Taste

Taste looks like a feeling and behaves like a calculation. You see a layout and something is wrong before you can say what. You read a sentence and your hand reaches to cut it. You hear a plan and a small alarm goes off, even though every individual part sounds reasonable. People call this instinct, but instinct implies it came for free. It did not. Taste is compressed judgment. It is thousands of past decisions, and their consequences, folded down into a reaction fast enough to use in the room.

This is why taste feels instant and is almost impossible to fake. The speed is real, but the speed is the last thing to arrive. Underneath it is a long history of paying attention to outcomes. You shipped the clever thing and watched it confuse people. You kept the boring thing and watched it quietly work. You admired something, took it apart, and learned why it held together. Each of those is a data point. Taste is the model trained on them.

Taste is not preference

The cheap version of taste is preference: I like this, I do not like that. Preference is real but it is not load-bearing. It does not have to survive contact with anyone else's reality. Taste is the version of preference that has been corrected by consequence. It carries a memory of what happened the last time you trusted that feeling, and the time before, and the time it was wrong.

That correction is the whole point. A person with strong opinions and no feedback loop is not developing taste. They are decorating their bias and calling it conviction. The mark of real taste is not that someone holds preferences confidently. It is that their confidence is calibrated — they are quick where they have earned the right to be quick, and they slow down exactly where their history is thin.

You can hear the difference. Untrained taste explains itself in adjectives: clean, elegant, modern, off. Trained taste explains itself in consequences: this will confuse the new user, this will be expensive to change later, this asks the reader to do work the writer should have done. The second kind can be argued with, which is precisely why it is worth more.

How it is actually built

Taste is built by closing the loop between a choice and its result, on purpose, many times. Most people break that loop without noticing. They make the choice, move on, and never go back to see what it cost. The work felt finished, so the lesson was never collected. Years pass and they have experience without judgment — a long résumé of decisions they never audited.

The builder who develops taste does the unglamorous thing: they look back. They notice that the feature everyone fought for went unused. They notice that the paragraph they were proudest of was the one readers skipped. They sit with the gap between what they expected and what happened, and they let it adjust them. This is the same discipline that makes shipping valuable — reality is allowed to edit the idea, and the editor takes notes.

Attention is the other half. You cannot develop taste in a domain you refuse to look at closely. People with taste tend to be a little obsessive about the surface of their craft. They notice the kerning, the latency, the word that is doing two jobs, the seam where two systems were bolted together. Not because the details are sacred, but because the details are where the consequences hide. Simplicity is downstream of this kind of looking: you cannot remove what was not earning its place until you can see, precisely, what each thing is doing.

Taste is a form of speed

In practice, taste is an economic advantage, not an aesthetic one. The person with taste arrives at a good-enough decision in one move where others need five. They prune the option tree early. They do not waste a week building the thing that taste already flagged as wrong. They recognize the dead end from the entrance.

This compounds. Every decision you can make well and fast is attention you get to spend somewhere harder. A team with shared taste argues less about settled questions and more about real ones. They have a common sense of what good looks like, so they do not re-litigate the basics in every meeting. Taste, at the level of a group, is mostly agreement about what not to discuss.

The failure mode is taste that has hardened into rule. Taste should compress past judgment, not freeze it. When the world changes — a new medium, a new tool, a new kind of user — old taste can become confidently wrong. The strongest practitioners hold their taste lightly enough to notice when it has stopped tracking reality. They keep one channel open for the unsettling result that their instincts did not predict.

Why AI raises the price of taste

It is tempting to think that machines which can generate near-infinite options make taste obsolete. The opposite is true. When generating a candidate is nearly free, the bottleneck moves entirely to selection. The scarce skill is no longer making something passable. It is knowing which of the thousand passable things is actually right, and being able to say why with enough precision that the next iteration is better.

A model will happily produce ten designs, ten openings, ten architectures, all fluent, all plausible. Fluency is no longer evidence of quality, because fluency is what the machine is best at. What it cannot supply is the consequence-trained judgment that picks the one worth keeping and rejects the nine that would have quietly cost you later. That judgment is taste, and it has just become the rate-limiting step.

This is also the antidote to slop. Slop is output that no one with taste ever stood behind. The defense against an ocean of cheap, plausible output is not more generation. It is someone who can look at it and refuse most of it. As the cost of producing things falls toward zero, the value of being right about which things deserve to exist rises toward everything.

The honest version

Taste can be performed, and a lot of it is. There is a posture that imitates taste — confident dismissal, knowing references, the right vocabulary — without the underlying loop of choice and consequence. It survives until the work has to meet reality, and then it does not. Real taste leaves a track record. The performed kind leaves only opinions.

So the practical path is unglamorous and available to anyone. Make real choices, on real work, where the result will land on real people. Then go back and look at what happened, especially when it disappoints you. Take apart the things you admire until you understand the decisions inside them. Do this for long enough, with enough attention, and the judgment compresses. One day it will feel like instinct. It will not be. It will be everything you bothered to notice, finally fast enough to use.