He Paused His Master's and Built an 80,000 Star Repo
About this episode
Magnus Müller raised $17M for Browser Use — the open-source layer with 80,000+ GitHub stars that lets AI agents actually use websites. Midway through this interview he says out loud that if his integration with Anthropic's Claude Code succeeds, Claude Code can absorb the value natively. Then he opens his eval platform — Claude Code agents on cycle 10, ~100 hours of runtime, $100-200K/month on LLMs — and explains why he's calm: he hitchhiked from Germany to Iran with a hammock, and your brain screams with worries every night but everything ends up fine.
Key moments
- 00:00 He built an AI company while showering at a gym
- 00:57 How did two students build Browser Use from a coworking-space couch?
- 04:14 What does Browser Use actually do?
- 06:27 Will Amazon's lawsuit kill AI shopping agents?
- 10:28 Will AI start prompting you instead of the other way around?
- 13:04 How does his personal 6 AM AI agent work?
- 14:51 15:32 in the edited transcript). Claude Code replacement framing follows the strategic-integration read (46:29), not company-level failure. The eval-platform demo IS referenced with verbatim numbers: Claude Code agents on cycle 10, ~100 hours runtime, $60/run for 100 tasks, $100-200K/month LLM cost — all from the
- 15:41 What tech stack does a personal AI agent need?
- 18:20 What breaks when you build a browser agent from scratch?
- 23:15 How do you fix non-deterministic AI agents?
- 26:00 Can AI agents learn from each other?
- 32:19 What does a real production AI eval platform look like?
- 41:12 Why hire 19-year-olds over senior engineers?
- 46:05 Could Claude Code replace Browser Use?
- 50:38 What does hitchhiking to Iran teach you about being a founder?