Read the codebase before you change it
An AI that doesn't understand your project's patterns will write code that works but doesn't fit. Index the codebase first so new code matches what's already there.
Every project has its own style. How files are named, how errors are handled, how components are organized. When someone new joins a project (human or AI), they need to learn these patterns first. Otherwise they'll write code that works in isolation but clashes with everything around it.
At the start of a session on any project, take a few minutes to review the existing patterns. What naming conventions are used? How are files organized? What does the typical component look like? This context helps every subsequent decision match the existing style.
Re-do this after major reorganizations, when the patterns have shifted enough that old assumptions are wrong.
First session on any project you didn't build. Also after big restructuring on your own project, when things have moved around enough that the old patterns no longer apply.
Product leader shipping across enterprise SaaS, AI in production, and 0→1. Writing about what actually ships — not what sounds good in a deck.