep the shared activity row active instead of ending the live state, and all live states reuse one stable id; consumers keyed on the old thinking-indicator-row id or the separate thinking row need updating.

^{Macroscope summarized 64c8012.}

---

Note

Low Risk
Scoped to chat timeline row derivation and presentation; stable row id changes may affect tests or selectors that targeted the old thinking/work-live ids.

Overview
Unifies in-progress “Thinking” and the latest tool into one shared timeline row (live-activity-row) so the UI does not swap row ids or drop completed commands back to a past “Ran” state while the turn is still working.

deriveMessagesTimelineRows now treats the activity slot as live when a tool is running or the latest visible tool succeeded (workEntryIndicatesToolSuccess), keeping the Running shimmer on that row until the next assistant message. Declined (and other non-success) tools stop occupying the live slot; Thinking returns on the same row id. The thinking state uses a brain icon instead of the bot icon.

Tests and markup expectations move from thinking-indicator-row / per-group work-live:* ids to live-activity-row where the shared handoff applies.

^{Reviewed by Cursor Bugbot for commit 64c8012. Bugbot is set up for automated code reviews on this repo. Configure here.}

[macroscopeapp]

Approvability

Verdict: Approved at 64c8012

Macroscope's review found this PR approvable — This is a narrowly scoped chat timeline fix that keeps the latest successful command visually live until the next assistant message, with declined and failed tools retaining their existing behavior. Production changes are confined to row derivation and presentation, and focused tests cover the transition states.

Notes:

• No code objects were reviewed. Approvability was decided on eligibility alone.

^{You can add or adjust custom eligibility rules. Learn more.}

Dismissing prior approval to re-evaluate 64c8012