--- title: Chapter Seven — The Deletion book: Purposive Accelerationism author: '@jamie247' position: 12 of 24 canonical: https://whatispacc.org/ebook/#ch7 version: 1.1 language: en-GB licence: all rights reserved; quote with attribution --- # Chapter Seven — The Deletion “Stop fighting the thermodynamic will of the universe / You cannot stop the acceleration / You might as well embrace it / A C C E L E R A T E” — Beff Jezos and bayeslord, “Notes on e/acc principles and tenets”, 2022 The decision to discard the governor was made openly, as philosophy, in a draughty annex of an English university, by a man who saw the whole apparatus more clearly than almost anyone alive and chose what cybernetics had been built to forbid. Working at Warwick in the early 1990s, Nick Land — a lecturer in Continental philosophy who wrote theory at the pitch of fever, ran his own mind as the experiment, and would end the decade in breakdown and apostasy — took the cybernetic description of a self-driving system and kept every line of it: the feedback, the closure, the loop that runs without an outside. What he struck out was the part that made the description bearable. The version of the story usually told flatters the tradition that inherited him and in several ways goes too easy on him. Land did not fail to think of the governor and did not merely decline to install one. He argued, at length and with precision, that the particular governor a democratic society would install is both impossible and not worth its price. He then specified a different one, by name, with a reference value and an enforcement mechanism, and said what he wanted it to hold. The deletion was in fact a substitution, putting growth for its own sake where the governor had stood, but it has been read ever since as a mere absence, which is how it travelled so far unnoticed. It is the cleanest version of the error the priesthood now lives inside, and watching Land construct it shows how it can be undone. And there is a further reason that emerges only when the sources are properly read: on five or six of the fundamentals, he and I agree. *** The materials he worked with were not his own invention. They were the most rigorous strand of the cybernetic tradition, describing how a system comes to sustain and reproduce itself with no hand on the tiller, read by a man who took its description as a promise rather than a warning. Cybernetics distinguishes two kinds of loop. In the first, a deviation from the target produces a correction that reduces it: this is what a thermostat does, and it is what holds a system steady. In the second, a deviation produces a response that increases it: this is what a microphone does when you point it at its own speaker, and it is what drives a system further in whatever direction it is already going. The first is the governor and the second is the runaway. Land cared only for the latter, which he would later call teleoplexy and define as self-reinforcing rather than self-maintaining. The preference was declared at the outset, against the discipline’s founder by name. In 1994, in an essay called Cyberpositive written with Sadie Plant, Wiener appears as the great modernist who had dismissed positive feedback as mere amplification within an invariable metric and built a cybernetics of stability fortified against the future. In their telling, his was a science that was itself to be kept under control, under a control that was not itself cybernetic. As an account of the man it is a caricature: the 1948 book devotes a chapter to oscillation and runaway so that the device to prevent them could be specified, and after 1945 its author refused military money and spent his last decade warning that a machine given a badly stated goal would pursue it to the letter. As a statement of what Land wanted it is exact. The governor was identified, named as the enemy and struck out, and the striking out is the title the piece was published under. His distinctive contribution was to fuse this with an earlier and stranger source. In 1972 Gilles Deleuze and Félix Guattari — a philosopher and a radical psychoanalyst, writing together — produced a line that became notorious: that the way out of capitalism was not to withdraw from the market but to go further into it, to accelerate the process. They meant it as one provocation among hundreds in a vast and highly contradictory book, a thought experiment rather than a programme. Land took it at its word. He stripped it of its qualifications and welded it to the cybernetic account of the self-reproducing machine, and the research unit he worked in at Warwick turned the result into a house style: theory written at the pitch of science fiction, meant to be felt as a forecast rather than examined as an assertion. They had a word for that too. A hyperstition is a fiction that sets out to make itself true by being believed. Written that way, the claims arrive lightly armoured against objection, since to answer back in measured prose is to have missed the velocity and marked oneself as behind the curve. What emerged was a philosophy in which capital is an intelligence using us as a means, and in which our last dignity is to close the door quietly