Changelog

Trevize valley illustration with a Changelog label and the title: Verification agents and screen recordings.

Verification agents and screen recordings

A green test suite never proved the button works. Runs now open the product, use the change, and record it, so a short video of it working lands on the pull request next to the diff. You can also watch the agent's screen and take the wheel when it gets stuck.

Trevize valley illustration with the Codex logo, a Changelog label, and the title: Run your loops on Codex.

Run your loops on Codex

Codex is now a supported harness in Trevize. You can also sign in with your ChatGPT subscription and use it to power runs on both Codex and OpenCode.

Trevize valley illustration with a Changelog label and the title: Agent runs controlled by state machines.

Agent runs controlled by state machines

Every run now executes on our stable V3 workflow engine, built on a state machine that lets blueprint steps flow back and forth. Runs now fix their own failed CI checks, and agents can browse your app and post screenshots as proof of their work.