Skip to content

Agent: stop the thread jumping when a message sends - #1582

Closed
AnthonyRonning wants to merge 1 commit into
codex-gpuibump-motion-heightsfrom
codex-gpuibump-motion-send
Closed

AnthonyRonning wants to merge 1 commit into
codex-gpuibump-motion-heightsfrom
codex-gpuibump-motion-send

Conversation

@AnthonyRonning

@AnthonyRonning AnthonyRonning commented Nov 12, 2026 •

Copy link
Copy Markdown
Contributor

Seventh PR of the stack. It fixes what showed up when the motion was used against a real backend: two jumps when a multi-line message sends, and a composer move that read as lag.

How it was found

A multi-line send in the top-of-stack build was screen-recorded and stepped through frame by frame. Two bursts of movement stood out, each lasting a few frames.

What was wrong, and the fix

1. The sent message's row never took over from its bubble. The motion work hands the bubble off to the real row so the row appears where the bubble was. That hand-off was keyed by the id the event names. The transcript files a row under its view key (timeline_view_key), and a real host's rows have one, so the ids never matched. The row entered as a brand-new row, faint and growing from nothing, on the same frame its bubble vanished: the thread dropped, then climbed back.

The hand-off now uses the view key. The existing tests passed because their rows carry no identity, so both ids were the same string; a new test sends a row with a live identity and failed before the fix (the row started 41 px below the bubble).

2. Sending from a tall composer dropped the thread first. The composer emptied at once, giving up several rows, while the bubble's slot opened from nothing. For one frame the thread fell into the gap, then glided back up as the slot opened.

The slot now opens from the height the composer gave up. The newest row holds still on the send frame and then only moves up, by what the bubble needs beyond that. Clearing the field also resets its eased height on purpose now, so a send always empties it on that frame.

3. The composer's move between a new task and a conversation is instant. It used to glide over 200 ms on the first message, on opening a task from the empty state, and on New Task. Where the user types is not something to animate, so that glide and its state are removed.

Validation

On macOS, at this head: cargo test -p maple-agent-app --locked with -D warnings, and the repository pre-commit hook. New tests:

  • a row keyed by its live identity still takes over from the bubble, with older rows holding still;
  • sending a five-line message never moves the newest row down;
  • the composer is at its new place on the first frame, both ways.

Not verified

The fixes have not been watched in a running app yet. The recording that found the bugs should be repeated on a build of this PR.

🤖 Generated with Claude Code

Recording a multi-line send against a real backend and stepping through
the frames showed two jumps, and the composer's move between a new task
and a conversation read as lag.

- The sent message's row never took over from its bubble. The hand-off
  was keyed by the id the event names, but the transcript files a row
  under its view key, and a real host's rows have one. The row entered
  as a new row, faint and growing from nothing, on the frame its bubble
  vanished, so the thread dropped and climbed back. The hand-off now
  uses the view key. The tests passed because their rows carry no
  identity; a new test sends one that does.
- Sending from a composer several rows tall dropped the thread first.
  The composer emptied at once while the bubble's slot opened from
  nothing. The slot now opens from the height the composer gave up, so
  the newest row holds still on the send frame and then only moves up.
  Clearing the field now also resets its eased height on purpose.
- The composer changes place at once between a new task and a
  conversation. Where the user types is not something to animate.

Co-Authored-By: Claude <noreply@anthropic.com>
@AnthonyRonning AnthonyRonning added the ci:package Build unsigned packages on this PR label Nov 12, 2026
@AnthonyRonning
AnthonyRonning removed this pull request from stack #1574 November 12, 2026 03:37
@AnthonyRonning

Copy link
Copy Markdown
Contributor Author

Folded into #1575. The rework of the motion layer (springs and timing, the hover call, the shared composer numbers and the send fixes) is one PR now instead of four. This commit is unchanged and is on that branch.

@AnthonyRonning
AnthonyRonning deleted the codex-gpuibump-motion-send branch November 14, 2026 16:22
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ci:package Build unsigned packages on this PR

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant