Created a Github bot that can review PRs to catch bugs and security issues. It has been useful as a quick first check for issues before diving into architecture changes. Also, discovered while building this that current LLM eval tools suck.