Taste Is the Moat
As frontier models commoditize raw capability, the durable advantage in creative work shifts from who can make something to who can tell what is worth making.
Jul 9, 20263 min read
Every few months a new frontier model lands and the same argument restarts. One camp says the craft is finished. The other says nothing has really changed. Both are wrong in the same way: they treat capability and judgment as the same thing.
They are not. Capability is getting cheaper by the week. Judgment is not.
The floor rose. The ceiling moved.
A year ago, "can the model do it at all" was a real question. Today the honest answer for most creative tasks is yes, at a level that clears the bar for a lot of working professionals. Drafting, comping, scoring a rough cut, turning a brief into forty directions before lunch. The floor rose, and it rose fast.
What did not happen is the ceiling coming down to meet the floor. The distance between competent and right is still enormous, and models do not cross it for you. They hand you more options, faster, which means the bottleneck moves. It stops being production. It becomes selection.
The scarce skill is no longer making the thing. It is knowing, out of a thousand plausible things, which one is worth shipping.
That skill has a name. It is taste.
Taste is not a vibe
People hear "taste" and picture something soft and unaccountable. In practice it is the opposite. Taste is a compression of thousands of prior decisions into a fast, reliable judgment about fit. Fit to the brand, to the audience, to the moment, to the medium. It is legible if you make it legible.
The studios that will matter are the ones that treat taste as an operating system, not a mood:
- A point of view sharp enough to say no to work that tests well.
- A house standard that survives contact with a deadline.
- The discipline to kill the second-best idea even when it is genuinely good.
None of that comes out of a model. All of it can be applied to a model's output at a scale that was impossible when a human had to make each option by hand.
What this changes about how we work
The old pipeline was linear and expensive: brief, concept, produce, revise. Production dominated the cost, so you rationed ideas. You picked a direction early because exploring three of them fully was a budget.
Invert the cost of production and the smart move inverts too. You explore widely and cheaply, then spend your real budget on the edit. The value moves downstream, to the moment of choosing.
old: brief → 1 direction → produce (slow) → ship
new: brief → 40 directions (fast) → curate hard → shipThat second line only works if someone in the room can curate hard and be right. Otherwise you have just automated the production of mediocre options, which is worse than making one good one, because now you have to wade through the pile.
The uncomfortable part
This is good news for people with genuine judgment and bad news for people who were selling capacity. If your value was that you could execute, execution is now abundant. If your value is that you know what is worth executing, you just got a force multiplier.
We are betting the studio on the second thing. Not because it is romantic, but because it is the part that does not commoditize on the same curve. Models will keep getting better at making. They are not on a path to caring what gets made, or for whom, or why now.
Taste is the moat because it is the one input that still has to come from a point of view. Everything downstream of it is getting cheaper. Everything upstream of it is where the work moved.