Prefab or cathedral

I keep coming back to one thought. AI made code cheaper. That is what lets you build a cathedral instead of a prefab. Not a better prefab. A different thing.

Most software is prefab. Ready rooms, ready walls. You pick from a list and move in. It works and ships and nobody remembers it. Nothing wrong with that. Houses are fine. It is good at being a house.

A cathedral is different. Nobody ships a cathedral. Somebody stands on the site and builds it, not from a kit. Cheap code is what makes that possible now. Before, the stone itself cost years. Now the stone is almost free, so the only real cost is deciding what the building should be. Portal was my try at that. Not another app glued from parts everyone already uses. One board, drawn by the GPU, and the architecture we designed ourselves. Every tile is where it is because we put it there. Nothing lines up by default.

Portal The unreal engine for AI agents — one canvas, every model, every agent. →

These videos are the same idea, just smaller. One model painted the glass. Another looked at the frames and told me what it saw. A third one will do the voice later. Sound, image and motion used to be three different jobs. Now i can do all three in one evening, badly, but done.

What stuck with me: the gold letters on the covers were sampled from the film itself. Not a font i picked. The picture decided. You cant get that from a template. You have to want it.

Code got cheap because AI got good. That is the whole change. What costs now is time and caring about the result. Prefab asks for neither. A cathedral asks for both, and for the first time, almost anybody can pay it.

Every new project you start, you pick again. Prefab is faster. Cathedrals take longer. But they are the ones people keep.

More on the channel.