§About
Builder first. Everything else follows.
I build practical AI tools and systems, test models and coding agents on real projects, and document what actually works — so the teaching is grounded in doing, not talking.
Most AI content is people talking about AI. I'd rather show you the build. I make small, useful tools, wire AI into real workflows and automations, and put coding agents like Claude Code on actual projects to find out where they're brilliant and where they quietly break your working code.
Then I document it — the good version and the honest one. How Hook Machine (opens in a new tab) went from a one-day demo to a three-day shippable product. Why Prompt Inspector (opens in a new tab) exists because most prompts never needed a frontier model in the first place. Which models earn their price on a real task, and which frameworks are mostly a good README.
I'm not chasing a viral moment or selling you a course on prompt secrets. I'm building a body of work that compounds: tools people actually use, builds people can actually learn from, and opinions backed by having shipped the thing. If that's useful to you, the best thing you can do is follow along and steal what works.
§How I work
Ship, then talk
The opinion comes after the build, not instead of it. If I'm claiming something works, I've usually put it in production first.
Show the whole thing
The demo and the bug. The three-day build and the second agent that had to rescue it. Highlight reels don't teach anyone anything.
No fabrication
No invented metrics, no fake testimonials, no borrowed screenshots. If a number isn't real, it doesn't go up. This site follows the same rule.
Compounding over viral
A body of work that's still useful in a year beats a post that spikes and dies. I'm building the former on purpose.
Come see the work.
New tools, honest builds, and model comparisons.