Hey everyone — builder here. Three months ago I was revisiting a module I had built myself two weeks earlier. Sat staring at it for twenty minutes. Could not explain my own code. The AI had written significant chunks of it, I had reviewed and merged, and completely moved on. That moment bothered me more than I expected. We talk endlessly about LLMs having a context window like it is a fundamental limitation of the technology. We never apply the same framing to ourselves. Developers have a context window too. AI-assisted development is just filling it faster than any human brain can track. The existing answers did not work for me. Writing docs separately from coding means docs that lie within a week. Asking Cursor to document as it goes means a new README every three lines with zero architectural context. What I actually wanted was simple — documentation written automatically at the one moment developers never skip. The commit. So the changelog ID is the commit hash. The docs live in your repo as plain Markdown. The AI reads your actual codebase to answer questions, not its training data. Zero extra workflow. Everything stays in sync because it has no choice but to stay in sync. The thing I did not expect when building this — the `.devmem/` folder ends up being useful context for AI tools too. Cursor reading your actual module documentation is meaningfully better than Cursor guessing from the file it has open. One knowledge base. Useful for your teammates and your AI tools simultaneously. Still rough in places — module detection on unconventional project structures needs work and large messy refactor commits stress-test the capture prompt in ways I have not fully solved yet. Fork it, break it, tell me what is wrong. That is the only way it gets better. — Surya

DevMem

Ur codebase has a memory now.updates itself on every commit.

Ur codebase has a memory now.updates itself on every commit.