The Selector
The Selector
Every creative system has two roles: someone who produces, and someone who selects.
For most of history, production was the bottleneck. Books were hard to write. Films were hard to make. Songs were hard to record. Selection mattered, but it sat downstream — an afterthought to a supply that was always constrained.
AI inverts the ratio.
Now anyone can produce anything. A song. An image. An essay. A product page. Supply becomes infinite overnight. What was scarce — production capacity — turns into the cheapest input in the system.
The bottleneck moves.
It moves to the selector.
The person — or institution, or brand — that decides which thing among infinity gets brought forward. Out of the ten thousand songs the model just generated, which one matters? Out of the million product variants your tools created, which one a customer should ever see?
Selection used to ride along with production. The same entity wrote and decided. Now the two split. Producing is free. Selecting becomes the entire job.
This is not a new role. It already exists in every cultural field. You just stopped noticing it.
A&R at a record label. An editor at a magazine. A curator at a museum. A buyer at a fashion house. A critic. A producer like Rick Rubin who, by his own admission, does not produce — he selects what is already there.
These were always quietly the most powerful nodes in their networks. AI makes that explicit.
Selection cannot be trained.
This is the part that confuses people. They assume that taste, since it is observable in outputs, must be learnable from outputs. It is not.
Taste is a function of who the selector has become — the books they read, the people they spent time with, the failures they survived, the things they walked away from. It is the residue of a particular life.
A model can mimic the surface of a selector's choices. It cannot reproduce the formation that led to those choices. That formation requires time and exposure that no training run compresses.
This is why cultural capital is durable. Production scales infinitely. Selection scales only with lives lived.
Three classes of selector accumulate the anchor function.
Curators. People and institutions whose explicit job is to choose. The New Yorker. Criterion. The buyer at a Milan boutique. The A&R at a label. Their value is the discount they save the audience from sorting through noise. AI multiplies the noise. Their value compounds.
Tastemakers. Individuals whose perspective is the selector. Rick Rubin's choices became Rick Rubin. Steve Jobs's taste was the asset Apple built around. The Hermès atelier does not follow trends — it embodies a position that others orbit. The taste lives in the person. The person can be brand-anchored. The taste cannot be cloned.
Trust-bearers. Brands and institutions that earn the right to vouch. The Michelin star. The university degree. The investor's reputation. They don't make the thing. They certify which thing deserves your attention. In an output-flooded world, certification is the scarce service.
This is the cultural capital pillar.
Money × tech earns by routing capital through new rails. Culture × history earns by anchoring attention to specific human signals in a sea of machine-made output.
Both are infrastructure. The first is the infrastructure of value movement. The second is the infrastructure of value recognition.
Both compound slowly. Both are hard to fake. Both become more valuable as their commodity counterpart — the rails-vehicles, the produced output — gets cheap.
So this is the thesis.
When making becomes free, choosing becomes the asset.
The companies that matter in culture for the next decade will not be the ones that produce the most. They will be the ones that anchor attention to a particular sensibility — a human one, formed slowly, defended consistently.
Production is infinite. Selection is the moat.