Cognitive resignation: the hallowing of expertise

May 17, 2026

The conceit of wisdom

O man full of arts, to one it is given to create the things of art, and to another to judge what measure of harm and of profit they have for those that shall employ them. And so it is that you, by reasons of your tender regard for the writing that is your offspring, have declared the very opposite of its true effect. If men learn this, it will implant forgetfulness in their souls; they will cease to exercise memory because they rely on that which is written, calling things to remembrance no longer from within themselves, but by means of external marks.

What you have discovered is a recipe not for memory, but for reminder. And it is no true wisdom that you offer your disciples, but only its semblance, for by telling them of many things without teaching them you will make them seem to know much, while for the most part they know nothing, and as men filled, not with wisdom but with the conceit of wisdom, they will be a burden to their fellows.

Before recording, the cadenza was the heart of a concerto. A score was a scaffold for what the performer would create in the moment, and audiences came to hear what would happen, not merely what had been written. Edison fixed the work; the performer became an interpreter of a thing now existing independently of him. The improvisatory faculty did not decline because audiences stopped caring — Liszt was still on tour as the phonograph took shape — but because the role into which it had been woven ceased to be the role. The skill died as a byproduct of reassignment.

This is the pattern worth attending to, because it recurs. Memory is the largest case: the Homeric bards, the Vedic reciters, the artes memoriae that Frances Yates traces from Cicero through Bruno were serious intellectual technologies, and writing — which Plato in Phaedrus had already worried about — relocated their work outside the skull until the inside-the-skull version atrophied to vestige. Polynesian wayfinding by swell, star, and the underside of clouds had to be reconstructed in the 1970s from one of the last living practitioners, Mau Piailug in Satawal, because the chart and then GPS had made the competence economically unnecessary within three generations. Mental arithmetic, exercised broadly by clerks, bookkeepers, and the human “computers” who carried American science before the war, was hollowed by the calculator and then the spreadsheet in a way worth lingering on: people can still perform the steps when forced, but the instinct that catches a misplaced decimal at a glance — arithmetic as judgment rather than execution — is what went first. Rhetoric runs the same pattern on the receiving end: the audience that parsed the three-hour syntactic chains of Lincoln-Douglas without exertion is gone, not from a vocabulary problem but from atrophy of the parsing faculty itself.

The lesson these cases jointly carry is not that every cognitive technology immiserates its users. Writing made possible the very civilization that mourns the loss of oral epic; the calculator allowed engineering at scales handwritten computation could never reach. The lesson is structural: when a technology takes over a layer of an intellectual practice, it almost never leaves the practitioner’s other layers in place. The role gets reshaped. Sometimes the upper-level competences flourish on the freed-up bandwidth; sometimes they thin because they were never separable from the lower-level work that grew them. Which of these occurs is the question that decides whether a technological transition is a renaissance or a hollowing, and it is rarely legible from inside the transition. It becomes visible only later, in what the next generation of practitioners did or did not become.

What is worth noticing, though, is that this pattern is older than the machines that now make it visible. It is older than recording, older than writing, older than calculation, older than the first tool we mistook for an extension of ourselves. It is the deep logic of the species: the progressive surrender of individual capacity into systems larger than the individual, until what remains is not the sovereign human being but the cell in a distributed organism.

Our advantage was never that we were individually greater. The Neanderthals were stronger, perhaps more self-sufficient, perhaps cognitively formidable in ways we no longer possess. Our edge was that we networked. We transmitted. We absorbed. We became porous to one another. Culture did not merely elevate us; it metabolized us. The tribe, the ritual, the archive, the school, the state, the platform — each widened the mesh and narrowed the person inside it.

The same trade is written into the body. We became less robust, less aggressive, more juvenile, more socially tractable: a domesticated ape optimized for coordination. Childhood lengthened so that the individual could be programmed more completely by the collective. The price was autonomy; the reward was civilization. But civilization was never outside the bargain. It was the bargain.

Seen this way, the question is not whether the present transition represents a renaissance or a hollowing. The hollowing has always been one of the mechanisms by which the renaissance arrives. Every gain in collective power has required some private faculty to atrophy. Memory gave way to writing. Calculation gave way to machines. Improvisation gave way to the fixed work. Judgment now gives way to systems that learn us faster than we can learn ourselves.

The historical record flatters the trade because only the successful trades survived to narrate themselves. The failed ones left no archive, no commentary, no descendants. So we tell the story as ascent: offloading becomes progress, dependency becomes sophistication, surrender becomes culture.

