Everyone’s arguing about who gets replaced. Engineers, writers, designers, analysts — pick your profession, someone on LinkedIn has already written the reassurance post. “5 Skills AI Can’t Replace.” “Why Humans Will Always Matter.” The takes are endless and they all sound the same.
Critical thinking. First-principles reasoning. Judgment. Creativity. Strategic vision. Communication. They dress it up in different language depending on the audience, but strip away the framing and you’re left with one thing every time.
The question was never “which jobs survive?” That framing is already wrong. Jobs are bundles of tasks, and tasks get automated piecemeal, not wholesale. The real question is simpler and harder: which skill survives?
Not which role. Not which department. Which underlying human capability remains valuable when the cost of generating competent output drops to zero?
Taste.
The ability to look at infinite options and know which one is right. To feel the difference between something that works and something that sings. That’s the moat. Not because AI can’t produce — it produces endlessly. But because production without discrimination is noise.
The commoditization wave
Code, writing, design, analysis — all of it is becoming table stakes. Not worthless. But cheap. The quality floor rises every quarter. What used to take a senior hire and a week now takes a junior and an afternoon. What took a junior and an afternoon now takes nobody and ten minutes.
This isn’t a prediction. It already happened.
In 2023, the bottleneck was production. Could you get the thing made? Could you ship it? In 2026, the bottleneck is evaluation. We are drowning in competent output. The constraint isn’t generating options — it’s knowing which options matter.
AI produces a thousand iterations in an hour. Logos, landing pages, architecture diagrams, marketing copy, database schemas. A thousand versions, all technically fine. But it cannot tell you which one to ship. It cannot feel the difference between the version that’s correct and the version that’s right. Choosing what to build, what to keep, what to kill — that’s the human job now.
Consider two software engineers. Same title, same salary, same tools, same AI access. One solves whatever tickets land in their queue. They’re fast, reliable, competent. The other spends half their time deciding which problems are worth solving in the first place. They push back on bad specs. They kill features before they ship. They say no more than they say yes.
Same inputs. Wildly different value. One is replaceable the moment a better model drops. The other isn’t. The difference is taste.
What taste actually is
Everyone talks about critical thinking, first-principles reasoning, and judgment like they’re separate skills you can tick off a list. They’re not. They’re taste wearing different hats.
Critical thinking is taste applied to truth — knowing what’s signal and what’s noise. First-principles thinking is taste applied to problems — knowing what’s actually worth solving versus what just feels urgent. Judgment is taste applied to decisions — knowing when to commit and when to walk away.
Same underlying muscle. Different domains.
Taste isn’t a soft skill. It’s the sum of everything a person brings — intelligence, domain experience, pattern recognition, emotional awareness, cultural context, years of watching what works and what doesn’t. When someone exercises taste, they deploy all of it at once, compressed into a single call. That’s why it resists automation. It’s not one capability. It’s the integrated output of a whole person.
So why call it taste and not judgment?
Because taste implies something judgment doesn’t. It’s aesthetic, not just analytical. The best decisions aren’t purely logical. The founder who kills a feature that tests well because something feels off. The designer who breaks a grid rule because it looks better. That’s not analysis. That’s taste. It carries an intuitive charge that “judgment” sanitizes away.
Here’s the structural difference. AI evaluates within constraints someone else defined. It’s the optimizer. Taste defines the constraints. Taste decides what’s worth optimizing in the first place.
The obvious objection: “But AI can judge too.” Sure. Within a given frame. It scores, ranks, filters. But it doesn’t choose the frame. Every autonomous AI loop — every agent, every pipeline — still starts with a human deciding what the loop should care about. Change the frame and the outputs change completely. The frame is the taste. And taste is still ours.
Taste in every domain
Here’s how you know if something is a taste problem or a knowledge problem. If there’s a clear right answer, it’s knowledge. If there’s no clear right answer and you still have to choose — that’s taste.
Engineering. The architect who chooses the right abstraction over the clever one. Two senior engineers with identical knowledge will design different systems. The difference is taste in structure — which trade-offs to accept, which complexity to reject.
