# Design Breeding Captures Method

Idea capture is the first stage of an operating loop.

A person speaks from intuition. The words arrive tangled: taste, fear, product promise, scene direction, half-metaphor, a rejection of something she has not fully named. The old workflow would transcribe the tangle and call the transcript a brief. The new workflow asks what in the tangle can run.

The first pass is compression. What has to stay true? What is only local costume? Which objection is actually a principle? Which delighted sentence is a gene? The point is not to summarize the speaker. The point is to extract enough first-principles structure that the next artifact can be generated from it.

The second pass is operationalization. First principles become mutation rules. Preserve the living trait. Reopen the accidental traits. Produce siblings, not one answer. Let the population show the operator what her taste meant. When she reacts, treat the reaction as new data about the generator.

This is why design breeding is different from ordinary mockup production. A mockup asks, "Is this the page?" A breeding run asks, "What did this page teach the generator?" The output is not only a selected tile. The output is the parent, the population, the feedback, the rejection signal, the mutation rule, and the stop note.

Without that record, the next pass inherits a specimen. It copies palette, silhouette, copy density, and UI furniture because those are visible. Similarity replaces understanding.

With that record, the next pass inherits a gene. If the living thing was motion around a center, the container can disappear. If the living thing was softness, the geometry can mutate. If the living thing was trust before explanation, the form can wait.

The artifact matters because it is visible proof that the loop reached the world. A process note can claim that capture became principles and principles became action. A working surface can show it. Six variants on a screen, a button with motion around it, a transition into a call to action, a small form, an ending state: the method stops being a theory and starts answering to taste.

That is the 2026 change. The expensive part used to be crossing from thought into artifact. Now the expensive part is having enough taste, boundary, and procedure that the artifact is not random. AI tools collapse the cost of making visible candidates. They do not remove the need to know what should be preserved.

The system needs organs because the same run is not the same object everywhere.

The experiment workshop is the generator. It is internal, messy, moonshot-oriented, and allowed to overproduce. The public graph is the brain. It compresses the method that survived into a node another mind can read. The company is a body of work. It imports stable building blocks, obligations, promises, privacy boundaries, and work orders, then sends pressure back into the brain when reality answers.

These organs are not silos. They are symbiotic. The boundary is what lets them exchange without becoming confused. The workshop should not be forced to speak like a company commitment. The company should not inherit raw lab chaos. The public graph should not leak private context, but it should be able to publish the method when the method stands on its own.

Design breeding captures method at the moment a spoken intuition becomes an executable lineage.

The visible artifact is not the decoration around the idea. It is the receipt that the idea crossed the boundary into operation.