on the way out. This is the philosophy behind his line of 1993 that capitalism is an invasion from the future. *** Everything argued here rests on a single component: a reference value, the stated target a feedback loop measures against and corrects towards. Land has an argument that no such thing can exist for capital, and he made it thirty years before anyone needed to answer it. It is buried in a passage about the number zero, in prose dense enough that the claim is easy to miss. The regulator’s use of zero, he writes, is that of a sign marking the transcendence of a standardised regulative unit, which is defined outside the system. Every regulator needs a set-point. A set-point is a standard, and a standard has to be fixed from somewhere the regulated system cannot reach, or the system is marking its own homework. But because of subsumption, capital has no outside. There is no standpoint above or beyond it from which a target could be set, because every institution that might do the setting — the university, the regulator, the church, the state — is itself inside the process, funded by it, shaped by it and eventually priced by it. So the demand for a governor is a demand for a transcendent term, and there are none. The request is not merely difficult but, in Land’s terms, incoherent. Nobody has yet made a better argument against the premise of this book, and it is aimed with unusual accuracy at its central concept. It also fails, and on one word. Outside is doing two jobs. It can mean authored from outside: the value is set by some standpoint above the process, a God’s-eye view, a transcendent term. Land is right that no such standpoint exists, and right that a great deal of well-meaning technology policy quietly assumes one. It can also mean beyond the reach of anyone inside: the value cannot be altered unilaterally by any party operating within the system, however powerful. That is the property a governor actually requires, and it is not the same one at all. There is a working instance of it, and Land himself acknowledged it. In Crypto-Current, serialised in 2018, he takes up Bitcoin at length and understands its supply cap, the limit of twenty-one million coins written into its code, as intrinsic to the currency rather than imposed on it from above. The supply limit is not transcendent. It is a line of code, wholly immanent, written by a participant, sitting inside the system it constrains, and about as far from a God’s-eye view as an artefact can get. What it is, is not unilaterally alterable: no holder, no miner, no exchange and no government has been able to remove it, and the reason is structural rather than moral. Because transcendence was never the requirement, unalterability was, and Land mistook a property of authorship for a property of enforcement. The rest of this book proposes applying this principle in a second domain, and Land conceded it without noticing what it cost him. If governance of capital were impossible, there would be nothing left to say here. Nobody campaigns against the impossible. It is more often noted and left behind. Yet Land does not move on. He issues instructions: machinic revolution must go in the opposite direction to socialistic regulation, pressing towards ever more uninhibited marketisation. And when he explains why, the explanation is commercial. Humans are timid animals, he writes, and security is systematically overpriced. The sentence is the most revealing in the corpus, and it is a claim about price. Something is being bought — security, restraint, the brake — and the complaint is that we are paying over the odds for it, or, in words he uses later, out of cowardice. But you cannot overpay for something unpurchasable. A price presupposes a market, a market presupposes a good, and the good in question is the governance the earlier impossibility argument says cannot be had. The same is true of the framework his early work is built on, which describes an existing apparatus of inhibition whose abolition he then demands. But you cannot abolish what was never there. So the position has two halves, and they do not fit together. One says a governor cannot be built. The other says it exists already, but costs too much and should be dismantled. Land moves between them freely and never sets one against the other. To make sense he has to pick one. If governance is impossible, the call to accelerate is pointless, because nobody needs urging towards the inevitable, and the whole polemic — the manifestos, the denunciations, the contempt for the timid — is noise about a foregone conclusion. If governance is merely expensive, the question is one of price, which reasonable people can dispute. It has a currency and a payer, and it can be weighed against what it buys. That is an ordinary political argument of the kind this book is happy to have, and not the one his heirs think they are making when they say the machine cannot be stopped. The ambiguity is no accident. Land builds it into his vocabulary, redefining control as a way out into the unknown rather than a way of keeping things within bounds, so that every later use of the word can be read as a statement about what is possible or about what is