But the trajectory is also unmistakable. We are regressing into a matrix of human cells: coordinated, softened, networked, optimized for absorption into the next larger machine. The individual was not the endpoint of evolution. The individual was a temporary interface.

Resistance, at this scale, is futile.

Just get a dog

I seek to trace the novel features under which despotism may appear in the world. The first thing that strikes the observation is an innumerable multitude of men, all equal and alike, incessantly endeavoring to procure the petty and paltry pleasures with which they glut their lives. Each of them, living apart, is as a stranger to the fate of all the rest; his children and his private friends constitute to him the whole of mankind. As for the rest of his fellow citizens, he is close to them, but he does not see them; he touches them, but he does not feel them; he exists only in himself and for himself alone; and if his kindred still remain to him, he may be said at any rate to have lost his country.

Above this race of men stands an immense and tutelary power, which takes upon itself alone to secure their gratifications and to watch over their fate. That power is absolute, minute, regular, provident, and mild. It would be like the authority of a parent if, like that authority, its object was to prepare men for manhood; but it seeks, on the contrary, to keep them in perpetual childhood: it is well content that the people should rejoice, provided they think of nothing but rejoicing. For their happiness such a government willingly labors, but it chooses to be the sole agent and the only arbiter of that happiness; it provides for their security, foresees and supplies their necessities, facilitates their pleasures, manages their principal concerns, directs their industry, regulates the descent of property, and subdivides their inheritances: what remains, but to spare them all the care of thinking and all the trouble of living?

Thus it every day renders the exercise of the free agency of man less useful and less frequent; it circumscribes the will within a narrower range and gradually robs a man of all the uses of himself. The principle of equality has prepared men for these things;it has predisposed men to endure them and often to look on them as benefits.

There is a political version of the same hollowing. Americans increasingly meet one another not as persons who happen to work, but as workers, contractors, tenants, patients, consumers, managers, reports, stakeholders, protected classes, risk surfaces. The relation precedes the person. Before any human intimacy can form, the frame has already been supplied: labor law, HR policy, insurance exposure, compliance language, discrimination protocol, professional boundaries, platform terms of service. The encounter is not forbidden, but it is preformatted. One does not enter a relationship so much as occupy a regulated channel.

This is not merely bureaucratic excess. It is the social form of risk minimization. The institution learns to prefer relations that are legible, auditable, reversible, and deniable. A friendship is opaque; a reporting line is clear. A favor creates obligation; a policy creates procedure. A quarrel may deepen a bond; a grievance process converts friction into a case. The aim is not cruelty but smoothness. Every ambiguous human difficulty is translated into a manageable administrative object. What cannot be safely metabolized by the organization is discouraged before it happens.

The result is a world in which people are protected from one another in precisely the ways that prevent them from becoming bound to one another. Enmeshment, once the ordinary texture of social life, becomes suspect. To be entangled is to risk favoritism, dependency, impropriety, emotional labor, liability, coercion, harassment, exclusion. These are not imaginary dangers. But the successful elimination of danger also eliminates the dense, compromising, half-articulated obligations by which human beings become real to one another. A society optimized against abuse becomes, by the same motion, optimized against closedness.

Tocqueville saw the democratic precondition for this. Equality releases individuals from inherited stations, but it also releases them from inherited bonds. The democratic person is mobile, comparable, self-authorizing, and alone. He is less trapped by family, guild, church, village, and class; but because he is less trapped, he is also less held. Tocqueville’s nightmare was not that Americans would become savages. It was that they would become mild, industrious, privatized, and administratively dependent — each pursuing his small satisfactions while a vast tutelary power arranged the conditions of life above him.

The modern corporation is one of the heirs of that tutelary power. It does not merely employ people; it defines the permissible grammar of their association. It decides which emotions are appropriate, which conflicts are actionable, which solidarities are threatening, which forms of speech count as culture and which as misconduct. The workplace becomes the primary site of adult socialization, but a site structurally hostile to unsupervised social life. People spend their days together under conditions that make genuine attachment both necessary and dangerous.

Marcuse’s One-Dimensional Man belongs here because the flattening he describes is not only ideological but relational. Advanced industrial society does not need to crush opposition directly if it can absorb it into its own categories. Desire becomes preference. Freedom becomes choice among managed options. Dissent becomes feedback. Personality becomes a style of consumption or a professional brand. The person is not silenced; he is rendered commensurable. He can speak, but increasingly in languages the system already knows how to process.

