When making is cheap, taste decides

For most of the history of design, making was the expensive part. A photograph needed a shoot, a page needed a developer, a campaign needed weeks. The cost of making kept the volume down, and a low volume gave every piece a chance to be looked at properly.

That cost has collapsed. A model now produces a credible image, a tidy landing page or a reasonable post in less time than it takes to describe it. What has not become cheaper is knowing which of the hundred options is right, which is almost right, and which should never leave the building. When making is cheap, the scarce part moves to the decision. The old word for that decision is taste.

Taste is not a preference

Taste is often waved away as personal. The programmer and investor Paul Graham argued against that more than twenty years ago, in his essay Taste for Makers: “Saying that taste is just personal preference is a good way to prevent disputes. The trouble is, it's not true.” His conclusion was practical: “We need good taste to make good things.”

The record producer Rick Rubin gave the clearest working definition on 60 Minutes in 2023. Asked whether he knew how to work a soundboard, he said: “No. I have no technical ability. And I know nothing about music.” Asked what he was paid for, he answered: “The confidence that I have in my taste and my ability to express what I feel has proven helpful for artists.”

There are two parts in that answer, and both matter more now. Knowing what is good. And being able to say why, clearly enough that someone else can act on it. The second part is the one a machine needs from you.

Why the machine drifts to the middle

A generative model is trained on what already exists, and without direction it produces the most likely version of what it has seen. That is useful for a first draft and fatal for a brand. Dylan Field, who runs the design platform Figma, made the point in an interview with Stratechery in June: AI is trained on the distribution, so the people using it have to be the force that pulls the work outside it. He added a warning for teams that now produce faster than ever: “Going fast in the wrong direction is not progress, it's a dead end.”

Design leaders seem to agree. In the AI in Design 2026 report by Designer Fund and Foundation Capital, based on a survey of more than 900 designers, only 5 percent of leaders said they were placing less emphasis on execution quality. Many argued that taste, judgement and the ability to hold a high standard matter even more, now that good enough output is easy for anyone to generate.

References are taste you can show

Taste is hard to put into words, which is exactly why it has to be put into references. A folder of photographs a brand is proud of says more about light, crop and colour than a page of adjectives. A few pages you admire say more about spacing than any brief. The references you reject say even more.

This is how we work with images. Reel builds every style from real references, with light, lens and set chosen the way a photographer would choose them, and turns it into a recipe that is tested until the client approves it. Studio does the same for words: a brand's DNA, voice, positioning and copy rules, captured once and locked, so everything that follows sounds like the company.

Taste that lives only in one person's head leaves the company whenever that person is on holiday. Taste written down as references and rules can be applied a thousand times, by people and by machines.

The final touch

The last part of taste is the least visible. Jony Ive, Apple's former chief design officer, described it at Stripe Sessions in 2025 with the example of a cabinetmaker who finishes the back of a drawer that nobody will see. “I believe that we have this ability to sense care,” he said, and we sense carelessness just as well.

There is evidence that he is right. In a test by Ipsos and Syracuse University this year, most viewers could not tell which ads had been made entirely by AI. Yet even without knowing, they rated the human made work as more eye catching and more imaginative.

In practice, care is a list of small things. The crop that is a little tighter. The word that is almost right and gets replaced. The image that is technically fine and still goes in the bin. Machines are very good at the first draft and the second. The final touch is a decision someone has to make, and stand behind.

Where people stay

That is why we keep people at two points in every process: at the ideas, and at the picks. In Reel, ideas, references and styles are agreed before anything is made, and a person picks every image and every clip that is used. In Tempo, the month arrives prepared and tested, and a person approves every post. The machine makes the options. A person says no to most of them.

We want AI to take the making, the variants and the routine, and to make all of it faster. The ideas, the judgement and the responsibility for what goes out stay with people, because that is where a brand is decided. Reel describes its own future as a studio that learns what performs, makes more of it, and asks you only for taste. Even at the end of that line, taste is the part that stays with a person.

What a company can do

Write your taste down before you automate anything. Collect the twenty pictures, pages and posts you are proudest of, and the ten you would never publish, with one line on each that says why. Turn them into references and rules your tools can read.

Name the people who pick, and give them time to do it properly instead of squeezing approvals between meetings. And keep the bin full. A team that throws away most of what the machine makes is a team with taste.

Making will keep getting cheaper. Deciding well will not.