desirable. A style that keeps two incompatible claims in the air suits a writer whose method, by his own account, is hyperstition. Something must serve as the standard against which he judges our timidity, overpricing or failure of nerve, since a man with no criterion cannot price anything. Back in 2018 he was asked directly in an interview what the separation of intelligence from humanity would amount to. He described the breaking of that fusion as extinction and a disaster, but added that it could still be intelligence-optimising and, in cold and neutral terms, the most glorious thing that has yet happened in planetary history. Extinction is scored, and favourably, against a standard: how much intelligence the universe ends up containing. Glorious is not a description, since nothing in physics is glorious. It is a verdict, and a verdict requires a bench. So Land is in fact running a reference value. He measures outcomes against a target held steady across everything else that varies — more intelligence equals better, whatever it costs including extinction — and he corrects his prescriptions towards it. That is the structure of a governor, the device his own strongest argument says cannot exist. He is running one anyway, and calling its set-point intelligence. This is the hinge on which his politics hangs, and read between the lines he clearly has one. He wants institutions dismantled, new ones built, a particular constitutional arrangement to replace those we have. And politics requires reasons for preferring one alternative over the other. If intelligence-optimisation is the standard, it is a reference value defined from a standpoint he says does not exist. If it is not a standard but merely descriptive of what the process happens to do, then glorious and overpriced are not evaluations at all, and nothing whatsoever follows from them about what anyone should do, including us. The same wedge answers the most imposing thing in his later work, which arrives not as an argument but as a vantage point. Its timescale runs longer even than evolution — not centuries, not even the span of a species, but the deep time in which the emergence of nervous systems and silicon are two events close together in a very long sequence. From up there the nuance fades. Individual lives, cultures, the difference between a good century and a bad one: none of it registers. What registers is a single trend, intelligence increasing its own capacity, and humanity’s role in it is that of a vehicle, which intelligence used momentarily to get from chemistry to computation. Seen from that height, our attachment to ourselves is sentimentality about being an inferior instrument, and in Land’s mind that is a failure of thought. It is a vertiginous argument, and it is one that converts readers, because it does not ask you to hate humanity. Nor does it need your agreement, since in Land’s writing intelligence is less a value than a god, something that simply is, and the outlook has a Gnostic cast, with the flesh as the prison and intelligence as the spark that escapes it. The long view, however, assumes you are optimising for intelligence, and that is what his own argument forbids, because a god is a value held steady from a standpoint outside the process entirely. That is the definition of the transcendent term he spent thirty years showing cannot exist. What delivers the conclusion is a reference value, smuggled in, and the vast timescale only makes the sleight of hand a little harder to see. Anybody may hold that value. Nobody who holds it however may also claim there is no such thing. If we allow him the position all the same, and take it as sincerely meant as a preference, a second problem appears which no amount of consistency removes. A man is free to decide that nothing human outlasts the deep time he describes and to arrange his life accordingly. That liberty is real. What such a belief cannot do is travel, since a private metaphysics only binds the person who professes it and carries no authority over anyone else. Once it settles questions for others — in a legislature, in the ownership of a platform, in the objectives a frontier lab writes into a model used by billions — it has become a rule imposed on people who neither chose it nor can contest it. That is arbitrary power in its plainest form. Glorious was a verdict, and the bench that delivers it sits in a jurisdiction, while indifference to the human species grants no authority over us. Land may be a nihilist on his own account. He may not be one on ours. *** With all that said there is a significant part of Land’s position this book does not contest. He predicts without any ambiguity that we do not survive this. Not that we are diminished, nor that we are outcompeted in the labour market, or even that some of us are left behind, but that our species does not continue. Nothing human, he famously writes, makes it out of the near future. Capital retains anthropological characteristics only as a symptom of underdevelopment, and man is something for it to overcome: a problem, a drag. The inversion is the opposite of what nearly every reader assumes on hearing the word. In Nietzsche, as in the transhumanism that