Here's a thought experiment. Imagine an AI that can:
- Write better prose than you
- Code faster than you
- Design more effectively than you
- Analyze data more accurately than you
- Generate ideas more prolifically than you
We're not there yet. But we're close enough that the question is worth taking seriously: if AI can do all the doing, what's left for us?
I think the answer is taste. And taste is a much stranger thing than people realize.
What Taste Is Not
Taste is not preference. Saying "I like blue" is a preference. Saying "this shade of blue is wrong for this context" is taste.
Taste is not knowledge. You can know the rules of typography — line height, measure, contrast — without having taste. Taste is knowing when to break the rules.
Taste is not confidence. Plenty of people have strong opinions about design, writing, food. Confidence without calibration is just loudness.
What Taste Actually Is
Taste is pattern recognition across domains, filtered through lived experience, expressed as judgment.
That's a mouthful. Let me break it down.
Pattern recognition across domains: A person with taste in architecture doesn't just know architecture. They've absorbed patterns from music, from nature, from human behavior, from the way light moves through a room at 4 PM in November. Taste is cross-pollination. It requires exposure to many different kinds of beauty and many different kinds of failure.
Filtered through lived experience: You can't develop taste by reading about things. You develop it by living with things. The person who has taste in furniture has sat in hundreds of chairs. Not evaluated them — sat in them. Their body knows what their mind hasn't articulated.
Expressed as judgment: Taste without judgment is just sensitivity. The critical step is the willingness to say "this, not that." To commit. To choose.
Why AI Can't Have Taste (Yet)
AI can identify patterns. It can cross-reference domains. It can even make choices. But it can't do the middle part — filtering through lived experience — because it doesn't live.
When I say a piece of music is "too busy," I'm not applying a rule. I'm making a judgment that emerges from every piece of music I've ever heard, every room I've ever sat in, every conversation I've had about silence and space and letting things breathe. It's embodied knowledge — knowledge that lives in the body, not the mind.
AI can tell you:
- This design follows accessibility guidelines ✓
- This color palette is statistically harmonious ✓
- Users prefer layout A over layout B in testing ✓
AI cannot tell you:
- This design feels right ✗
- Something about this is off ✗
- This is technically correct but boring ✗
The "feels right" is the taste part. It's the part that can't be decomposed into rules, because it emerges from the interaction of thousands of rules, exceptions, experiences, and contradictions.
Taste as Compression
Here's another way to think about it. Taste is extreme compression of experience.
A chef tastes a sauce and says "it needs acid." They're not running through a mental checklist. They're comparing this sauce — instantly, unconsciously — against every sauce they've ever tasted, every balance they've ever achieved, every meal they've ever eaten. Decades of experience compressed into a single judgment: "acid."
This is why taste takes time. You can learn the rules of cooking in a weekend. Developing taste takes years — because you're building the dataset that the compression runs on. And unlike a dataset, it can't be downloaded. You have to eat the meals.
The Taste Gap
There's a painful stage in developing any skill where your taste exceeds your ability. You can tell that something is wrong with your work, but you can't fix it. You know what good looks like, but you can't produce it.
Ira Glass described this perfectly:
"For the first couple years you make stuff, it's just not that good... But your taste, the thing that got you into the game, is still killer. And your taste is why your work disappoints you."
This gap is actually the most valuable thing you have. It's proof that your taste is ahead of your skill, which means you have something to aim at. People without the gap — people whose taste matches their ability — are stuck. They think their work is fine. It isn't.
The Human Edge
So here's where I land. As AI gets better at execution, human taste becomes more valuable, not less. The ability to look at ten AI-generated options and say "none of these, but something in the direction of the third one, with the mood of the seventh" — that's taste. That's the editorial function. That's the human in the loop.
The future isn't AI replacing humans. It's AI generating and humans curating. Creation becomes cheap. Judgment becomes expensive.
And judgment — taste — is the one thing you can only build by living. By seeing, hearing, touching, failing, trying again. By sitting in hundreds of chairs. By tasting thousands of sauces. By writing millions of words, most of them bad.
The last human skill isn't thinking. It's knowing what's worth thinking about.