How AI is used

This site is about patterns that help humans and AI agents produce high-quality code together, so it seems only fair to say exactly how AI is involved in producing the site itself. The short version: the posts are written by a human and reviewed with AI; the code is drafted by AI and directed and reviewed by a human.

The posts

Every blog post starts as a fully human-written draft. AI then acts as a reviewer, not a ghostwriter:

• It fixes mechanical issues — grammar, idioms, typos — and those fixes are applied directly.
• For structural concerns (pacing, repetition, ordering), it proposes options, and I choose the one that fits — or none of them.
• After a final self-review, remaining small issues get the same treatment: AI suggests, I decide.

The ideas, the opinions, the structure, and the words are mine. Every post carries an end-note linking back to this page.

The code and the site

The relationship inverts for code. The site — and the libraries, skills, and tooling in the wider project — is largely AI-drafted: I direct the work, review the results, and decide what ships. Chrome pages like this one follow the same path: AI drafts, I rewrite and approve.

That inversion is deliberate. The whole project explores how far human-directed, AI-produced code can go when it’s held to explicit patterns and guards.

What AI doesn’t do

• No AI-generated posts. A post is never drafted, in whole or in part, by AI.
• No AI opinions presented as mine. Judgements about patterns and trade-offs are my own.
• No unreviewed output. Nothing — prose or code — ships without a human reading it.