Proof and Imagination

Tens of thousands of people use Atypica every day. It’s an AI agent for consumer insight — a tool that lets you talk to simulated users before you talk to real ones. And if you watch those tens of thousands of people long enough, a pattern emerges. They all fall into one of two groups.

The first group comes to prove something.

They already have an answer. Maybe it’s a hunch. Maybe it’s a genuinely brilliant intuition about what their customer wants. But they can’t just say it. Their boss wants data. The organization runs on logic, on quantification, on the comforting architecture of rationality. So they need a process — a way to translate what they already believe into a language that everyone else will accept. They’re not looking for discovery. They’re looking for certainty.

The second group comes to imagine.

They don’t have an answer. They barely have a question. They sit down with an open mind, start a conversation with an AI, and wait for the thing they never would have thought of on their own. They’re not looking for confirmation. They’re looking for surprise.

Both groups call what they’re doing “consumer insight.” But they are doing entirely different things.

Proof is an act of convergence — you draw a circle, then you fill it in. Imagination is an act of divergence — you don’t know where the edges are, which is precisely why you might end up somewhere genuinely far away.

Here’s the thing about being smart: it doesn’t save you. No matter how intelligent you are, you still have one set of experiences, one cognitive frame, one way of filtering the world. Everything you can think of is bounded by everything you already know. The value of AI is not that it’s smarter than you. It’s that it puts you in a space outside your own experience — lets you hear voices you don’t normally hear, and see the structural gap between your mental model and the way things actually are. That gap? That’s insight.

David Deutsch, the physicist, wrote in The Beginning of Infinity that knowledge doesn’t grow by accumulating more data. It grows by producing better explanations — what he calls good conjectures. A good conjecture is one that’s hard to vary without breaking it, which is what gives it real explanatory power. But good conjectures don’t appear out of nowhere. They need a precondition: you have to first encounter something that surprises you. Insight is the raw material of conjecture.

We are living in an age where supply is approaching infinity. Everyone can use AI to write code, to design, to generate content. Every advantage on the supply side will eventually be competed away.

The only scarce thing left is the imagination to see demand that no one else has seen. And that imagination begins with a single, unexpected insight.