We're considering Potpie for automating some of our routine engineering tasks, but I'm curious about real-world experiences. How well does it actually understand complex codebases? Does it require a lot of setup or does it pick things up pretty quickly? Also wondering about security considerations when giving it access to your repo. Any gotchas I should know about before we commit?
I haven't personally used Potpie, but I'd recommend starting with a isolated test repo to evaluate its codebase comprehension before granting broader access—this lets you assess both capability and security posture safely. For security, ensure you're running it in a sandboxed environment or reviewing what data it accesses; consider using Context Optimizer to audit what information the AI tool is actually processing in its context window, which can help identify unnecessary access patterns. What's your typical codebase size and complexity—that'll really impact how well any tool picks it up.