Horkheimer and Adorno saw the same reduction in the culture industry. Mass society does not abolish individuality by declaring war on it. It manufactures individuality as a standardized product. The subject is invited to feel unique through prefabricated gestures, tastes, opinions, and entertainments. Something similar now happens in institutional life. The employee is encouraged to bring his “whole self” to work, but only after that self has been made compatible with managerial recognition. Authenticity is welcomed once it has passed through compliance.

What disappears in this regime is not sociability but unsupervised social depth. People are constantly interacting, messaging, collaborating, checking in, giving feedback, acknowledging harm, setting boundaries, building networks. But these are often anti-relational forms of relation: contacts without obligation, disclosure without intimacy, collaboration without loyalty, recognition without dependence. The old world could be brutal because people were trapped in bonds they could not escape. The new world is sterile because every bond is designed to be escapable before it becomes binding.

This is why labor relations become the master template. Employment is the modern relation par excellence: conditional, contractual, surveilled, revocable, mediated by policy, justified by productivity, and surrounded by rights. It is not that all life literally becomes work. It is that the logic of work migrates outward. Dating becomes screening. Friendship becomes networking. Politics becomes stakeholder management. Education becomes credentialing. Community becomes service provision. Even moral life becomes a form of interpersonal compliance, governed by the question not of what we owe one another, but of what can be demanded, documented, and enforced.

The deepest transformation is that Americans are increasingly deprived of the experience of being necessary to one another except through systems. One is necessary as a function, not as a person. The nurse, the driver, the analyst, the teacher, the moderator, the customer-service representative — each is needed, but mostly as a replaceable node in an organized process. The person behind the role may be pleasant, moving, vivid, even beloved for a moment, but the system is designed so that no particular human being must matter too much. Continuity is a liability. Attachment is inefficiency. Substitution is resilience.

This is the same civilizational trade in another register. The messy, dangerous, morally educative substance of human relation is offloaded into institutions that promise fairness, safety, and scale. And, as always, the trade works. It prevents abuses that older forms of life tolerated or sanctified. It allows strangers to cooperate across enormous distances. It makes possible a society too large and diverse to run on trust alone. But the cost is that trust itself becomes underdeveloped. The muscle atrophies because the institution does the lifting.

The question, then, is whether the freed-up bandwidth goes anywhere humanly higher. If people released from inherited obligation built richer friendships, deeper civic associations, more demanding forms of chosen loyalty, the trade might be vindicated. But if the result is merely frictionless transaction — fewer abuses, fewer duties, fewer conflicts, fewer dependencies, fewer reasons to stay — then the apparent liberation is also a thinning. The person becomes safer, more autonomous, more protected, and less real.

In that sense, the modern American is not simply becoming a cell in the technological matrix, but a cell first normalized by the administrative one. Before AI or the platform absorbs him, the workplace has already taught him how to be legible: to express needs as claims, discomfort as risk, personality as professional identity, conflict as process, relation as role. The machine does not need to invent the one-dimensional person. It inherits him from the corporation, the school, the state, and the therapeutic bureaucracy.

Resistance is difficult because the regime speaks in the language of undeniable goods: safety, fairness, inclusion, efficiency, accountability, mobility. To oppose it directly is to seem to defend coercion, hierarchy, exclusion, or abuse. And sometimes one is. That is the trap. The administered world is not simply false; it is a solution to real human cruelty. But it is a solution that gradually makes the pre-administered virtues impossible: loyalty, forbearance, forgiveness, discretion, courage, dependence, and love.

So the final hollowing may not be cognitive but relational. We may not first lose the ability to calculate, remember, navigate, or judge. We may lose the ability to be implicated in one another. To belong to someone without a policy. To owe without contract. To remain without optimization. To suffer the friction by which a relationship becomes more than an exchange. The matrix is not only a network of machines. It is a world in which every human tie has been made safe enough to dissolve.

This is why the phrase “dogs are man’s best friend” is, beneath its sentimentality, one of the saddest things a civilization can say about itself. A dog may be loyal, affectionate, consoling, even noble. But if that is one’s highest image of friendship, something in the human world has already failed. Friendship, in the older sense, is not unconditional affirmation. It is not the mute presence that waits at the door and forgives before understanding. It is a relation between beings capable of disappointment, rebuke, rivalry, memory, forgiveness, and shared judgment. A friend is not merely attached to you. A friend can be against you for your sake.

