I wanted software that could build itself. Not generate code and cross its fingers - actually build. Plan the work. Write it. Catch its own mistakes. Ship it clean.

So I built an engine that does exactly that. It breaks a vision into pieces, implements each one, validates what it wrote, and only moves forward when the work holds up.

Sixty stories shipped. This site is one of them.

That was supposed to be the whole project.

But the engine didn’t know anything. It could build whatever you told it to - brilliantly, reliably - but it had no opinion. No taste. No memory of what worked before.

I kept thinking: what if it didn’t just build what I asked for, but understood why I was asking?

So I started mining knowledge. Real expertise. Real voices. Not scraped - curated. Creators I trust. Advice I’d actually follow. Distilled into something a system could reason over.

Now the engine had instincts.

One voice isn’t enough. A single opinion is just a guess with confidence.

So I built a council. Independent knowledge systems - each with its own perspective, its own expertise, its own blind spots. They don’t take turns. They converge. And nothing ships until they agree.

When they disagree? That’s where it gets interesting. Disagreement isn’t failure. It’s where the best decisions hide.

The council could reason. The engine could build. The knowledge was deep.

But it was still building for everyone. Not for you.

So I built the last piece - an intelligence that learns who you are. Not from a quiz. Not from a form. From the way you talk. The choices you make without thinking. The things you keep coming back to.

And then one morning, over coffee, they connected.

The engine. The knowledge. The voices. The identity.

The answer isn’t generated. It’s selected.