Context Engineering for AI

Originally published at collinwilkins.com This one is long and a more detailed follow up to my first article on the topic on this Two things before we get into it: first, you'll walk away with at least one thing you can apply this week to get more consistent results from your AI coding tool. Second, the examples throughout use Claude Code — that's my daily stack, not an endorsement. Every principle here applies to Cursor, Copilot, or whatever you're running. Let's start with an example that probably happened to you this week... You asked your AI coding tool to add a new API endpoint. It generated exactly what you need — right naming convention, file location, and imports. You closed the task in 15 minutes. Next morning, you asked for another endpoint. It used a naming pattern from a framework you dropped three months ago. The file landed in the wrong directory. It imported a library that's no longer in the dependency tree. You spent 40 minutes cleaning it up. Then a teammate tried the