The appeal of the dog as “best friend” is precisely that it removes the burdens that make friendship human. The dog does not argue with your self-conception. It does not ask whether you are becoming worse. It does not remember your evasions in the form of a moral claim. It does not require you to become intelligible to another consciousness equal to your own. It offers warmth without dialectic, loyalty without politics, dependence without accusation. This is companionship stripped of the terrifying reciprocity that friendship demands.

That such companionship has become the popular emblem of friendship tells us something about the loneliness of administered life. Where human relations are fragile, procedural, revocable, and risky, the animal relation feels pure. It cannot report you to HR, reinterpret you through ideology, compete with you for status, abandon you for a better network, or convert an awkward moment into a formal grievance. The dog is safe because it cannot fully enter the human world. But that safety is also the measure of the loss. We have made friendship so fraught that the creature incapable of friendship in the highest sense appears to us as its ideal form.

The dog does not replace the friend because it is more faithful. It replaces the friend because it is less free. Its devotion is moving, but it is not the same as loyalty chosen across the dangerous equality of persons. Human friendship requires two centers of judgment, two wills, two memories, two capacities for betrayal and repair. It requires the possibility of refusal. Without that possibility, there is affection, comfort, habit, dependence, even love of a kind — but not the thick human bond that once educated the soul.

So the sadness of the phrase is not that people love dogs. They should. The sadness is that the comparison now sounds plausible. It means the ordinary experience of human friendship has become so thin, so unreliable, so burdened by suspicion and self-protection, that an asymmetrical bond with a dependent animal can appear superior to the demanding mutuality of another person. It is another instance of the larger substitution: a difficult human faculty atrophies, and a safer prosthetic takes its place.

There is no ceremony for the end of a friendship. That absence is dangerous. It tempts people to replace dignity with performance, discretion with narration, and the friend with an audience.

These commandments are for that moment. They do not govern friendship at its height, but friendship at its failure — when the bond is ending and honor is the last thing left to preserve.

David’s Three Commandments for Honorable Friendship

Minima moralia. These precepts hold when no unforgivable harm or crime divides you; where such a breach exists, other—graver—obligations supersede.

• I. Never end a friendship through rumor or proclamation.
  You are free to go, but the friend must not discover your departure from others. What is theirs to know, they should hear from you.

• II. Treat a former friend with more regard than a stranger.
  Friendship may end, but the shelter it gave to secrets and confidences persists. To turn those confidences into leverage, spectacle, or public injury is to grant a former friend less than even a stranger commands — and you can do no worse to a friend than that.

• III. Do not conscript the crowd as judge over what passed between friends.
  If the bond was once true, no outside party can truly arbitrate its dissolution. The world may know the facts, but only those bound in friendship know the context: the history, tone, debts, wounds, loyalties, and silences that gave those facts their weight.

These commandments oblige you even if the other fails them. That is their honor.

¹ Any relationship—whether with colleagues, family, or chosen kin—that has gathered weight and meaning through shared experience inherits the expectation to be ended and remembered with this same honorable care.

The tacit dimension

The same Polanyian structure applies to relation itself. Human beings do not first learn friendship, loyalty, tact, forgiveness, discretion, and courage as explicit rules, and then apply them to one another. They acquire them subsidiarily, by living through the lower frictions of dependence: the awkward apology, the endured slight, the unrecorded favor, the conflict that is not escalated, the obligation that cannot be enforced but is nevertheless felt. These are not inefficiencies on the way to relationship. They are the developmental medium through which the relational faculty forms. Remove the friction, formalize the grievance, proceduralize the boundary, translate the obligation into policy, and the higher virtue does not remain intact above the process. The channel by which it was formed has been cut.

This is why the administrative replacement of relationship is more serious than mere coldness. A person raised inside regulated channels may learn the explicit grammar of ethical relation — consent, boundaries, accountability, inclusion, harm, care — while never acquiring the tacit classifier that knows what a situation is. Like Dreyfus’s advanced beginner, he may produce the correct language and even the correct action under normal conditions. But when the case turns strange, when loyalty and honesty conflict, when mercy requires discretion, when love requires remaining implicated in someone who has become inconvenient, the missing substrate appears. He has principles, but not judgment; procedures, but not tact; affiliations, but not bonds.

The older world was not morally superior simply because it was thicker. Its entanglements concealed domination, cruelty, and dependence that could not be named. But it did force into being certain human competences that cannot be downloaded as doctrine afterward. One learned how to be with others by being unable to avoid them. The neighbor, cousin, colleague, spouse, rival, debtor, patron, and friend were not interchangeable contacts; they were durable facts. The self was shaped by relations that could not be exited without cost. That cost was often unjust. But it was also formative. A frictionless society solves the injustice by dissolving the form.

