In a world where you can generate apps, images, and stories almost on demand, taste is what remains.
Not taste as in "nice colors." Taste as in judgment. The ability to look at what exists, imagine what ought to exist, and tell which next move narrows the gap.
That applies to interfaces. It applies to stories. Most importantly, it applies to problems. The best builders are not only better at solving problems. They are better at choosing them.
Taste is subjective?
Of course taste is partly subjective. One person loves brutalist architecture. Another sees a concrete apology.
Taste is also cultural. Different cultures have different values and belief systems. Those values guide which problems feel urgent, which tradeoffs seem acceptable, and what kind of interface feels respectful1.
But "subjective" does not mean "random." Taste can be trained. You can get better at seeing. You can learn why one product feels inevitable and another feels assembled from parts.
Without taste, products become merely ok. They either collapse into uniformity, the same cards, the same gradients, the same UX, or they become a random pile of clever details. Both are failures of judgment.
We should not settle for ok. The goal is to make something you are proud to put your name on. Something opinionated enough to be disliked by the right people. Something linearesque.
AI can help you move faster. It cannot decide what is worth moving toward.
Cave, Reader
We should be absurdly bullish on AI. I want it to replace as much drudgery as possible. I use it for code, summaries, language learning, exploration, and the thousand small tasks that used to leak time.
But if you want to live a tasteful life, the important decisions still have to pass through a human.
Tools can produce options. They cannot carry responsibility. They cannot know what you refuse to ship.
Honing Taste
Think of how you solve a Maths problem: You sit down and tinker. The problem is hard, so you stare at it for a while. Maybe you try three approaches and none of them work. Maybe you sneak a glance at the solution because you are tired.
Then the solution looks obvious. So obvious, in fact, that you convince yourself you would have found it.
But the next time you see a similar problem, you get stuck again. Your brain played a trick on you. It mistook recognition for understanding. You did not learn the pattern. You memorised the shape of an answer.
To actually learn Maths, you have to struggle with the problem long enough for the structure to become yours. You need to fail in five directions. You need to build the internal machinery that lets you generalise.
Taste works the same way.
To tell what is from what ought to be, you have to make things. Design ten bad interfaces. Launch five products. Write the ugly first version. Delete the clever paragraph. Notice why the thing in your head died on contact with the real world.
Books help. Theory helps. AI helps.
But they cannot do the seeing for you. Taste is learned by contact.
TL;DR
Hold the brush yourself.
Use the helpers. Use them aggressively. But do not outsource the part where you decide what good means.
Write the first version yourself. Draw the outline yourself. Sit with the pixels long enough to know what they want to become.
Slow down there, so you can speed up everywhere else.
Be tasteful.
Footnotes
-
Consider the difference between information-maximising Japanese design languages and more minimalist American ones. ↩