Hi HN, I'm the creator of Hound, an open-source tool that uses AI agents to hunt for deep logic bugs in codebases. It works by building knowledge graphs of the code, spawning agent teams to generate and refine hypotheses, and simulating expert auditing processes. In tests, it replicated 10/30 findings from a real audit contest. It's language-agnostic and runs locally via CLI.
More details and examples in my blog post:
https://muellerberndt.medium.com/unleashing-the-hound-how-ai...
Excited to hear what you think!