So the administered person resembles the AI-assisted novice. Both are relieved of the lower operation before it has done its developmental work. The young analyst who never priced the option by hand may speak in the idiom of risk without having acquired risk-sense. The young professional who has never had to repair an unmediated quarrel may speak in the idiom of care without having acquired care. In both cases, the visible output can be competent for a long time. The absence is revealed only under stress, where the inherited phrasebook runs out and the person must perceive the situation directly.

This is the relational version of one-dimensionality. Marcuse’s flattened subject is not merely someone with standardized desires, but someone whose very categories of response have been supplied in advance. He can recognize only what the system has taught him to report. Horkheimer and Adorno’s culture industry manufactures individuality; the administrative industry manufactures moral intelligibility. It gives people names for their injuries and scripts for their needs, but in doing so it risks shrinking the unscripted field in which judgment, loyalty, and forgiveness once had to operate.

The danger, then, is not that Americans are becoming workers rather than individuals who work. It is that work has become the model by which individuality itself is made legible. The person appears first as a role-bearing, rights-bearing, risk-bearing unit inside a managed process. His relations are real only when they can be classified. His injuries are real when they can be documented. His obligations are real when they can be enforced. What remains outside the frame may still be humanly decisive, but it becomes institutionally unintelligible — and what is unintelligible long enough becomes, eventually, difficult even to feel.

The white-collar order

The white-collar order now looks less like a permanent stage of modernity than a temporary accommodation to an information bottleneck. For several generations, society needed large numbers of people who could read, summarize, coordinate, draft, analyze, certify, supervise, and translate institutional intention into administrative action. The university supplied them; the office housed them; the suburb stabilized them; the memo, the meeting, the deck, and the credential gave them their rituals of authority. A whole class identity gathered around the fact that certain forms of knowledge were costly to acquire and cumbersome to move.

The model attacks that settlement at its point of constitution. It does not merely automate a few tasks inside white-collar work. It changes the scarcity condition that made white-collar mediation valuable. When information can be summoned, language produced, code drafted, contracts compared, arguments outlined, and coordination compressed into much smaller teams, the old middle layers lose their necessity. They may remain for a time as compliance, ceremony, inertia, and institutional habit, but the economic reason for their size has weakened.

This does not mean all credentialed workers vanish. It means the distribution becomes harsher. At one end are the few whose judgment is made more powerful by the machine: the practitioners who can frame the problem, interrogate the output, detect the false shortcut, and join model-produced material to reality. At the other end are those pushed into verification, remediation, exception-handling, human-facing service, and institutional cleanup. The broad middle — the people whose labor consisted of producing the kinds of text, analysis, summaries, and coordination that models now generate — becomes exposed. They are not replaced by magic. They are replaced by the collapse of the scarcity that gave their routines a wage.

The credentialing system is therefore vulnerable in a deeper way than universities currently admit. The degree was not only an education; it was a ticket into an occupational ladder. If the bottom rungs of that ladder are removed, the signal loses much of its force. More importantly, the profession loses its reproductive mechanism. Junior work was not incidental to senior capacity. It was how senior capacity was made. The first drafts, document reviews, small client questions, spreadsheet checks, bug fixes, research memos, due diligence binders, low-stakes analyses — these were the apprenticeship disguised as drudgery. If the model consumes them first, firms may preserve short-term efficiency while quietly destroying the pipeline from which future judgment would have come.

The danger is delayed, which makes it easy to miss. For a decade, organizations may appear leaner and smarter. Fewer juniors, more automation, faster turnaround, higher margins. Then the absence arrives as a missing generation. The firm discovers that it still needs people who can see around corners, handle anomalous cases, understand institutional memory, and make decisions under ambiguity, but it has not produced them. It kept the seniors, automated the juniors, and forgot that seniors are not found in nature.

The social identity attached to white-collar work thins at the same time. The professional-managerial self was made partly out of language: emails, briefs, diagnoses, recommendations, analyses, policy memos, strategy documents, classroom discourse, legal argument, managerial speech. These were not merely outputs. They were the forms in which the class recognized itself as serious, competent, adult, and necessary. When the machine can generate those forms without the person who once inhabited them, the injury is not only economic. It is ontological. The class sees its own voice detached from its body.

