Deep Research progress now writes to a durable audit trail as agents run. That means:
- Refresh-safe progress. Reload the property workspace mid-run and every thinking step that already executed reappears in the chat accordion. No more 'where was I?'
- Catch up later. Close the tab during a Deep Research run, come back hours later, and the full thinking-step history is still there.
- Cards and chat in lockstep. The section cards on the left rail and the chat accordion poll the same canonical status feed, so what you see in one always matches the other exactly.
Combined with the 5–10 minute Deep Research and 7–20 minute Financial Pro-Forma runtimes, long-running analyses feel a lot less fragile.