Open source AI code review - that's the right move. Devs won't trust closed-box AI with their codebase. How are you handling false positives? That's usually where automated code review gets annoying - flagging things that are intentional.