browser-use just dropped video-use -- and it''''s exactly what it sounds like. AI-powered video editing through coding agents like Claude Code and Codex. Drop raw footage in a folder, describe your edit in natural language, and get final.mp4 back. No timeline, no presets, no menus. It handles everything: cutting filler words (umm, uh, dead air), auto color grading, 30ms audio fades, subtitle burns, and even generates animation overlays via HyperFrames, Remotion, or Manim -- all spawned as parallel sub-agents. The repo just hit trending on GitHub with 700+ stars. What makes this wild is the self-evaluation loop -- the agent renders every cut boundary, reviews its own output, and only shows you what passes inspection. Session memory persists in project.md so it picks up where you left off next week. This is the direction tooling is heading: not another GUI app, but agent-native workflows where you describe what you want and the system figures out the how. For anyone building automation pipelines or content workflows, this is worth a deep dive. Full breakdown by Alex Hitt linked below. Full video link https://youtu.be/pSeyHZ1Q-BM