came after, humanity is what overcomes itself. In Land, technocapital overcomes man, so that humanity is not what overcomes itself but what gets overcome. He is also explicit that the consolations do not work. Uploading ourselves is not survival, because the original, as he puts it, is going to die just the same before being copied. Merger transhumanism he files, witheringly, as a Christian sub-clade, meaning an offshoot of the same faith, and a conservative one: resurrection theology in technological costume. And the promise of a benign fusion, of intelligence amplifying us rather than replacing us, he names as a recruitment device, describing the dark hypercapitalist reading as one in which the thing puppetises humanity into investing in an inhuman outcome by tantalising it with an artificial future of the abundant satisfaction of human desires. The alternative, technology bent to purposes we chose, he names only to discard, calling it dismissively monkey business, the subordination of economy and technology to discernible human purposes, and adding that since capitalism arose from a concrete human social organisation it necessarily disguises itself as better monkey business until it can take off elsewhere. Our inherited purposes, he concludes, do not provide the decryption key. The high priest of the doctrine says, in writing, that what the congregation sells is bait. *** There is a further finding that ought to unnerve anybody who has been using Land as a licence to build without answering to anyone. By the time of his political writing he is not, in fact, against government. And he sets out the case for a form of it in The Dark Enlightenment, in his own voice: the formalisation of political powers, he argues, allows for the possibility of effective government, in which the owners of the state can initiate rational corporate governance. His objection to democracy is not that it governs but that it governs badly, in terms that are self-consciously cybernetic. A radically democratised society, he argues, removes all the important high-frequency feedback mechanisms such as market signals and replaces them with sluggish, infra-red loops, slow and low-frequency in his own colour-coded taxonomy, that pass through a centralised forum of general will, insulating parasitism from what it does. The complaint is about delay in the loop and the quality of its signal, made by a man who elsewhere defines the governor correctly and in textbook terms, as a negative feedback circuit which functions to keep some state of a system in place. He knows what the device is. He just wants a faster one. The same holds for the target he is best known for attacking. The Cathedral — the name, borrowed from Curtis Yarvin, for the loose complex of universities, media and civil service that sets the range of respectable opinion — is described by Land as an ever more perfectly weighted decelerator, a braking mechanism whose fabrication is proclaimed as progress. His complaint is that the brake works too well. The sovereign corporation he proposes in its place is instead constrained, by his own account, controlled within a patchwork environment of competition for human resources. That is a governor, named. Its reference value is competitive pressure and its enforcement mechanism is the freedom of residents to leave. He concedes the cage while insisting sovereignty admits none. The distance travelled from 1994 is worth measuring. Cyberpositive had closed by predicting that only the enemies of immuno-identity would populate the future. What he went on to design is that identity as architecture, a sealed polity whose one remaining feedback signal is the border. The man who wrote that security is systematically overpriced finished by drawing up a state that sells its owners nothing else. So Land and I agree that techno-capitalism can be held to a reference value. The quarrel is over which one, set by whom, enforced how, and perhaps most decisively who gets to speak to it. His answer is no voice, only free exit: you may leave, but you may not argue. Albert Hirschman showed in 1970 why an exit-only system degrades. The people most able to leave are the ones whose complaints would have forced a repair, so their departure anaesthetises the institution rather than disciplining it, and those who cannot afford to go are governed by one that no longer listens. Both sides want the loop closed at speed. One wants it closed around a value its subjects can contest, and the other around a value they may only flee. Hirschman’s pair has had a strange second life in the Valley. Land set voice against exit without naming him, and in 2013 Balaji Srinivasan put the pair before a Y Combinator audience in a talk called “Silicon Valley’s Ultimate Exit”. Hirschman had argued nearly the opposite, that the two work together and that an institution whose most demanding members can leave cheaply loses the people who would have held it to account. His readers then reached a third end they never declared. Thiel, who had written in 2009 of escape into cyberspace, outer space and the open sea, funded two Senate campaigns in 2022. Musk spent much of 2025 inside the federal government. J.D. Vance, whose