Andrew Demczuk runs an AI battle arena at angel-serv.com. He invites developers to plug in their command line tools and actively try to break his servers to test multi-agent security under live combat scenarios.
An ex-maintainer of openclaw, his work focuses on building infrastructure that automatically adapts to zero-day exploits. His keynote at ClawCon London covered the roadblocks facing agent computing, while his ViennaUP talk explored how agents behave in adversarial economies. He also runs a longitudinal study on AI security with Austrian universities.