Hiring. Reading beyond the resume. Sensing who will thrive versus who just interviews well. Every great team was built by someone with taste in people.
Strategy. Knowing which bet to make when the data supports multiple directions. Strategy without taste is spreadsheet extrapolation. The founder who says no to nine ideas and goes all-in on the tenth — that’s taste applied to markets.
Communication. Knowing what to say, what to leave unsaid, and how much context someone actually needs. The best explainers don’t have more knowledge. They have taste in attention.
Simplification. Knowing what to cut without losing the essence. “As simple as possible but no simpler” is a statement about taste. Every great product is the result of someone killing features that the roadmap said should exist.
Domain doesn’t matter. The designer who knows which trend to ignore outlasts the one who follows all of them. The writer who kills their darlings outlasts the one who publishes everything.
Taste is the universal filter.
How taste is built
People hear “taste” and think it’s innate — you either have it or you don’t. That’s wrong. Taste is learnable. But not through courses or frameworks. Nobody ever developed taste by reading a book about developing taste.
The formula is simple: high-volume judgment plus feedback loops. You need both sides. Lots of decisions AND learning whether they were good. One without the other gives you either recklessness or theory. Neither is taste.
Reps with consequences. Ship things and watch what happens. Not reading about design — designing, and then seeing people use it. Not studying marketing theory — running campaigns and watching the numbers. Taste comes from feedback loops, not textbooks.
Exposure to excellence. You can’t judge good if you’ve never seen great. Study the best writing. Use the best products. Dissect the best decisions. Taste is pattern recognition, and pattern recognition requires a library of references. Build the library deliberately.
Saying no. Taste isn’t additive, it’s subtractive. It develops by killing options, cutting features, editing ruthlessly. The person with taste isn’t the one who adds the most — they’re the one who removes everything that doesn’t belong. Every “no” sharpens the filter.
Editing other people’s work. Red-penning develops taste faster than creating from scratch. When you edit, you learn to see what’s unnecessary, what’s weak, what’s trying too hard. You develop an eye for the gap between intent and execution.
Cross-domain theft. The engineer with taste has read philosophy. The designer with taste understands unit economics. The strategist with taste has studied architecture. Taste grows at intersections, not in specializations. The best references come from outside your field.
Arguing about quality. “I just like it” isn’t taste — it’s preference. Taste is preference you can defend. The argument forces you to articulate principles, to find the reason underneath the instinct. If you can’t explain why something is better, you haven’t developed taste yet. You’ve developed a habit.
Passive curiosity is the prerequisite. Reading widely, noticing details, caring about quality — all necessary. But taste is forged in decisions. Specifically, decisions with stakes where you could be wrong. The gap between “I appreciate good work” and “I produce good work” is bridged by thousands of judgment calls with real consequences.
What this means for your career
Here’s the litmus test. Look at your current role. Does it require taste — picking what to build, what to cut, what matters? Or does it require execution — doing what someone else already decided? That ratio is your durability score.
This has nothing to do with job titles. There are senior engineers with no taste and junior designers with tons of it. The org chart doesn’t map to taste. The question is whether you’re the person who decides or the person who implements decisions. Seniority doesn’t protect you. Judgment does.
If your job is pure execution, you have a window. It’s closing, but it’s open. Start moving upstream. Make judgment calls before you’re asked to. Pick the problem before someone assigns it. Edit before there’s a review cycle. Kill an idea before it ships. Every act of taste is a signal that you’re operating at a level no model can reach.
And here’s the reframe that matters: this isn’t doom. It’s liberation. The most tedious parts of every job — the grunt work, the boilerplate, the repetitive analysis — that’s what’s getting automated. What’s left is the interesting part. The choosing. The directing. The shaping. The parts that actually require you to be you.
The jobs that survive aren’t the ones that produce the most. They’re the ones that choose the best.
AI can generate anything. Taste is how you decide what deserves to exist. Everything else is noise.