I spent last Saturday in Shenzhen — the engineering city just across the border from Hong Kong, where I often spend a weekend — in the company of a close friend of mine, a designer named Seazov, whose particular field is the architecture of interiors. We have, over the years, returned several times to a small standing argument about his eye. He is, by any honest account, unnervingly good at the bare-eye measurement of anything. He can stand in a doorway, take it in for a few seconds, and tell you, within a half-centimeter or so, what the contractor’s tape will subsequently confirm. He has a phrase he reserves for this gift, half boast and half observation, which he produces whenever a colleague reaches for a ruler in his presence. “I do not carry a ruler,” he says. “My eye is the ruler.”

I have been turning the remark over since I first heard it from him, in late 2024. It is the sort of self-description that has not been available to most professions for a long time. There are crafts in which the body has remained a credible instrument — the designer’s eye, the perfumer’s nose, the pianist’s hand, the wine merchant’s palate — and to listen to a master of one of them is, among other things, to listen to a person whose body still carries inside it an apparatus that the rest of the world has long since produced in metal and plastic. Seazov is one of these. The ruler lives, in his case, behind his eye.

I find myself thinking about him for a separate reason, however. The matter I have been trying to work out, for the better part of two years, is not really about the persistent gifts of designers and perfumers and pianists. It is about a different sort of yardstick — the one we have always carried, more uneasily than any other, for the differences between human minds — and what has begun to happen to it in our own decade.

What I have noticed, over the past two years or so, in conversations with friends and colleagues who would once never have allowed themselves the latitude, is a certain new willingness to make observations about other people’s minds outright. To say that a particular person reasons more carefully than another. That a certain colleague has unusually weak abstraction. That a certain founder is unusually quick. The remarks are made in much the same tone in which one might once have said that someone was a faster runner: with a kind of clinical, not unfriendly, slightly bored acknowledgement of a fact about the world. In 2018, in 2020, even in 2022, they would have been considered, if not actively rude, then at least the mark of an uncultivated speaker — the kind of person who had not yet been informed that certain things were not to be said. They are not, in 2026, considered any such thing.

I confess I was slow to put a name to what was happening. The first explanation that came to me was the obvious one: that we had become, in the small ways one does, slightly coarser, less observant of the formal courtesies that had governed conversation for a generation. This is the explanation most people are inclined to offer, particularly those who would prefer that the world become coarser slowly enough not to require any adjustment of their own opinions. But I have come to think this is not what has happened, or at least not the most important thing that has happened. The change is real, but it is downstream of something else. What has changed, I would now say, is not the manners. It is the location of the thing being discussed.

*   *   *

Consider, for a moment, what happened to physical strength.

For most of human history, to comment on another person’s body was a delicate matter. Strength, like intelligence, was understood to be a property of the person, inseparable from the person, and the language available for discussing it was the language of insult and praise rather than of measurement. A man was strong or he was weak, and the words carried evaluative weight. To say a man was weak was, in any straightforward sense, to take something from him.

This changed, with surprising speed, once force itself began to be produced outside the body. The eighteenth century gave us Watt’s separate condenser; the nineteenth gave us the steam press, the locomotive, the Krupp hammer that drew crowds at the 1862 Exhibition; the early twentieth gave us, perhaps more important than any of these for the purposes of public consciousness, the photograph of the man lifting the bar. The outputs of these machines could be measured in newtons and horsepower and joules, and once the measurements existed, they became available for the comparable evaluation of human muscle as well. A man could be said to bench-press one-twenty or two-twenty in much the way one might describe the capacity of a small motor. The vocabulary was the same. The vocabulary was, importantly, not insulting. Charles Atlas, taking his advertisements to the back pages of the comic books and asking the kicked-sand-in-the-face boy to consider a course of improvement, was making, almost without anyone noticing, an argument about the measurability of personhood.

By the late twentieth century the elaboration was nearly complete. Athletic ability had been disaggregated into a dozen specific measurements — vertical leap, hundred-meter time, reaction speed, maximal oxygen uptake, the ratio of fast-twitch to slow-twitch muscle fiber — and to say someone had a poor vertical leap was no longer an insult, because it was a measurement, and a measurement, by its nature, carries less moral charge than a verdict. A. J. Liebling could write, in his sweet-tempered books on prizefighting, of fighters whose particular gifts were precisely catalogued: the man who could only fight at a certain weight, the man whose reach exceeded his footwork, the man whose chin would not survive the third round. No one, reading him, felt that anyone had been insulted. The whole conversation had been moved out of the register of personhood and into the register of functional description. People still cared, of course, about being good at sports. But the comparison itself, the bare act of placing two human bodies side by side and observing their differences, had become merely interesting rather than transgressive.

