TL;DR — While a thinking model works between steps, the agent screen now says it is waiting for the model and for how long, not "Listing" a tool that already finished. Auto-verify's checks show as checks.
Fixed
- The agent screen no longer shows a finished tool as the one running.
Between two steps the timeline kept the last tool under the active stage ("Listing: plugins", with the step bar) while the header said "Choosing the next step", for as long as the model took. At max effort that can be a quarter of an hour. It now says "Waiting for the model · 16:02", counted from when the model was asked; a retry does not restart the count. A tool's label shows only while the tool runs. The narrow agent panel says the same in its title and no longer highlights the finished call.
- Auto-verify's checks show as checks. While the build, typecheck or tests
ran after a change, the screen said the model was being waited on and named the last edit. VERIFY is now the current stage, with "Running checks" and how long they have taken, until the results are in and the model is asked to fix what they found.
- The step count no longer drops to 0 during the second tool of a reply.
Every tool result reset it, so a command that was the second call in one reply ran under "step 0/50" and an empty bar.