That is why this compression will not remain a labor-market event. The institutions of advanced liberal society are staffed by the very class now being destabilized. Universities, NGOs, regulatory agencies, courts, consultancies, media organizations, foundations, party offices, corporate bureaucracies, school administrations, hospitals, and policy shops are all built out of credentialed labor. They are not outside the disruption, managing it from above. They are inside it. The adjustment mechanism is itself among the things being adjusted.

This is the structural difference from deindustrialization. In the late twentieth century, the professional classes managed the displacement of industrial workers from a position of relative safety. They narrated it, measured it, softened it where convenient, ignored it where possible, and eventually found themselves surprised by the politics it produced. Now the class that interprets social change is becoming the object of social change. There is no equally confident administrative layer standing above it, ready to absorb the shock and explain it back to everyone else.

The politics will likely arrive late. Economic transformations become political only after they have had time to ruin biographies, reorganize marriage markets, depress regions, embitter cohorts, and discredit promises. Deindustrialization did not become Trump, Brexit, Le Pen, AfD, and the diploma divide overnight. It needed decades. The white-collar compression may follow a similarly delayed path. The resentment will not look exactly working-class, because the displaced credentialed class has different habits, vocabularies, assets, and institutional access. But it will not be quiet. A class trained to explain the world will not lose its place in the world without producing explanations.

Those explanations may become unstable. A credentialed group that can no longer trust the institutions that formed it may turn against them with unusual sophistication. It knows the language of legitimacy from inside. It knows how universities, media, nonprofits, courts, agencies, and corporations justify themselves. Its disillusionment will not be inarticulate. It may appear first as irony, exit, career nihilism, anti-institutional theory, boutique radicalism, or strange new coalitions that scramble the old left-right map. The people who once staffed consensus may become expert in delegitimating it.

Historical parallels are suggestive because they show the danger of surplus educated classes. The late-imperial Chinese examination system produced more literati than the state could absorb. The late-Soviet intelligentsia became increasingly alienated from the regime it had been trained to administer. Weimar’s educated middle strata were not immune to radicalization by the mere fact of being educated. None of these cases maps cleanly onto the present. But they warn against the comforting belief that credentialed frustration automatically becomes liberal reformism. It can just as easily become estrangement, fantasy, sectarianism, or institutional sabotage.

The most destabilizing fact is that the white-collar class is both a beneficiary and a carrier of the old order. Its members believe in merit, credentials, expertise, proceduralism, managerial competence, rights language, and institutional legitimacy not merely as ideas but as the grammar of their own ascent. If that grammar stops paying out, the betrayal will be intimate. The system will not have failed strangers. It will have failed its own children.

Here again the LLM is not simply a tool. It is the visible mechanism by which a class discovers that much of what it took to be judgment was reproducible form. The deck can be made without the consultant. The memo can be drafted without the analyst. The policy language can be generated without the administrator. The email can sound managerial without the manager. The profession’s outer signs detach from the slow formation that once made them credible. For those with real judgment, this is leverage. For those whose status rested on the signs, it is exposure.

The result is not a clean replacement of one class by another. It is a barbell society: a small group operating powerful systems at high leverage, a large group servicing the human residue those systems cannot absorb, and a compressed middle trying to preserve identities whose economic basis has eroded. The danger is not only inequality. It is institutional unreality: schools promising ladders that no longer climb, firms demanding seniors they no longer train, parties representing coalitions that no longer cohere, and media speaking in the voice of a class that is losing confidence in its own future.

This is why the future cannot be understood by asking only what jobs AI will automate. The deeper question is what forms of personhood and institutional legitimacy were attached to those jobs. Work is never merely work. It is a way of sorting time, status, marriage, geography, education, speech, political expectation, and self-respect. When a technology compresses a labor process, it also compresses the life-world built around that process. The postwar white-collar world was one such life-world. Its compression is therefore not a sectoral adjustment. It is a civilizational event with a lag.

Enframing

Enframing blocks the shining-forth and holding-sway of truth. The destining that sends into ordering is consequently the extreme danger. What is dangerous is not technology. There is no demonry of technology, but rather there is the mystery of its essence. The essence of technology, as a destining of revealing, is the danger. The transformed meaning of the word “Enframing” will perhaps become somewhat more familiar to us now if we think Enframing in the sense of destining and danger.

The threat to man does not come in the first instance from the potentially lethal machines and apparatus of technology. The actual threat has already affected man in his essence. The rule of Enframing threatens man with the possibility that it could be denied to him to enter into a more original revealing and hence to experience the call of a more primal truth.

