About

This is one operator’s field notes on using AI: the tools I build, the strategies that survive a real week, and what my own usage costs.

Nothing here is a recommendation. Where I quote a number it is measured, dated, and attached to something I built myself.

The method is the same on every post:

1. State the mechanism. What actually drives the outcome, not what sounds like it should.
2. Rank the explanations. The competing ones, including the boring ones, including my own first guess.
3. Publish the falsifier up front. What I predict, the number, and the date it gets checked. A post with no checkable prediction is a story, not an argument.

Every post carries a date and a URL. Corrections are appended to the post, dated, and the original wording stays. Nothing is rewritten to look better than it was.

Who is writing. One person. This site is where the work is, not the person. No photographs, no biography, no employer, no location, and there will not be any. The address below is an alias that reaches me and nobody else; it is not a personal address, and mail sent to it reaches one inbox.

One thing this site cannot do. It cannot tell you whether it is working. There is no analytics, no comment count, no follower graph, and there will not be one. The method applies to the posts, not to the site. A post that states what would make it wrong is checkable by you, without me counting anything.

Contact: contact@fakemap.com

One address, answered by me. No newsletter, no comments, no analytics.