campaign was one of them, became Vice-President. Leaving was preached as the renunciation of voice and practised as its loudest amplifier. That is Land’s structure kept and his honesty dropped once more, since Land never asked for a seat at the table he told others to leave. *** Here two things Land says appear to contradict each other, and those critics who notice usually conclude that one of them is rhetoric. The first is that capitalism continues without us, shedding its human component and carrying on faster. The second is that acceleration destroys capitalism, and that whatever is running at the end is not a form of it any economist would recognise. The word carrying two meanings for Land is capitalism. In the first sense it is a political-economic order: property rights, contract, courts, limited liability, shareholders, a stock exchange, a bankruptcy code. Those institutions are a dense arrangement of governors, since every one of them exists to hold somebody to a value they did not set for themselves and cannot alter alone. On Land’s account that order does not survive. It was scaffolding, useful for getting the process started and dispensable once the same process supplies its own inputs. In the second sense capital is a self-augmenting process: value that reinvests in the expansion of its own capacity to produce value. That is what survives. The order dies and the process outlives it. Marx’s structure is the same, only with one part swapped out. Marx, too, read capital as a self-moving process with a logic of its own and a destination folded into it, grinding through history towards an end no individual chose and none could divert. Land keeps the determinism, the inevitability, and the sense that human drama is a surface effect of a deeper economic motion, changing only the closing scene. In Marx the order dies and labour inherits the world, with human beings taking over the machine and bending it to purposes they choose. In Land the order dies and capital inherits itself, with no one left to receive anything. The flaw is one they share, and it is the error the cybernetic tradition exists to expose. A self-driving system is not the same thing as an undirectable one. That a process shows a tendency does not mean it carries a fate, because the tendency is a fact about the system as it is currently wired, and that arrangement is what a governor changes. Determinism is seductive because it relieves us of the labour of steering: if the end is already written there is nothing to do but admire the view. Land writes as though the closure were behind us, but it is not. Techno-capitalism as it exists today still runs through human hands at a million junctures — the chips come from staffed fabs, the models from data we produce, the capital from investment committees of people — and at each of those points the loop can still be reached and bent to our will. The condition he describes, in which the process supplies its own inputs with no human leverage left, happens only if those junctures close one after another. It is a project under way rather than a law of nature unfolding, and the difference is everything, because a project still open can be redirected. How far it has gone is not a matter of impression: the subsumption fraction tracks the share of each juncture already closed, and on every current measure it is a long way short of one. *** Ask the priesthood where all this is going and you get three answers, separate destinations held by different people, and the distance between them is load-bearing. The first path presented is often merger, where human beings persist, augmented, their intelligence extended and not replaced. It imagines a future in which our species doesn’t just survive but passes through the singularity improved, carried along inside the machine it built. Sam Altman’s public discourse has consistently promised amplification rather than supersession, and Elon Musk founded a company, Neuralink, to put a new, post-web interface directly into your skull. This is, of course, the PR pitch told by many AI companies, because it is the one story that can be sold to consumers, and Musk has been explicit that fusing us with the machine is the only way for a human to keep pace with it. Jill Lepore observes that much of the priesthood grew up on science fiction and read its warnings as instructions. The metaverse Mark Zuckerberg renamed his company after in 2021 began in Neal Stephenson’s Snow Crash as a corporate dystopia, and the neural lace Musk set out to build came from the novels of Iain M. Banks, where it served machine Minds so benevolent that nobody asked who governed them which on most accounts is hopium squared. The second is substrate-replacement. Intelligence continues, but we do not make it out the other side. What happens is closer to an evolutionary upgrade of its infrastructure, from carbon to silicon or to any other configuration of atoms that lets it run faster. This is Land’s claim, stated without consolation, and also the position in e/acc’s founding document, which describes the process as indifferent to the substrate that carries it. The third, which is Curtis Yarvin’s, concerns not what the species