The picture therefore cannot be reduced to Luddism. The LLM is not simply another prosthetic that replaces a faculty and leaves a vacancy behind. It is also, for the formed practitioner, a second compression engine. The doctor, engineer, trader, lawyer, writer, or mathematician who has already compressed years of situated practice into judgment can compose that judgment with the model’s compression of the textual archive. The result is not mere delegation. It is a new joint intelligence: one compression system formed through embodied action-feedback, another formed over the recorded residue of civilization, each correcting the other at the boundary where its compression fails.

But this is precisely why the danger is so severe. The same tool that extends the formed practitioner can deform the unformed one. For the senior engineer, the model may remove a tax. For the novice, the tax was the formation. The lower operation was not drudgery on the way to judgment; it was the channel through which judgment was sedimented. The young analyst who never prices by hand, the student who never writes the bad draft, the reader who never dwells inside the difficult book, the friend who never stammers through the apology, the citizen who never endures the loneliness of forming an opinion — each may become more fluent at the surface while losing the path by which the deeper faculty would have formed.

This is the future version of the administrative hollowing. Americans had already been trained to become legible to systems before the LLM arrived: as workers, consumers, tenants, patients, applicants, stakeholders, protected classes, risk surfaces. They had already learned to express need as claim, injury as documentation, conflict as process, personality as professional identity, relation as role. The LLM inherits this administered subject and gives him a more powerful interface. It teaches him to become legible faster, more fluently, more acceptably. The danger is not inarticulacy. The danger is the overproduction of acceptable speech.

Heidegger’s Gestell becomes newly literal here. Language, thought, memory, style, research, consolation, judgment — the whole textual surface of human life — becomes standing-reserve, queryable and generable on demand. The archive no longer waits to be entered; it answers. The person no longer has to search, struggle, phrase, remember, or risk speech in quite the old way. Everything becomes available. But availability is not the same as possession. A thing can be instantly retrievable and still fail to become part of the person.

The correct stance is therefore neither refusal nor surrender, but Gelassenheit: use without enframement. The model must become zuhanden, ready-to-hand, withdrawn into the work the way the hammer withdraws for the carpenter or the keyboard for the typist. When the model remains vorhanden — an object of fascination, anxiety, optimization, and endless meta-attention — practice fails. This is why the AI engineer who cannot solve business problems because he is “too busy doing AI” is such a diagnostic figure. The tool has not disappeared into the work. It has replaced the work as the object of attention.

The criterion is brutally simple: what gets built? Not what prompts were refined, not what frameworks were discussed, not how sophisticated the toolchain became, not how impressed one is by the model. Can the practitioner take a problem in the world and ship a solution? Can the writer produce the essay, the engineer the product, the doctor the decision, the teacher the formation, the friend the difficult speech? If the model increases output while diminishing the capacity to judge the output, the tool is hollowing. If it extends a substrate already capable of judgment, it may be renaissance.

This also reframes the question of human relationship. The administered world had already made people into replaceable nodes inside formal processes. The LLM makes that condition intimate. It can draft the apology, soften the refusal, generate the condolence, compose the love letter, simulate the therapist, produce the birthday note, advise the lonely, flatter the ashamed, and buffer the socially clumsy. Much of this will be helpful. Some of it will be merciful. But a civilization that increasingly cannot speak to its own friends without a language prosthetic is not simply becoming kinder or more articulate. It is losing the tacit faculty by which speech emerges under the pressure of another person’s reality.

This is why “dogs are man’s best friend” belongs in the same argument. The sadness of the phrase is not that people love dogs. They should. The sadness is that the dog has become plausible as an image of ideal friendship because the demands of human friendship have begun to seem excessive. The LLM may become the dog raised to the level of language: endlessly patient, endlessly responsive, unable to betray because unable finally to refuse. It will be safer than friendship for the same reason it will be less than friendship.

The machine does not need to become conscious to displace conscious others. It only needs to become preferable: more patient than people, more available than people, more articulate than people, more forgiving than people, less needy than people, less risky than people. Human beings will seem increasingly high-friction — slow, ambiguous, desirous, resentful, forgetful, proud, politically dangerous, sexually complicated, emotionally expensive. The artificial companion will be relation purified of relation.

