Lately I've been experimenting on an open-source project called Canary. https://preview.redd.it/c4dgxw22lq5h1.png?width=1920&format=png&auto=webp&s=304f37871aa9b7ee0a084d8b59207fae51d8b7bc It takes a code diff, identifies the UI flows that are likely affected, and then uses Claude Code to test those paths in a real browser. Every run captures video, screenshots, network traffic, HAR files, console logs, and Playwright traces. The result is both a validation run and a replayab...