What made the difference was not, ultimately, a change of heart. It was the arrival of an external yardstick. Once force could be produced and measured outside the human body, the human body could be measured against the same scale, and the measurement no longer felt like an attack.

*   *   *

Intelligence, almost alone among the human capacities, did not undergo this transition for a very long time.

The last two hundred years have steadily externalized one human capacity after another. Movement was externalized by the engine. Calculation was externalized by the abacus, then the calculator, then the spreadsheet. Memory was externalized by the printing press, the library, the database. Vision was externalized by the camera. Voice was externalized by the telephone. Pattern recognition, in narrow domains, was externalized by the various technical systems of the late twentieth century. But reasoning, judgment, abstraction, the synthesis of language, the formulation of arguments — these stayed, stubbornly, inside the human being, and because they stayed inside, evaluating them remained a personal matter rather than a functional one.

The twentieth century did not, for want of trying, succeed in settling this in public. Francis Galton, in the 1880s, set up his anthropometric laboratory at South Kensington and measured the head circumferences and reaction times of London visitors who paid threepence for the privilege of being recorded by science; Binet, in 1905, devised his scale to identify French children needing extra schooling; Spearman, around the same time, proposed his general factor of intelligence; Lewis Terman, at Stanford, gave the work its American name. By mid-century the apparatus had become an industry, and by the 1970s it had become a scandal — Gould’s Mismeasure of Man arrived in 1981 with the air of a long-overdue eviction notice. The trouble with the whole enterprise was not, in the end, that it failed to measure anything. The trouble was that it tried to measure something that existed only inside people, with no external referent, and so every score functioned as a verdict on the person rather than as a measurement of a function. There was no second object against which a human mind could be compared without the comparison feeling, to all involved, like a comparison of human beings to one another. It was as if one had been asked to compare two athletes in a world before stopwatches: one would have only adjectives, never numbers, and the adjectives, being adjectives, would do moral work.

This is what changed, suddenly, around late 2022.

Whatever else may be said about the wave of large language models that arrived in those years — and a great deal has been said, much of it foolishly — it constitutes the first industrial-scale production of something we are willing to call intelligence outside the human body. It is partial. It is uneven. It is, in many respects, oversold. Anyone who has spent serious time with the systems is in no doubt about their failures, which are not subtle. But the fact remains that we now sit in front of objects which formulate arguments, synthesize across domains, write reasonably, plan over horizons, and produce judgments that we then evaluate. We rate them. We rank them. We compare GPT against Claude against Gemini against the various open-weight systems, and we do so in a register that has become, in the past three years, remarkably clinical. We say one model has weak long-horizon planning, that another has poor calibration, that a third writes well but reasons poorly, that a fourth executes well but cannot strategize. The language is the language of functional measurement, of the sort one would have applied a century ago to small motors.

I recall a conversation last winter, in an office in Beijing, where I have lately been spending my time, in which two gentlemen I have known for some years were discussing the relative merits of three different models for a piece of strategy work. They spoke without rancor and without metaphor, the way two engineers might once have discussed the relative virtues of competing pumps. The vocabulary was completely available to them. Neither felt that he was being indelicate. The whole exchange lasted perhaps fifteen minutes, after which they returned to the kind of conversation they would have been having in 2010, in which the subjects compared were people. The transition between the two conversations was, I noticed, frictionless. The language did not change. Only the noun did.

And this language, once available, has begun to migrate. It is beginning to flow back into the way we describe people, not because we have collectively decided that the old taboo against ranking human minds was foolish — most of those I know have not decided any such thing, and would deny it strenuously if asked — but because there is now, for the first time, a yardstick that did not exist before, and once a yardstick exists, comparisons made against it stop feeling, in quite the old way, like attacks on personhood.

*   *   *

A few things follow from this, and they are easier to confuse than to separate.

