When Intelligence Becomes Cheap, Taste Becomes the Real Bottleneck

Hatched by john ke

Jun 08, 2026

10 min read

3 views

88%

What happens when the machine can already do the work?

For most of modern history, the hard part was doing the thing. Writing the report, making the sketch, migrating the codebase, calling the customer, launching the business, learning the skill. Effort was the bottleneck, and intelligence was rationed like a luxury good. That world is ending faster than most people realize.

The uncomfortable question is not whether machines can now produce convincing output. They already can. The real question is this: if intelligence is cheap, what becomes scarce?

The answer is not just compute. It is not just money. It is not even time, although time remains brutally finite. The new scarce resource is judgment under abundance: knowing what to ask for, what to ignore, what to multiply, and what to trust. Once machines can generate a thousand plausible options, the value shifts from production to selection, from raw effort to direction, from being the engine to being the navigator.

That shift explains why a simple prompt can produce a startlingly good line drawing of a face, why a founder can spin up revenue with almost no product, and why the next decade may reward a strange mix of ambition, taste, and psychological stability more than credentialed expertise.

When intelligence becomes abundant, the winner is not the person who can think the most. It is the person who can aim best.

The first collapse: from making things to specifying things

There is something almost surreal about asking a model to recreate a face as a notebook sketch, in blue and white ink, with the hand of the artist visible in frame, and getting back something that looks intentional rather than merely generated. That kind of output is not impressive only because it is pretty. It is impressive because it reveals a deeper shift: the artifact is no longer the bottleneck, the instruction is.

In the old world, the gap between imagination and execution was wide. You had to be able to draw, code, design, or write well enough to carry the idea across the finish line yourself. In the new world, a good enough specification can collapse that gap. The prompt becomes a kind of compressed intent, and the model becomes an execution layer.

This creates a strange inversion. The better the model gets, the less valuable generic competence becomes, and the more valuable precision of intent becomes. Anyone can say, “Make this look nice.” Far fewer people can say, “Preserve the expression exactly, change the medium to ink sketch, use notebook paper texture, keep the proportions faithful, and leave visible traces of an in-progress hand-drawn process.” That second instruction is not just a prompt. It is a demonstration of visual judgment.

This matters far beyond image generation. In software, the equivalent is not “build me an app,” but “migrate this module safely, preserve these invariants, surface these edge cases, and produce tests that reflect the actual failure modes.” In writing, it is not “write a strategy memo,” but “make the argument crisper, remove hedging, surface the hidden tradeoff, and preserve the original hierarchy of claims.” In business, it is not “find me an idea,” but “scan for distribution wedges where a small, fast bet can exploit an overlooked asymmetry.”

The machine did not eliminate craft. It moved craft upstream.

The second collapse: from effort as virtue to leverage as virtue

For a long time, the cleanest way to signal seriousness was to work harder. Long hours, heroic effort, grinding through pain, staying late, burning weekends. That ethic still matters, but it no longer guarantees leverage. If a task that used to take ten days now takes ten minutes with the right tools, then effort alone is no longer the scorekeeper.

The more important metric is shots on goal per unit time.

This is the hidden logic connecting a successful tiny app, a fast business experiment, a large code migration, and a personal health problem that has resisted conventional treatment. Each of these is a domain where the old approach was constrained by low iteration speed. You could only try so many hypotheses, so many variations, so many forks before your patience, your budget, or your calendar gave out. Now the cost of trying has fallen so dramatically that the bottleneck has shifted to courage and clarity.

A useful mental model here is to think of the new economy as a multiplication stack:

Model quality determines how much raw capability you can access.
Compute budget determines how much of that capability you can deploy.
Workflow design determines whether the capability actually compounds.
Judgment determines whether the output becomes value.
Psychological endurance determines whether you keep iterating when the first answer is wrong.

Most people optimize only the first layer, as if buying a powerful engine were enough. But the real advantage comes from tuning all five. A founder who can run multiple agents on the same problem, test several business angles in parallel, and rapidly discard dead ends can compress months of learning into days. A writer who uses the best model to challenge their assumptions, tighten their prose, and generate counterarguments is no longer merely assisted, they are amplified.

In the era of cheap intelligence, the scarce skill is not producing more. It is deciding where multiplicative effort belongs.

That is why “spend more compute” is not just a technical recommendation. It is a new managerial philosophy. If one model gives a decent answer, ask three. If one agent finds one route, send another through a different path. If a task seems too large to start, break it into pieces and let the machine help you define the pieces. The most important question has changed from “Can this be done?” to “How many high-quality attempts can I afford?”

The real moat is not intelligence. It is appetite for uncertainty.

There is a seductive misunderstanding that emerges whenever tools get smarter: people assume the winners will simply be the smartest people. Intelligence does matter, but it is not the whole story. In a high-leverage environment, raw IQ without action is just unused potential. What matters more is the appetite to act before certainty arrives.

