Solo founders/ small team who are writing code with Codex or Claude Code:
Speed vs Accuracy ❌
Speed vs Codebase maintainability ✅
Accuracy per session is mostly solved by the frontier models, but my team was struggling with codebase maintainability.
As the product grows, AI agents change files, code gets rearranged, specs become outdated very quickly, and there’s no clear history of why decisions were made. There is a lot of overhead per session. There is no accountability inside the agentic system. If any bug is found inside the system, who is responsible for finding that bug and solving it without rearranging the whole codebase?
I’ve been facing this exact problem on my prod codebase. So, I have a solution for that. I Increases my codebase's auditability, documentation, and accountability, and it reduces the chances of architectural changes and code changes during an iteration.
So, I want to write the solution as a blog. Before I write about it: is this a real pain for you?
Reply with your biggest codebase-management problem, or just say "codebase". If enough people are dealing with it, I’ll share the system I have developed for my codebase. Happy to help. :)