Some rooms tell you exactly where you stand the moment you walk in. This one did. Engineers with years on me, laptops already open to codebases they’d clearly been shaping for days. And me, with an empty editor and six hours on the clock.

I could have read that as a disadvantage. It was the whole bet: same clock for everyone, six hours, and the only thing that counts at the end is what runs.

The idea: talk to it, watch it think

I built SENTIENT, a generative-UI architecture canvas. You talk to an AI about what you want to build, and instead of handing you a wall of text, it draws you a plan. A roadmap. An interactive node graph you can move through, expand, and reason with.

Most AI gives you paragraphs. I wanted one that gave you a map you could walk around inside.

That meant, in six hours, wiring up the conversation layer, turning free-form intent into structured architecture, and rendering it as a living graph on a canvas. From a blank screen.

How you move that fast

People assume speed like that is about typing fast or knowing every API by heart. It isn’t. It’s about three things I still lean on every day:

Six hours later, SENTIENT ran. You could talk to it and watch a project take shape as a graph in front of you. And it won, over a room of people who’d started long before I sat down.

What it actually proved

The head start other people had wasn’t the moat. Knowing exactly what to build, and building nothing else until it ran, was.

SENTIENT was one of seven or eight tools I’ve shipped end-to-end, most of them self-taught. But it’s the one I point to when someone asks whether I can actually build, or just talk about building. Six hours. One blank screen. That’s the answer.