That is why the future may reward an unusual combination: intelligence, grit, and self-belief. Intelligence helps you see the opening. Grit helps you keep going after the first failure. Self-belief keeps you from retreating every time the possibility of embarrassment appears. Together, they form a kind of entrepreneurial immune system.

Think about a person who says, “I’ll cure this disease,” or “I’ll build a billion-dollar company,” or “I’ll create a category that does not exist yet.” In the old economy, that statement often looked like delusion because the path between aspiration and outcome was too long and too expensive to traverse repeatedly. In the new economy, the same statement may be less absurd because each attempt is cheaper, each feedback loop is faster, and each failed hypothesis can teach you more before your runway expires.

This does not mean overconfidence is good. It means bounded boldness has become valuable. The best operators are not reckless. They are asymmetrical.

An asymmetrical bet has three properties:

The upside is large enough to matter.
The downside is survivable.
The learning rate is fast enough to improve the next attempt.

That is a far better definition of “rational risk” than the old obsession with minimizing pain. A safe job with no equity can feel prudent while silently capping your upside for decades. A well-structured experiment with clear downside and large upside may look messy, but it can radically increase your probability of a meaningful outcome.

The psychological twist is that the same environment that creates opportunity also creates anxiety. When there are too many possibilities, the mind can become overwhelmed by the fact that every hour spent in the wrong place is expensive. The answer is not to panic harder. It is to narrow the field with discipline. The people who win are not the ones who feel everything. They are the ones who can act without being broken by the scale of the possibilities.

The new hierarchy: taste, targeting, and resilience

If machines can generate competence on demand, what remains uniquely human?

Not just creativity in the abstract. Not just intelligence in the old sense. The edge now belongs to three things that are harder to automate than most people realize: taste, targeting, and resilience.

Taste is the ability to recognize quality before it is popular. It lets you tell the difference between a technically correct output and a strategically useful one. In image generation, taste decides whether a sketch feels alive or merely accurate. In business, taste decides whether a feature is clever or essential. In writing, taste decides whether an argument is sharp or just verbose.

Targeting is the ability to choose the right problem. This matters more than ever because a smart model can assist you with almost anything, which means the opportunity cost of choosing poorly has increased. The central question is no longer “What can I do?” but “What should I use this capability on?” A brilliant migration nobody needs is still wasted motion. A perfectly executed product nobody wants is still a dead end.

Resilience is the capacity to remain functional while the world reorders around you. As tools get more powerful, the people who use them best may also experience the most intense pressure. They see how much can now be done, how quickly others can move, and how much slack still exists in their own lives. That can generate either disciplined focus or paralyzing comparison. Resilience is what keeps ambition from curdling into anxiety.

A useful way to think about the future is this: AI compresses execution, but it expands the value of selection. When execution gets cheap, bad choices become more visible. When experiments get cheaper, taste matters more because you can afford to try more things, but you cannot afford to be directionless. When the machine can draft, code, plan, and brainstorm, the human who can pick the right direction becomes the real scarce asset.

This is why the most valuable people will not simply “know AI.” They will know how to conduct it. They will set the task, define the constraints, inspect the output, and revise the premise. They will treat the model as a force multiplier for judgment, not a replacement for it.

Key Takeaways

Move upstream from doing to specifying. The highest-value skill is increasingly the ability to define the task precisely, not just execute it well.
Measure shots on goal, not just hours worked. Use AI to increase the number of credible attempts you can make in a week, month, or quarter.
Spend compute where uncertainty is expensive. Apply multiple models, agents, or passes to problems that are important, complex, or ambiguous.
Choose asymmetric bets. Pursue opportunities with large upside, limited downside, and rapid feedback loops.
Protect your mind from opportunity anxiety. The pace of possibility can become mentally corrosive. Discipline and focus are now competitive advantages.

The deeper shift: from productivity to possibility management

The temptation in moments like this is to talk about productivity. That is too small a frame. Productivity assumes the task is already chosen and the only question is how to do it faster. The bigger change is that the number of feasible tasks has exploded, and with it, the burden of choosing among them.

That is why this era rewards a new kind of person: someone who is neither passive nor frantic, neither naive nor cynical, but capable of managing possibility. Such a person knows when to let the model do the work, when to keep digging, when to multiply attempts, and when to stop. They are comfortable letting intelligence be cheap because they do not mistake cheap intelligence for cheap judgment.

If the old economy rewarded those who could bear long periods of effort, the new one rewards those who can bear long periods of ambiguity without losing their nerve. The person who can stay calm while directing abundant tools through uncertain terrain will have an enormous edge. Not because they know everything, but because they know how to keep learning faster than others can react.

The real revolution is not that machines can now sketch a face, write a memo, or help launch a business. It is that the distance between intention and outcome has shrunk enough to make ambition itself more consequential. When that happens, the rarest resource is no longer intelligence. It is the human capacity to choose well, move boldly, and remain intact while the world gets faster.

That is the game now. Not doing more. Not thinking more. Aiming better, faster, and without breaking.