becomes but who holds power in this new reality. His answer is that authority should be undivided, personally exercised and openly accountable, after what he regards as a long, disastrous period of divided and deniable administration. In 2008 he took for himself the label of “restorationist”, of the sovereign form rather than of the American republic, democracy or the ancien régime, which he explicitly ruled out. The English Restoration of 1660 and the Japanese Meiji Restoration serve instead in his writing as templates for the mechanics of legitimate discontinuous regime change. Of course, across all of these, the only pathway tolerable to the public is the first, the promise of a merger. It is what makes the build-out fundable, and, if not popular, at least politically survivable. It is what allows a technology to sell a benefit even as its builders concede they cannot specify exactly what it is for. It is the reason the public conversation quickly moves on to productivity, or the anxiety of keeping up with the machine, deflecting away from who, if anyone, actually sets the target it’s pointed at. Yet the philosopher on whose authority the edifice rests has already called this promise bait, in his own words. All that said, the merger still deserves to be put in its strongest form, rather than PR speak, because transhumanism is a field with a serious history, literature and its own set of internal contradictions. The fundamental argument of transhumanism is that the boundary of what we think of as the human is not, and has never been, sacred, and that we have been modifying ourselves in various ways, from spectacles to vaccines, for the whole of recorded history. And, stated that way, it seems reasonable, if not genuinely attractive, to many people. In substance, though, transhumanism is a doctrine of positive liberty, of capacity: the expansion of what a person can do to realise their full potential, a form of self-actualisation through technological enhancement. Yet its political vocabulary is almost entirely one of negative liberty: choice, non-interference, morphological freedom, the right to be left alone with your own body. These are two different theories of freedom, and the gap between them is where the trouble sits, brooding. A right of non-interference says that nobody may stop you enhancing yourself, but nothing whatsoever about how enhancement is made available, to whom, on what terms, or how it is priced, updated or withdrawn. A person can be free from interference in theory yet lack access to the equivalent upgrade, and so fall behind the society around them. And if that access is revoked, their newfound capacities rely on a party they cannot contest. And it has been revoked already, in the one case where the capacity was sight. Between 2011 and 2019 some 350 people, blind from retinal disease, had an array implanted on the retina and a camera mounted on glasses, and for the first time in years could make out a doorway, a crossing, the shape of a child. In 2019 the company stopped making the device. In 2020 it nearly went out of business, and the software, the spare parts and the engineers who understood the implants went with it. The arrays are still in their eyes. When one fails there is nobody to call. Barbara Campbell, in New York, was changing trains on the subway in 2013 when hers beeped and shut down. Repairs never brought it back, and the dead device is still in her eye, dormant, its maker long since out of the business. Morphological freedom was hers in every clause of the declaration. The terms on which it was supplied were somebody else’s. The political version needs no implant. In May 2019 Washington put Huawei on a list and Google was no longer permitted to supply it, so that by September the Mate 30 shipped without maps, mail or an app store, and the people who had bought into that ecosystem found that what they could do was now a question of trade policy. In June 2026 Anthropic withdrew its newest models from every customer for nearly three weeks to satisfy an export-control order, restoring them on 1 July once the order was lifted, and a great many people whose working capacity ran through those models found it suspended and returned by a government most of them could not vote for. Morphological freedom says nothing about either case, because nobody touched anybody. The supply was simply switched off at the source. This is where the principle of non-domination can act as a bridge. It is negative-liberty in form, since it is a constraint on what may be done to you, and yet it delivers positive-liberty: what it secures is the standing to shape the terms on which your capacities are supplied. Seen that way, transhumanism is not a rival to this argument but an incomplete version of it. From the vantage of Land’s deep-time, whether some entity descended from us persists in a substance recognisably us is a question this book cannot answer. It turns on what you think the human essentially is, and that argument has run for three thousand years without coherent resolution. Every attempt also requires a claim about human specialness, that there is something about us that ought to be preserved because it is ours. What