Why I built this

I maintain the UNICORN Binance Suite — a set of open-source Python packages for building trading systems on Binance, some of it running in production for years. I work with an AI coding agent on it every day, and early on I put real effort into the harness around that: conventions, memory, context that lets the agent actually learn the codebase instead of starting from zero every session.

That paid off in a way I didn't fully expect. I can hand the agent a piece of code and ask it something about the project, and it answers like it's been there the whole time — because, in a sense, it has. Watching an AI go from knowing nothing about a codebase to genuinely proficient in it, just by giving it the right context to learn from, is something I find genuinely cool.

That's what led to the actual insight: if context can make an AI proficient in a codebase whose history it never lived through, the same context makes a human developer portable between codebases too. The knowledge stops being trapped in one person's head — it travels with the project instead.

And the "why" itself is basically exhaust from that process, not separate work. You define the conventions once, and after that it just keeps accumulating out of the conversations you're already having — the longer you work on a project this way, the more it's built up on its own.

Keep the Why is that idea, built out. Nothing more complicated than: this seemed worth doing, so I did it.

More on how I work with AI agents day to day: blog.technopathy.club · GitHub