Long-form versions of what I keep learning: one published, one I am writing in public.
A field workbook for engineering with AI coding agents. Turn each repeated mistake into a small, tested guard that lives in your repository and works with whichever agent, git hooks and CI system you use.
“If an AI coding assistant makes a mistake, we can correct it. But if it makes the same mistake next week, perhaps the problem isn’t the agent’s memory.”
The method
From the book’s blog
Turn unfinished ideas into small experiments with AI
AI made building cheap. You can have a working prototype in an evening. What did not get cheaper is everything around the build: choosing what to make, getting anyone to see it, and deciding honestly whether it worked. This book is about that part.
Built around one loop
I’m writing it in public: new chapters land on Leanpub every week, free while it is in progress, and I post along the way on X.