The drafting and the editing run through the same production setup these notes are about. I work with my production agents and coding agents to turn my topics and my real failures into pages. I read, edit, and approve every one before it ships.
Before a page goes up I run it past a review council I built: several AI judges, each with its own rubric and its own bias, grading against a quality bar. I run it across the site, by hand. It caught a page that contradicted itself and a statistic that did not add up, and those pages got rewritten or cut. You can read how the same gate works inside the system in the essay on certification gates.
The plumbing is boring on purpose. An ideator reads my recent code activity and proposes topics. The site has a staging copy and a production copy, and promotion is one button I press myself. The preview cards are rendered at build time. None of that decides what is true or what is worth saying. I do.
What the AI does not do: pick the topics, live the production failures, or sign off. The scars are real. The judgment here is mine. If a paragraph here rhymes with something you are building, the door is at the bottom of every page.