Building the harness.

The path here was anything but straight — and that turns out to matter.

Ten years building and running The Grillin Greek was an education in something no curriculum offers: how to create a functional system within a living ecosystem. Logistics, supplier relationships, customer trust, a kitchen that has to work at 7pm on a Saturday regardless of what broke at 4. When you truly discover what to build and then provide for the people in your community, it teaches a great number of lessons — about scale, about trust, and about what people actually need versus what they say they want. Those lessons map directly to the work now.

Art school taught the discipline of creation under constraint — trial, error, intention, philosophy, the kind of mindful practice that makes you slow down enough to see what you’re actually building. Five years inside nonprofit organizations added a different layer. You watch institutions accumulate processes designed to protect their own continuity rather than serve their mission. Administration sealed from the reality of the people below it. Resources disappearing into overhead. What stays is a visceral understanding of the disconnect — and a clear-eyed view of what it takes to close it.

Then Unity, for the logic of interactive systems. Then Python, and everything that followed. Then the frontier models arrived and the entire frame shifted. Andrej Karpathy has called this moment the emergence of a new kind of computer — one you direct through language rather than code. Open source models catching up to proprietary ones in months. Reasoning systems that plan before they generate. Agents that persist, learn, and operate while you sleep. We are watching the most compressed science fiction revolution in human history unfold in real time, and every builder who understands what’s happening has an obligation to be building.

No CS degree. Karpathy said it plainly: the hottest new programming language is English. But understanding systems — how they fail, how they scale, how people actually use them versus how you expect them to — that understanding is what separates a tool from infrastructure. The creative operator who can think clearly, communicate with precision, and understand both the problem and the person it affects — that person has leverage that no bootcamp produces. The harness is the work: the infrastructure, agents, memory systems, and coordination layers that let frontier AI do useful work for real people in the real world.

What I'm Building Right Now

Infrastructure & Tools