AI

Much of what I publish here is written with AI. I hand agents rules to discourage bad habits: theirs and mine.

The ideas, the framework, and the judgment about what’s worth saying are mine. The drafting, copy-editing, and refinement are usually a collaboration with an LLM. Every co-authored post carries a colophon at the bottom naming who did what. If one of them is wrong, that’s on me, not the model.

The same goes for code. I commit as @nonrational. When I delegate significant steering to a machine, it commits as @nonreagent, so the two ledgers stay separate.

I publish how I work because people I admire published how they work, and I got better by reading it. My dotfiles, my agent config, the skills I use to write and to learn from others: all public. The time I spent making them and the things I learned aren’t meant to be hoarded. Take the parts that suit you and leave the rest. Make it your own. The same tools in someone else’s hands make something I couldn’t have made, and that makes us all richer.

Taking freely obliges you to say where it came from. I name what I borrow, partly as a breadcrumb back to the source and partly because someone went first and that deserves to be visible. The colophon is that same habit pointed at a machine. I don’t run a quieter rule for probabilistic character generators.

Saying so is the easy part.¹ I write here to reason an idea all the way through, and it isn’t through until it’s clear on the page. I’m in no rush to get there. The writing has to stand on its own.