Where Does Judgment Come From?
The Hidden Dead End of the AI Native Era
A follow-up to “The Hollowed Center of Judgment” and a companion to “Function Sovereignty.”
There is a question that has been sitting in my mind for months now. I want to put it down here.
In February, I posted “The Hollowed Center of Judgment.” The argument was this: in domains that can be computationally closed — neither the customer-facing front nor the product-development back end, but the vast middle layer in between — what remains for humans is direction and final judgment, the two poles of the inverted-Π structure.
In such domains, the people who issue direction and pass final judgment are, today, abundant. They are the ones who have spent decades accumulating experience.
But something is already happening, most visibly among engineers on the U.S. West Coast: young people without this judgment capacity are no longer needed. Demand for inexperienced new graduates is collapsing, rapidly, in plain view of the global market. And the senior generation — the so-called gray hair — is becoming visibly stronger in the AI era.[¹]
While gray hair remains active, society holds together. They function as the last line of defense in judgment. But when the day comes that those without sufficient experience make up the majority, and those currently issuing direction and passing final judgment are gone — who will steer?
There is a deeper problem layered on top of this.
The next generation will lose the very opportunity to acquire judgment in the first place. People without judgment will sit in the seat of the responsible person. They will “approve” AI outputs, but they will possess no embodied knowledge with which to discern those outputs.
The accountability-laundering structure becomes complete: no one is actually judging, no one is actually accountable, yet society keeps deciding things.
This is the problem that will not leave my head.
Judgment Is Distilled Experience
People who can issue direction and pass critique accurately today — and there are not many of them — are almost without exception people who have spent enormous time doing the work of judgment themselves. Try something. It fails. You get stuck. Trouble shows up. You improvise. Three heads find wisdom one cannot. You let it rest, then return with fresh eyes. From these endless cycles, certain key points and pressure points emerge — every kind of work has them, in vast quantities.
And here is something often overlooked: very few real jobs can be reduced to a single objective function. Even something as simple as selling beverages through a vending machine involves multiple stakeholders — the location owner, the buyer, the operator — whose values must be balanced. Value judgment in real work is fundamentally human. A project can be a numerical success and still die because someone, somewhere, decided it failed on terms that no metric captured.
Judgment is shaped by friction with the raw — the pressures, the troubles, the rub of reality. This is true even in the hardest material domains. Walking the production line, understanding the procurement flow, the inbound, the outbound, watching how people actually move through the space, optimizing all of it together — none of this is simple. Through massive system rollouts, the introduction of new machinery, the implementation of new processes, humans learn judgment.
“This won’t work.” “This is a non-starter.” Behind such judgments lies an enormous accumulation of experience. Sensing what is off, what flows smoothly — these are deeply embodied activities. Cyber-physical, in the human sense.
That accumulation is the womb of senior leaders’ judgment, direction, and critique. The capacity to judge that we place at the top of organizations is, in fact, compressed and distilled experience from below — from the raw and the operational.
The Invisible Pipeline
Society has never explicitly designed for the cultivation of judgment. And yet, generation after generation, people capable of direction and critique have been reliably reproduced.
Why?
Because judgment has been cultivated, almost automatically, as a by-product of operational work.[²] Give a person with reasonable potential enough exposure to real work, and the capacity grows.
In the act of cutting purchase orders, the sense that “something about this payment feels wrong” develops. In the act of reading contracts, the nose for “this single clause is dangerous” develops. In the act of arranging logistics, the intuition that “this arrangement will fail on the morning of the event” develops. In the act of writing code, the judgment that “taking this shortcut here will become technical debt in six months” develops.[³]
The stated purpose of these jobs is to cut purchase orders, read contracts, arrange logistics, write code. But as a by-product of doing them, the capacity to size things up — society’s quiet lifeline — has been cultivated, freely and almost automatically.
There is one more important thing to say. Most of these jobs are not ones that people originally chose because they “loved” them. Logistics, procurement, credit assessment, quality assurance, compliance, maintenance and operations — the bulk of the work that actually keeps society running consists of jobs whose existence most people are not even aware of until they are assigned to them. The depth and interest of such work reveals itself only after one engages with it. Or, before any sense of interest emerges, the friction of the work itself begins shaping judgment.
In recent years, young people are often told to “find what you love” or “pursue your passion.” There is nothing wrong with that on its own. But from the standpoint of cultivating judgment, treating “what you love” as the sole entry point is decisively insufficient. Most of the work that sustains society cannot even be loved in advance — its existence is unknown. And it is precisely the experience of being thrown into work one did not know existed, and wrestling with it, that produces transferable judgment. Those who only do what they love become well-versed in what they love. But the cross-domain nose required for direction and critique does not necessarily grow there.
This is what the apprenticeship system, on-the-job training, and indenture have been, across human history — an unconscious mechanism of intelligence transmission. No one designed it explicitly. The social system simply produced it. Work gets done, and people grow. People grow, and work gets done. The social reproduction of intelligence was embedded in this loop. And the loop operated independently of whether the individual loved the work or not.
