practice - context right in the repo

Practice

I think this mostly means having something like a documentation context graph right in the codebase alongside the code.

Status 🟡

My context is in my main markdown graph (more specifically the documentation context graph), which is in a different repo and is copied in. The challenge is that it's about Meadow, but also links to many adjacent things about psychology, indie hacking, etc. This doesn't feel quite right, yet, and I'm struggling to find an optimal solution.

Related

There's this related idea of carefully providing context so the agent can use it. For example, there's another practice: practice - observability stack in dev and test that empowers the agent to figure problems out on its own.

Sources

blog post - OpenAI harness engineering 2026-02