Taste Is Trained

Taste gets talked about like a gift: some designers have the eye, the rest of us install component libraries. It's a comforting story because it excuses not practicing. But nobody's instinct for whether an easing is right came pre-installed - the people we call tasteful have simply run more reps of noticing.

The training loop

• Consume deliberately. Emil Kowalski's starting point: you can't produce quality you can't perceive. Use good software slowly. Open the menu again. Why does this feel solid? What is that shadow doing?
• Articulate what you notice. Vague admiration trains nothing. "The exit is faster than the entrance" is a rep; "this feels nice" is not. Raphael Salaja's essay frames taste as exactly this - learned discrimination, sharpened by naming the differences.
• Recreate, then compare. Build the interaction you admired, put it side by side with the original, and study the gap. The gap is the curriculum.
• Change one variable at a time. Emil's judgement exercises are A/B comparisons: same component, 150ms vs 250ms, ease-out vs ease-in-out. Judging pairs is how you calibrate faster than judging singles.

Why it matters more now

When AI can generate a plausible interface in seconds, execution stops being the bottleneck - discrimination becomes the job. The skill of telling which of five plausible options is right is taste, and it only comes from the reps above. The tools got faster; the eye still has to be earned.