The first is that the language used to talk about cognitive difference is, slowly, disaggregating. The crude question of whether a person is smart, which was always the question that did the worst damage, is being replaced, unevenly, by a more granular description: what a person can plan, over what horizon, with what working memory, with what tolerance for ambiguity, with what calibration. This mirrors what happened with physical fitness once the yardsticks arrived. The athlete used not to be simply strong or weak; he had a certain endurance, a certain explosive power, a certain reach, a certain rate of recovery. The bluntness of the old vocabulary gave way to a finer one, and the fineness, paradoxically, made the comparisons less injurious, because no single measurement was ever the whole verdict.

What I expect to see, in the next decade, is the formation, in ordinary conversation, of something like a cognitive operating profile. It will not be a number. It will be a pattern of strengths and gaps, of the kind any decent talent evaluator already keeps privately about the people they know well, except that we will become more willing to say these things aloud, in something like the manner in which one might say that a friend has good cardio but a weak left knee. The strangeness of this prospect, in 2026, comes from the fact that it has not yet become fully permissible. By 2030 I suspect it will be the ordinary way of speaking, and the social commentators will write their pieces deploring it, and most of us will go on doing it anyway.

The second is that the connection between holding an opinion and being a particular person is loosening. For most of recorded history, an opinion was a piece of the speaker, in the literal sense that defending it was defending oneself. This is the deep reason most disagreements collapse into ego defense: there is no way, structurally, to attack a position without also attacking the person who is currently inhabiting it. Once the opinion can be lifted off the speaker and examined by an external system, however, this changes. One can ask, of any view, whether its evidence chain is complete, what its strongest steel-manned counter looks like, where the framing is biased, whether the argument is reaching past what the evidence will support. One asks these questions, increasingly, of a machine rather than of the speaker, and the machine answers without the speaker losing face. The opinion becomes a kind of module — held by the speaker for the moment, but available to be revised, refactored, replaced — in a way that does not feel like personal defeat. It becomes, in something like the sense of software engineering, a draft under version control. I find this, on balance, civilizing, though I admit it is taking some getting used to.

The third, and to me the most consequential, is that debate is losing its ancient function as the primary mechanism for the production of truth. For most of human history, the only way to combine the information held privately in many heads, and the reasoning held privately in many heads, was to bring those heads into the same room and let them argue. The argument was, in a real sense, the cognitive engine of a society. The clash of views was where information got merged and where logic got pressure-tested, because there was nowhere else for the merging and the pressure-testing to occur. This is what gave parliamentary debate, courtroom adversary, and academic disputation their genuine, not merely ceremonial, importance.

Once the merging can be performed upstream of the argument, by external systems capable of holding many positions simultaneously, the argument loses much of that function. What remains for debate to do, and what it will continue to do well, is the coordination of values, the formation of consensus among humans who have to act together, and the making of public commitments. These are not small functions; they are, in a sense, the only ones that ever genuinely required the room to be full. What goes away is the older sense, found throughout classical writing on rhetoric, that the argument itself is the place where truth is found. Increasingly, the truth has been found before the argument starts. The argument is now about what to do about it.

Dissent, in this new arrangement, undergoes a small but interesting reversal. The person who insists on raising the awkward counter-position, the colleague who can be relied upon to find the embarrassing exception, the contrarian who slows the meeting down: these were once treated, organizationally, as friction. They will increasingly be treated as a kind of signal, an indication of where the system has a blind spot, provided there is something in the organization capable of absorbing the contrarianism and converting it into a structured input. The annoying person, in other words, has a future. He may, for the first time in his career, be paid for what he is doing rather than tolerated despite it.

*   *   *

And this, I think, is the part of the change that begins to matter most.

The traditional model of leadership — the one on which most of the institutions of the last century were built, and about which there exists a management literature whose volume has not been matched by its utility — is a model of dominance. The leader was the person who, in any sufficiently contested room, could outargue the people in the room, hold ground against challenge, and impose a decision through some combination of judgment, force of personality, and willingness to absorb risk. Winning the argument was, in the end, what made them the leader; the subordinates’ acquiescence was the proof of the proposition. The libraries of business schools are stocked with books about this figure, written for the most part by their admirers.

I do not think this is going to be the dominant model in twenty years.

What is replacing it, in slow stages, is a different model, which I would call integrative, though the word does not entirely capture it. The new leader is not, primarily, the person who can defeat the other voices in the room. The new leader is the person who can metabolize them. They take in the information held in private heads, the local knowledge of the operations team, the legal anxiety of the compliance officer, the cultural intuition of the regional manager, the worry of the engineer who has seen this kind of failure before, and they produce, out of these inputs, a synthesis that is more complete than any single voice could have offered. They are still the ones who take the final call. They still own the outcome. What has changed is the source of their authority, which is no longer the suppression of voices but the integration of them.

