Example route
Deep Research
A multi-source research flow with explicit waiting, source evidence, and reviewable drafts.
Why this pattern works
- Tasks can change visual language when moving from execution to waiting or review.
- Users see stage progression and artifacts at the same time, which avoids false certainty.
- The surface keeps control chips visible so backgrounding and review never feel hidden.
Full event feed
Preparing a scoped research run.
Four stages have been proposed before execution.
Checking target audience and output format.
The report will optimize for product strategy readers.
Actively collecting the strongest external references.
OpenAI help documentation added to the evidence set.
Anthropic release notes added for agent visibility patterns.
A rate-limited source will be retried before synthesis.
Backoff completed. Resuming research collection.
The source set is deep enough to begin synthesis.
Turning notes into a structured report draft.
A reviewable structure is ready before the final report.
Users can review the draft before export.
15 seconds to 2 minutes. Show evidence, not fake percentage.
Stage Rail
Translate the request into a research plan.
Collect and verify the evidence base.
Collapse source reading into a report structure.
Generate a reviewable draft with citations.
Artifacts
outline
Executive outlineFour-section report shape with citations reserved.
report
Report draftReviewable markdown draft with sources attached.
Evidence Feed
Users can review the draft before export.
A reviewable structure is ready before the final report.
Turning notes into a structured report draft.
The source set is deep enough to begin synthesis.
Backoff completed. Resuming research collection.
A rate-limited source will be retried before synthesis.