The Rise of the Taste Worker
We've entered an era where taste—the ability to discern what's valuable from what's noise—is the only human skill that scales. Welcome to the post-knowledge-worker world.
Since the dawn of the knowledge economy, we’ve told ourselves a comforting story about talent: hire for attitude, train for aptitude. Growth mindset conquers all. Work hard enough and you’ll figure it out.
That story served us well. It’s now dangerously incomplete.
At KOHO, we’ve watched AI transform what “good” looks like in ways that have forced us to completely recalibrate how we think about talent. And a recent interview with product legend Gokul Rajaram crystallized something I’ve been circling for months:
We’ve entered an era where “taste”—the ability to discern what’s valuable from what’s noise—is the only human skill that scales.
Peter Drucker coined “Knowledge Worker” in 1959 to describe the shift from manual labor to thinking for a living. We’re due for a new term.
Welcome to the age of the Taste Worker.
The Production Bottleneck is Gone
The old world had a simple constraint: production was expensive. Code took time. Design took time. Copy took time. So we optimized for throughput. We built teams, processes, and hiring rubrics around the question: how do we produce more, faster?
AI has obliterated that bottleneck.
The cost of producing stuff—code, designs, drafts, prototypes—is approaching zero. What Gokul calls the era of “AI slop” is upon us: engines running rampant, producing infinite output, most of it garbage.
The new bottleneck isn’t production. It’s synthesis. It’s the ability to filter, evaluate, and direct that infinite output toward something that actually matters.
What “Taste” Actually Means
I’ve started using the word “taste” to describe the capability that separates people who thrive in this environment from those who drown in it. It’s not about aesthetics. It’s about:
- Judgment — knowing what to build before you build it
- Evaluation — assessing non-deterministic output (AI doesn’t give you the same answer twice)
- Orchestration — the instinct to direct tools rather than just operate them
- Editing — the eye to reduce infinite options to the one that matters
Gokul frames it as the Rick Rubin approach to product: the best people aren’t producers, they’re editors. They subtract. They curate. They know when to stop.
This isn’t a skill you can teach in a two-week bootcamp. I’ve come to believe—uncomfortably—that some of this is just innate.
The Dissolution of Roles
Here’s where it gets interesting: role boundaries are dissolving.
When a PM can prototype a working feature in an afternoon, when a designer can ship production code, when an engineer can conduct user research and synthesize findings—what does “role” even mean anymore?
The answer: role is becoming less about what you do and more about what judgment you bring.
| Role | Old Value | New Value |
|---|---|---|
| PM | Writing specs | Knowing which problem is worth solving |
| Designer | Pushing pixels | Understanding why one interaction feels right |
| Engineer | Typing code | Architecting systems that survive contact with reality |
AI handles the doing. Humans provide the discernment.
This means the middle-ground role—the person who manages processes without deep domain understanding—is evaporating. The manager whose value was coordinating humans? They better learn to coordinate agents too, or become an IC again.
There’s always been a tension between the “idea people” and the “skills people.” The Jobs and Wozniak dynamic. The visionary and the builder. The old world could afford that split because execution was expensive enough to justify deep specialization.
AI collapses that divide. When execution is cheap, having vision without the skill to evaluate and direct implementation is just having opinions. And having deep technical ability without the curiosity to know what’s worth building makes you a more expensive AI agent.
The winners aren’t idea or skill. They’re the people with an innate drive for both—curiosity that pulls them toward the right problems, and the ability to acquire whatever skills are needed to recognize good solutions. That convergence is taste. And there’s less and less room for people sitting on only one side of the old divide.
Why Attitude Isn’t Enough Anymore
This is the hardest thing I’ve had to internalize, because it conflicts with values I’ve held for my entire career.
I want to believe that effort and growth mindset can close any gap. But AI has fundamentally changed the physics. It used to take a team of people to multiply someone’s judgment—coordinators, executors, specialists turning a leader’s vision into output. AI now provides that leverage directly. One person with strong judgment and an AI toolkit can do what used to require a department. One person without that judgment just produces more slop, faster. The gap isn’t linear anymore. It’s exponential.
We’re seeing this in our own data. The gap between high performers and average performers is widening—not because average is getting worse, but because the ceiling is accelerating away. AI raises the floor (it’s genuinely harder to be terrible now), but it launches the ceiling for people with taste. The distribution isn’t a bell curve anymore. It’s a power law. A small number of people with strong judgment account for a disproportionate share of meaningful output. Remove the top contributor from any team snapshot and the standard deviation collapses. That’s not a normal distribution. That’s a new reality.
This has profound implications for how we hire. Gokul’s advice: stop talking, start doing. Work projects over interviews. Artifacts over anecdotes. You’re not filtering for credentials or culture fit—you’re filtering for agency. The ability to reject a premise, navigate ambiguity, and build without a roadmap. We’re looking for what I’ve been calling “reasoned mavericks”: people who drive consensus rather than wait for it, who create the roadmap rather than follow it.
The uncomfortable hiring bar in this world is simple: if the role can be performed end-to-end by an AI agent, it’s not a role anymore—it’s a workflow. The human in the seat needs to be the person who knows which workflows to run, when to override them, and what “good” looks like on the other side. That’s taste. And it’s becoming the price of admission.