Coding with AI: Notes from a year of shipping production code with AI as the coding partner.

Most writing about AI-assisted programming takes a side. This book takes a notebook.In Coding With AI: A Field Report, Stephane van der Aa reports back from a year of shipping a complex production platform — fifty-six data models, a six-system security stack, a working blockchain, real-time messaging, integrated AI inference — single-handedly, with an AI as the daily coding partner. The result is a working account of what actually changed, what didn't, and which disciplines make the difference between shipping faster and shipping faster bugs.Across thirty-one chapters, the book maps the AI-paired stack as it is lived in practice: the new shape of a session, the discipline of specs as source code, knowing when to delegate and when to drive, the failure modes that recur (drift, hallucination, cargo-cult code, the confidently-wrong diagnosis), the categories of bug that only surface on the second run, and the long-running case of a blockchain anchoring service that almost ate itself before the author noticed.The book is opinionated. It names disciplines with the receipts to back them, mistakes the author has made, and judgments he would now make differently. Where the evidence is mixed or his own thinking uncertain, he says so plainly. Its central argument: trust calibration is the load-bearing skill of the AI-paired era — and review is the single layer of the workflow that cannot be delegated.Not evangelism. Not dismissal. The field report from inside the work, for the experienced developer who has decided this is now the working environment, and wants to do it well. Read more