When the Pipeline Is Cut
The transition to the hollowed-center society severs this by-product pipeline.
Those raised from the start inside AI never receive the friction of the field. They never know the limit-condition environment in which judgment is forged under pressure.
Purchase orders no longer need to be cut. Contracts no longer need to be read. Logistics is arranged by AI. The first version of the code is written by AI. These “middle steps,” once the bulk of operational work, vanish cleanly.
This is, undeniably, a major step forward in productivity. But it also means losing the very pipeline through which the capacity to size things up has been cultivated, unconsciously, for centuries.
In the original Hollowed-Center essay, I noted that humans would increasingly specialize in handling exceptions — the crisis-mode response. But in most real workplaces, exception handling is performed by seniors with better judgment. Asking young people who cannot yet handle routine work to suddenly handle exceptions is like throwing untrained recruits onto a live battlefield.
What emerges instead is a generation able to produce plausible-sounding language (with the help of machines) but unable to actually judge or actually run anything. Direction in form only. Critique in form only. Final responsibility in form only. The inside, naturally, is hollow.
The Surface Is Just the Beginning
What is happening on the West Coast — the collapsing demand for less-experienced new graduates — is merely the first surface manifestation of this structural shift. What is truly frightening is that no one has an answer for what those new graduates will become in ten, twenty years.
In a world where you don’t need to cut purchase orders, by what means do you cultivate a young person’s instinct for “something is off”? In a world where you don’t need to write the first version of code, by what means do you let a young person experience the pain of technical debt?
The pipeline, having always “just existed,” never inspired the thought of being designed. By the time anyone notices, it is already gone. This is the fate of by-products.
This is not merely a problem of individual capacity development. It is a problem of how society as a whole cultivates and maintains genuinely deep intelligence over time.
The Real Dilemma
What makes this transition truly difficult is not that anyone is acting foolishly.
Get Kazuto Ataka’s stories in your inbox
Join Medium for free to get updates from this writer.
If anything, the opposite.
Organizations and individuals who exhaustively use AI become, in the short run, stronger. Purchase order processing, contract review, code generation, marketing campaign planning — those who embed AI deeply outpace others on speed, cost, and quality. At the level of individual operations, “not using AI” is no longer a rational choice.
In other words: the disappearance of the judgment-cultivation pipeline is the aggregate consequence of locally rational behavior. This is a classic collision between local rationality and systemic rationality.
Have young people spend time cutting purchase orders. Have them write code. Have them read contracts. Have them arrange logistics. In the short term, all of this is pure inefficiency. AI does it faster, cheaper, more consistently.
But there is judgment that grows only inside that inefficiency.
So the question is not “should we use AI or not.” If we don’t, we lose. But if we use it wrong, society loses its capacity for judgment over the long term.
Here lies the deepest dilemma of the AI native era.
This is not merely a discussion about education, nor about the younger generation.
The Counterargument: Observation Without Doing
Someone will say:
“But surely the source of judgment is not necessarily doing everything yourself. The falconer does not fly. The professor of surgery is not performing every new procedure with their own hands. The great editor has not written every text they edit. So shouldn’t extensive observation, critique, and refinement of AI outputs itself constitute a new form of accumulated experience?”
I have to admit that my own judgment, such as it is, did not come purely from doing everything with my own hands. A great deal of it came from watching extensively in the field — learning, comparing, sensing what was off, gradually understanding the underlying mechanisms and dynamics. So the simplistic claim that “you must have done it all yourself or you cannot judge” is not correct.
But here is the absolutely crucial point. For observation to actually grow into discerning judgment, there is an absolute prerequisite: the observer’s body must already carry a substantial reservoir of varied, raw experience.
The reason someone with high judgment in a domain can look at another’s work and instantly sense “this is off” is that they themselves once implemented, failed, and suffered in that domain. That embodied knowledge generates the resolution at which observation operates. It is what allows another’s experience to be chewed, compared, and felt as friction inside oneself.
A specialist surgeon who can spot the critical points of a new procedure — what is excellent, what is risky, where the pressure points are — does so because in their younger years, they performed many surgeries themselves, encountered unforeseen complications, handled them, sometimes failed, sometimes endured nights when they could not save the patient. A person whose surgical experience extends only to medical school exercises cannot see what is good or what is risky in the same procedure. A skilled editor who can look at a manuscript and decisively say “this section needs to be reworked” does so because the long experience of putting text into shape, of suffering and thinking through it, has been absorbed into their body. A person without that experience cannot see the difference, no matter how much they observe.
The Hollowing of Judgment
What makes the arrival of an AI-saturated training environment truly terrifying is precisely that it can skip even this first raw experience.
A person who has never implemented anything “critiques” AI-generated code. A person who has never written a contract “approves” an AI-drafted one. A person who has never been in the field “judges” AI-generated operational arrangements. A person who has never faced a customer “signs off on” AI-proposed marketing strategy.
