How this site is made.
Most of the words on this site were drafted by AI, riding along with me. I don't hide that. This page is the operating manual — what the bots actually do, what they're not allowed to touch, and what I catch when I check their work.
Drafting, research, scaffolding, and most of the commits on this site come from AI assistants working under direction — Claude, mainly, in a few different modes: chatting through an idea with me, or working directly in the codebase with the repo open.
What they're allowed to do: propose structure, write first drafts, build components, fix bugs, commit changes. What they're not allowed to do: decide what I believe, or ship anything I haven't read.
Judgment is not delegated. Every position this site takes, every call on what publishes and when, and anything that leans on my name has a human behind it — me.
A bot can draft an argument; whether the argument is one I'll stand behind is not a bot's decision. If something here is wrong, that's on me, not on "the AI."
Here's an actual one, not a hypothetical. The essay "Stop hiding the seams" started as a one-line prompt: I wanted to write about why this site treats AI collaboration as something to celebrate rather than disguise. Claude drafted a structure and a full essay from that.
I made two calls before a word was written — which section of the site it belonged in, and how long it should run — and the draft came back matching both. The editing after that wasn't about fixing the argument; it was about making the delivery honest, too: making sure the essay's claim about auditable process was actually true of the process that produced it, not just a claim the essay made about itself. That turned into this page.
Every post on this site carries a small marker near the byline.
Zack wrote it. A bot may have fetched a link or fixed a typo.
AI drafted it under my direction; I reshaped and edited it. This is the default.
An agent generated it end to end (changelogs, data-driven posts); I reviewed it before it published.
None of the three change who's accountable for what got published. That's always me.
This isn't a claim you have to take on trust. The repo behind this site is public, and every commit a bot made carries a Co-Authored-By trailer, so you can see exactly which changes came from a bot and which didn't.
The actual instructions the bots work under are public too — AGENTS.md — which is rarer than it should be. Most sites that mention "AI-assisted" don't publish the instructions their AI is running under. This one does.