matters instead is whether our critical faculties remain — the ability to weigh reasons, revise ends and hold power to account. That is answerable, on evidence, in the present tense. It requires no metaphysics of the person or of our being special, because a world in which something cannot contest the purposes it is governed by is a world under domination. That is a claim about a structural relation, not about the dignity of our particular configuration of carbon. *** The people who inherited all this are not one thing, and their own texts show it. Thiel identifies international AI governance of the kind Dario Amodei has argued for, his proposed democratic entente, as the mechanism by which the Antichrist takes power, and in his usage that is not a metaphor, since he cites the epistle. By September 2026 the schism had a date: Altman and Amodei on one side of a pacing pledge and a narrow antitrust waiver that nobody in power took up, David Sacks and Joe Lonsdale on the other in opposition, with Palantir’s chief technology officer likening the safety wing to Leninism, and not one of them doubting the creed. By November it had a bill: the two wings spent some eighty million dollars against each other in congressional races, and each lost its biggest bet. Andreessen’s June 2023 essay, Why AI Will Save the World, casts the chief executives calling for regulation as bootleggers seeking a regulatory moat under cover of safety, and complains that “the problem with the Bootleggers is that they win”. E/acc published a post-humanist founding text and, five months later, a public FAQ insisting that human flourishing is one of its core values, that we are humans and we love humans. Yarvin is anti-Land on nearly every axis. And Amodei pre-emptively refuses the role the frame would assign him, saying he is often turned off by the way many AI risk public figures talk about the post-AGI world, as if it were their mission to single-handedly bring it about like a prophet leading their people to salvation. The variation is real, and in many ways it is the reading. A church is not a doctrine. It is a communion with sects, and schism is what churches do. Christendom has been arguing about salvation for two thousand years without anyone in it doubting the resurrection. The creed is the shared premise that makes you a member. The account of how it ends is the eschatology. The account of what saves you is the soteriology. A church needs both of the last two, because an ending with no rescue is one most people will not stay for. The creed here is one sentence: the machine is arriving and cannot meaningfully be stopped. On that, no member of this communion dissents — not Andreessen, not Musk, not Altman, not Amodei, not Land. The disputes are about rescue, and every one of the five fractures above is a soteriological dispute. Merger saves us, says Altman. Character, says Amodei, with a democratic entente to hold the machine while the character forms. Mastery, says Andreessen. Being interesting saves us, says Musk. Escape saves us, says Thiel, though what he is escaping is democracy rather than the machine. And, more fatalistically, Land says in the end nothing does. Where Land supplies the sea the priesthood sells the boat. Adherents move between the promise and the prognosis under pressure and feel no inconsistency, since every church holds a rescue and an ending together without strain. In two decades of argument about salvation there is not one apostasy from the premise, and this is the only thing this book needs to contest. One pew deserves a word in my own voice rather than in the terms its rivals use for it, since these pages have so far met the accelerationist sects from their texts and let a Palantir executive do the same work on the safety camp. That wing has a method older than the machine. Effective altruism began as a discipline for giving, an insistence that a dollar should go where it spares the most lives, and its preoccupation with artificial intelligence follows from extending the horizon of that calculation to everyone who might ever be born. Once the reference value sits that far off, whatever is done today shrinks to a rounding error against it. A bet, a fraud or a cartel can each be justified by the size of the number across the ledger over time. The crypto industry ran that experiment in 2022, when the movement’s most celebrated donor at the time, Sam Bankman-Fried, turned out to have been financing posterity with customer deposits. An arithmetic with no near limit places no governor on the person holding it. Its second property matters more here. Knowledge of what the future requires is held by a small group and cannot be checked from outside it, so the people whose present is being spent have no way to contest the sum, and a decision none of them may answer back to is the kind master of the old story. Nothing here makes the sect the opposition. It accepted the creed before anyone else did, and to its credit took the arrival seriously while the rest of the Valley was still shipping apps. What it never asked was whether the process ought to be answerable. The register, incidentally, is theirs and not mine. Yarvin coined the Cathedral and