Agents that bring back video proof that the work is done
Published
How can AI replace clicking through every change a coding agent made to check it really works before merging?
- Replaces: clicking through every change a coding agent made to check it really works before merging.
- What to connect: E2E tests plus Claude recording videos and screenshots of the running app as required evidence before a task counts as done.
- How it works: Warp does the same in Warp Factories, where each agent gets a computer and its 3 to 4 minutes of activity are cut down to about 15 clips.
- Results: Warp says this has sharply cut its time to merge.
- Who it's for: developers and small teams who hand features to coding agents and review the results.
From the Morning Tokens issue of Oct 25, 2026