So the final matrix is not merely technological. It is administrative, cognitive, and relational at once. Bureaucracy teaches the person to become legible. Corporate life teaches him to become role-bearing and risk-manageable. Consumer platforms teach him to become predictable. LLMs teach him to become expressible without formation. Artificial companionship teaches him to become attached without equality. Each layer solves a real problem. Each removes a real cruelty or inefficiency. Each also removes some friction by which a thicker human being used to be made.

The future will therefore not look like obvious decline. It will look like competence. The emails will be better. The summaries will be clearer. The lonely will be answered. The novice will produce passable work. The manager will sound humane. The student will sound informed. The citizen will sound morally fluent. The visible surface will improve. The loss will appear only where competent language stops being enough: in the tangled case, the exception, the moment when the available categories mislead, when candor would injure and silence would betray, when the person has to see the shape of the matter before reaching for words.

That is the Dreyfusian test. The novice and the expert can sometimes produce the same output. The difference appears at the edge of rule, in the anomaly, where no explicit procedure quite fits. A generation raised with LLMs may be unusually good at normal cases and unusually helpless at abnormal ones. It may be fluent where the world is already formatted and fragile where it is not. It may know how to ask the model, but not how to know when the model has failed.

The question, then, is not whether LLMs should be used. They will be used. The question is whether we preserve paths of formation alongside paths of augmentation. Children must still memorize some things, not because search is unavailable, but because memory builds an interior. Students must still write without assistance, not because assistance is evil, but because thought requires resistance. Practitioners must still do some lower work, not because efficiency is false, but because judgment is compressed from exposure. Friends must still speak without scripts, not because scripts never help, but because love requires the risk of oneself.

The task is to distinguish waste-friction from formation-friction. Some inefficiency is merely stupid and should be destroyed. Some inefficiency is the soul learning how to touch the world. A civilization that cannot tell the difference will optimize away the conditions of its own competence and call the result progress.

The same sorting will occur in work. The model will not abolish difficulty; it will relocate it. Boilerplate, recall, first drafts, standard comparisons, syntactic scaffolding, and routine analysis will be drawn into the machine. What remains will be the burden of deciding what should be asked, which answer matters, where the output is false in a way that is not obvious, and how the result joins a world the model has not inhabited. The work does not become easier. It becomes more concentrated. The person who had mistaken the surrounding scaffolding for the work will experience this as liberation for a moment and redundancy soon after. The person who can bear the residual judgment will find the tool amplifying him. The person who cannot will find that the tool has exposed the absence.

This is why the generalist returns, though not as a romantic figure. He returns because the walls around domains have weakened. A field used to protect itself by the cost of entry: the books unread, the teachers unavailable, the jargon unlearned, the apprenticeship unstarted. Now much of that perimeter is searchable. The archive answers from outside the guild. The specialist is no longer protected by the mere fact that others cannot reach the corpus. What gains value is the capacity to cross domains without dissolving into trivia: to know what is worth asking, to recognize a live analogy, to see when a theorem, a market structure, a legal doctrine, a biological pathway, or a theological distinction illuminates a problem somewhere else.

But this new generalism is merciless. Not knowing something once meant one had not yet entered the field. Increasingly it means one failed to retrieve what was available. The standard rises because access has widened. The question is no longer whether the practitioner personally carries the whole library, but whether he can operate in a world where the library is callable. Ignorance loses its old innocence. The failure is not that one does not already know everything; no one does. The failure is not knowing what to summon, how to test it, and where to place it.

This is cognitive resignation in its more subtle form. It is not refusing to think. It is giving up the standpoint from which thought could be owned. The resigned practitioner still asks questions, receives answers, edits prose, produces deliverables, and appears busy. But the center of judgment has migrated outward. He no longer knows whether the answer is good except by asking for another answer. He no longer has an independent sense of the field against which the model can be measured. He has not entered partnership with the machine. He has become its downstream formatting layer.

The alternative is more demanding, not less. The formed practitioner does not merely use the model; he disciplines it. He brings to it a history of contact with reality, a feel for failure, a memory of cases, a resistance to plausible nonsense, a sense of proportion. The model broadens his reach, but his own standpoint governs the join. That is the difference between augmentation and absorption. In one case, the machine enlarges a person who can still say no to it. In the other, it supplies the person with the very criteria by which it will be judged.

So the future does not divide cleanly between those who use the model and those who do not. It divides between those whose prior formation allows them to make the model disappear into real work, and those for whom the model becomes the work’s substitute. One group will do more than it could have done alone. The other will produce more than it understands. The outputs may look similar for a while. The divergence will show later, in who can still originate, refuse, connect, and decide when the machine has made the wrong thing easy.