The Problem
A friend recently told me he’s stuck in a peculiar situation: he keeps starting projects, but abandons them halfway through. It’s not a technical limitation—he has all the tools. AI generates code, designs, copy. Theoretically, he can do ten times more than before.
The problem: more output, less satisfaction.
He described the feeling: “I get halfway through, the result is mediocre, and I don’t know whether to continue.”
This raised a question: when AI drives the cost of execution to nearly zero, what becomes valuable?
Taste.
Not the vague kind of “knowing what’s good.” I’m talking about knowing what’s right—for this specific context, this purpose, this audience. That judgment is the scarce resource in the age of AI.
The Illusion: Confusing Access with Ability
AI tools create a new cognitive bias.
Before, if you wanted to build a website, you had to learn HTML, CSS, JavaScript, or pay someone who did. That friction forced you to think—do I really need this? Is it worth the investment?
Now you describe what you want, and AI generates something that runs. It looks like you “did it,” but AI did it. You just called it.
The problem: we easily confuse the two.
“I can generate ten options” doesn’t mean “I know which option is right.” “I can iterate quickly” doesn’t mean “I’m iterating in the right direction.” “I have a result” doesn’t mean “this result is what I wanted.”
The inflation of access masks the atrophy of judgment.
My friend’s abandoned projects all follow the same pattern: starting is so easy that the “thinking it through” phase gets skipped. By the time you realize the direction is wrong, you’ve already sunk costs. Abandoning then takes more courage than never starting.
What Taste Really Is
Taste isn’t “knowing what’s good.” The internet is full of “good” things—award-winning work, trending products, master cases. Spend a day on Behance or Pinterest and you’ll bookmark hundreds of “beautiful” designs.
Taste is knowing what’s right—for this context, this purpose, this audience.
This requires two things: clear self-knowledge (who I am, what I want, what I don’t want) and sufficient reference points (what I’ve seen, compared, rejected).
AI’s problem: it makes the second part so easy that the first part gets skipped.
You don’t need to build your own reference system—algorithms recommend for you. You don’t need to make hard choices—generate ten versions and pick one. You don’t even need to pay for “giving up”—just generate another.
The result: your taste muscle never gets exercised.
How to Train Taste
Since generation is now cheap, judgment must be deliberately practiced. Here are methods I use:
1. Manufacture Scarcity
Actively limit your inputs. Only save three things per week that truly move you. Let everything else go. Then write down why they moved you—not “it’s beautiful,” but specifics like “this negative space creates anxiety” or “this color palette evokes a specific era.”
Scarcity forces depth, not surface skimming. When you know “only three this week,” you become more selective.
2. Build Your Personal Canon
Find your “bible”—five works, people, or projects you return to repeatedly. They form your coordinate system.
When something new arrives, ask first: is this better than what’s in my canon? If not, pass. This filters out 90% of noise.
My canon includes: Dieter Rams’ ten principles, a productivity tool I’ve used for a decade, a friend’s blog. Every decision, I unconsciously compare against them.
3. Delay the Call
When facing a problem, try yourself first, then open AI.
Even if it’s just a rough sketch, pseudocode, or bullet points. This “clumsy attempt” is your anchor. Only then look at AI’s output—you’ll have the ability to judge: where is it better than me? Where am I more accurate than it?
Without an anchor, you’re just an AI parrot.
4. Output Drives Input
Force yourself to produce something with judgment every week—analyze why option B is better than A and C, dissect a product’s design decisions, or post-mortem a project’s failures.
Without output, taste is just consumer preference. Only through output does intuition become testable, iterable system.
Final Thought
AI won’t weaken your taste, but using AI without thinking will.
The key question: what anchors your judgment?
If the answer is “algorithmic recommendations,” you’re just an extension of others’ taste. If the answer is “my own canon,” you have a real starting point.
When generation becomes cheap, judgment becomes expensive. And expensive things are worth investing in.
Work work. ⛏️