This is not, I want to insist, a softer kind of leadership. If anything it is more demanding. It requires framing skill, taste, an ear for bad-faith input, a willingness to take the final hit when the synthesis turns out to have been wrong. Dominance was, in its way, the lazier of the two models — one needed only to outargue. Integration requires the harder thing, which is to listen with discrimination. The people who will do it well are not, by and large, the loudest in the room. They are the ones who can be relied upon to have heard everything that was said, including the parts that were said quietly, by people whom the older model would have ignored.

Put as simply as I can: in the older arrangement, leadership was the art of being stronger than the people around you. In the newer one, it is the art of turning the people around you into a more complete system.

*   *   *

These four things — the demoralization of cognitive comparison, the loosening of opinion from identity, the decline of debate as the engine of truth, and the shift from dominance to integration — are not four different trends. They are four projections of one shift, which is that intelligence has begun to leave the body.

The brevity of the formulation conceals the size of the change. We have grown used to thinking of artificial intelligence as a kind of capability upgrade, a tool that does more for the same effort. I do not think that is the most important thing about it. The most important thing about it is that it has moved intelligence to a new address. It used to be the case that to evaluate intelligence one had to evaluate a person, and to argue with intelligence one had to argue with a person, and to lead intelligence one had to dominate persons. None of these implications survives the externalization. The social technologies built on top of them — the ones we have been carrying around inside our institutions, our offices, our friendships, our marriages — are quietly losing their grip.

I should say, before I close, where I think this argument has to slow down, because I do not believe in its triumphal version and I do not want this piece to read as one. There are conflicts no amount of synthesis dissolves. The fight over the marketing budget is not a fight about information; it is a fight about resources, and laying out the information more cleanly does not make either side surrender. The fight over who gets the credit, who keeps the office, who carries the blame, is not a fight about reasoning; it is a fight about position, and it is as old as the office itself. The fight over whether short-term growth is worth long-term reputation is not a fight about logic; it is a fight about value, and there is no machine that can prefer-rank values on a human’s behalf without becoming, in some way, the human. These three classes of conflict are the same in 2026 as they were in 1926, when Henry Luce was choosing his cover stories. The new arrangement does not touch them. It only makes the cognitive conflicts, the ones that always were tractable in principle, more visibly tractable in practice.

And there is a further worry, specific to the new arrangement, which is that the integrative layer is not neutral. If the system doing the synthesis carries a bias, integration silently becomes amplification of the bias. The same machine that can absorb dissent can also flatten it, and one cannot tell, from the outside, which of the two is happening, unless one has been very careful about the design of the integrative apparatus. This is the central technical and cultural risk of the new model of leadership, and any honest defense of integration has to take it seriously. To wave it away would be to repeat, in a new key, the error the technology industry has made many times already, which is to mistake an unspoken assumption for a neutral default.

*   *   *

I have spent my professional life inside the technology industry, and for most of it I would have said, with some confidence, that the question of how intelligence is distributed among human beings was not a question I would discuss in public. The discussion struck me as politically dangerous and personally indecent. I noticed, in the days since that Saturday afternoon — and at perhaps a dozen smaller occasions before it, occasions which, taken individually, would not have seemed worth recording — that I have stopped feeling either of those things quite as strongly. The reason, I have come to think, is not that I have grown coarser. It is that something has happened to the object of the discussion. The thing I was once reluctant to compare across people has, in part, left the body. It has acquired a yardstick. The comparisons are still consequential, and still touch real lives, and still ought to be made with some care. But they are no longer, in the old way, comparisons of human beings to one another. They are comparisons against a measure, and that, as Liebling’s prizefighters discovered a century ago, is a quieter kind of comparison than one might have feared.

The old etiquette is loosening. I find, when I think back to Seazov in Shenzhen, that I do not mourn it. His eye will remain his own; no engine I know of will measure a room better than he does. But for the human capacity we have always been most reluctant to compare across people, the yardstick has stepped quietly out of the body and into the world. The old etiquette served a generation that did not have such an instrument. We have one now, and the honesty it makes possible is not, in the end, a cruelty. It is something closer to what most of us have always been quietly aware of, and have been waiting, perhaps, for permission to say.