How I work

I design and ideate myself. The architecture, the data model, the thing that makes a tool worth building — that's my work, on paper, before anything runs.

For the build, I direct teams of up to 10–12 AI agents working in parallel. Each agent gets a scoped brief; I review, integrate, and verify what comes back. It's the same job as running an engineering team, compressed.

I was a software engineer before any of this existed — years of shipping code by hand, no AI anywhere. Process engineering and systems engineering came after, and each was the same skill at a different altitude: find the actual problem, then design the smallest system that removes it. The agents didn't replace anything; they multiplied an engineer who was already there.

I'm transparent about the AI use. Every case file on this site says how it was built, and because I've done this job without the tools, I can tell when the output is wrong. The verification evidence — tests, screenshots, live deployments — is mine to stand behind either way.

When I'm exploring a domain, I prototype rapid-fire: twenty-plus small tools in a day is normal. Most get thrown away. The ones that survive become case files.