Formally, the critique, the approval, the judgment, the sign-off — all of it is performed. But the inside is empty. Because the bodily basis for discerning what is good from what is wrong simply does not exist.
Neither the person nor those around them can see that the judgment is hollow. The form is impeccable. The outputs are delivered. The organization appears to be running. Meetings proceed. Approval seals are stamped.
But the “judgment” being made there is not judgment. It is ratification of AI output. And the person doing the ratifying does not even realize they are merely ratifying. Because they hold no bodily measure with which to distinguish ratification from judgment.
This is the deepest problem of the hollowed-center society. It is what I previously called the accountability-cleaning structure — or perhaps more accurately, the accountability-laundering structure — completed at the generational level.[⁴]
What Can Be Done
What, then, should we do? The interventions lie across three layers.
The first is the redesign of “what humans must pass through” in education.
Shift the goal of education from output efficiency to the formation of intelligence itself. Concretely: build a social norm that deliberately separates “what AI can do for you” from “what you must struggle through yourself.” In sports, no one says “weight training is inefficient, so we’ll skip it.” As judgment’s equivalent of strength training, we must consciously preserve inefficient routes — and this must be a socially shared design choice. Not excluding AI, not delegating to AI, but deliberately preserving the processes humans must pass through. And — without making “what you love” the only entry point — preserving the experience of being placed deliberately into domains one did not even know existed.
The second is the deliberate design of judgment cultivation within organizations.
What was once obtained free of charge as a by-product of operational work must now be deliberately reconstructed in a world where AI has taken those operations. Create places where young people are made to carry real responsibility, on purpose.[⁵] Systematically train the skills of critiquing, modifying, and redesigning AI output. Articulate the judgment patterns of the gray hair generation — not as know-how but as the process of “what signals to watch for, what to doubt, what to verify.” Pass that on as transferable practice. The Data Scientists’ Society of Japan’s recently revised skills definition (v6), which I have led, frames the role as “value creation leader/designer” — this is one attempt to enter that territory.
The third is the deliberate preservation of domains AI cannot reach — the “open worlds.”
The kinds of experience that exist only in matter, body, and the singularity of time. Care for children, care for elders, craftsmanship, contact with the natural world, embodied labor in the field. These are domains AI structurally cannot enter, and at the same time, they are the last strongholds where the first raw experience of human judgment grows. The “Valley of the Wind” project we have been working on is, ultimately, contiguous with this argument. Stepping aside from the “efficient life” prescribed by urban optimization algorithms, and preserving places where one’s own body comes into contact with the world — this is not merely a regional discussion. It is a social attempt to preserve the soil in which the next generation’s judgment can grow.
What Will Remain
What will hold value to the end of the AI era is, most likely, not knowledge. It is the traces of long wrestling with reality. Without those traces, real judgment does not grow. And without judgment, we cannot decide how to use AI — or how to rewrite its optimization functions.
The move we should make now is not to sharpen our technique for using AI as it becomes more and more capable. It is to redesign — in education, in organizations, and in the very architecture of society — the mechanisms through which humans continue to wrestle with reality, even in the age of AI.
Quiet work. Tedious work. Work whose results take time.
But if we do not act here, society will, with certainty, grow thin. And for some time, no one will be able to see it growing thin.
That this is a major fork in the road — that we must be conscious of this, and begin laying the seeds for the future — this is what is being asked of us now.
ps. Japanese original version is available here:
「AI native時代の袋小路 ― 見立てる力はどこから来るのか」
https://kaz-ataka.hatenablog.com/entry/2026/05/26/232654
[¹]: In fact, looking at the dynamics of overseas AI-focused ventures: companies that hire tens of thousands of experienced specialists at premium rates to teach AI (to model direction) are growing rapidly, while in the labor market, hiring of engineers and freelancers in the 22–25 age range is collapsing. The decline in demand for younger workers without “specification, design, and direction” capacities grounded in experience is becoming visible. See: Lodefalk, M., Löthman, L., Koch, M., & Engberg, E. (2026). Same storm, different boats: Generative AI and the age gradient in hiring (Working Paper №2/2026). Örebro University School of Business. https://hdl.handle.net/10419/339320
[²]: At least among those with the perceptive capacity to become leaders. Those without it tend to learn from templated versions of leaders’ wisdom rather than from direct cultivation.
[³]: Most of these jobs are not even known to exist by the individual before they are assigned. But through the experience itself, the skill grows.
[⁴]: See my earlier essay, “The Hollowed Center of Judgment.”
[⁵]: In a conversation with Yoichi Ochiai at the start of last year, the phrase “limit-condition training with AI” (shuraba with AI) emerged. The idea is neither to strip AI away and throw young people bare-handed into the field, nor to let AI handle the field entirely, but to have them pass through real high-pressure situations together with AI. In his workshops with elementary and middle school students, Ochiai gives challenges that cannot be completed without thinking together with AI — and the children, given such challenges, become visibly more alive (“kira-kira shitekuru”). This is an important hint about the direction of judgment cultivation in the AI era.
