Ask AI, rebuilt as one console: inline evidence, a visible plan, live artifacts
Chat and the composer are now one unified console:
- Test evidence — screenshots, traces, failure detail — shows up inline in the conversation, not only inside a separate artifact.
- The composer shows a context readout and the agent's own plan before it acts.
- Thread telemetry is available behind the context chip, for anyone who wants to see what the assistant actually used to answer.
- Fixed along the way: dropdowns rendering off-screen, dead whitespace under messages, the context meter failing to appear, and questions getting dropped when a conversation branches.
See Ask AI.