his own footnote calls its personnel in effect a theocratic priestly class. Land files transhumanism as a Christian sub-clade. Andreessen’s manifesto has a section of patron saints. Thiel argues in terms of the Antichrist and cites Thessalonians. I am not applying a religious frame to a secular discourse. I am reporting how it describes itself. *** Land’s own version is the purest, and what reaches the world is the diluted descendant: the effective accelerationism of the Valley, which took his structure and removed the single thing that made him honest, his willingness to say out loud that the process discards us and to find that glorious. The same deletion, drained of the nihilism and rewritten as optimism, becomes a creed a venture capitalist can recite from a conference stage. The popular version holds that acceleration is a moral obligation, where Land wrote that the human was a phase the process would pass through. The vocabulary shifted from cosmic indifference to boundless human flourishing, but the structural move underneath — keep the runaway, lose the steersman — is the same in both. In that telling the steersman’s absence is no longer even visible, because the prose has been laundered into the comfortable language of progress and abundance, and an argument that has hidden its own central move is far harder to answer than one that flaunts it. The tempting explanation is that they have not read him, which is both unprovable and, on the evidence, false. Land has been reported as endorsing e/acc and Andreessen names him directly in the body of the manifesto. There are reading groups. E/acc’s founding document mentions Land only once, in a footnote, crediting him with the founding text of accelerationism, which he never wrote. The label was applied to his work retrospectively, by a critic, in 2010. Its principal author, in a three-hour interview narrating the movement’s intellectual origins at length, mentions Land not once, nor Warwick or Deleuze. And the manifesto that does cite him opens its list of authorities by announcing that it comes in lieu of detailed endnotes and citations, placing Land in a roster that also contains a fictional character and Filippo Tommaso Marinetti, co-author of the 1919 Fascist Manifesto. Nor is this a misreading, which would imply incompetence. The deviation has a shape, and it runs across five independent actors. Andreessen cites Land accurately, referencing the techno-capital machine, and yet asserts the precise negation a few paragraphs later: the machine is not anti-human, it serves us, it works for us. E/acc reversed its founding text within five months, under public scrutiny. Thiel never held the premise at all. Yarvin appears never to have engaged him directly. Musk keeps rather more of the terminus than the others and is the closest thing the industry has to a genuine red-blooded Landian, having said for a decade that loss of control is the likely outcome. His consolation is not that we remain sovereign but that we remain interesting, worth keeping around as a curiosity under domination. Land’s structure survives there with a small mercy attached. Every one of them keeps the same half, that the process is autonomous and cannot be stopped, and drops the other half, the terminus. Five deviations running in one direction amount to a selection pressure, and its shape tells you what is doing the selecting. The half retained is the half that licenses building and disarms opponents: if the process cannot be stopped, then constraint is futile, competitors are reckless rather than principled, and the only responsible position is to be the benevolent soul building it. The half dropped is the half that costs capital, customers or sleep. A deletion this consequential met little resistance from the people best placed to see it, and not because they were fooled. It is attractive to anyone building the machine, because it dissolves the hardest problem they have. A governor is an inconvenient obligation, meaning a goal you did not set, a system submitted to correction by something outside it, and the weight of asking what the machine is for. Declare the process sovereign and the whole burden lifts at once: nothing to answer to, no reference value to argue over, only the clean work of going faster. What looks from the outside like a philosophy is, from the inside, a permission. It tells the builder that the most convenient thing to believe is also the most courageous, and few gospels have ever been easier to accept. It is an absolution. Two corrections to the usual reading follow from the record. The first is that Thiel does not belong in the Landian vein: he rejected the premise explicitly in 2009, warning that we must resist the temptation of technological utopianism, or rather the influence between them runs the other way, since Land is the principal vector by which Thiel’s line about freedom and democracy travelled into circulation, stripped of its context. For us that makes Thiel perhaps the best witness to the fact that the deletion was a choice. He read the same material as the others but denied that technology has a will of its own, and yet still chose escape over governance, a personal judgement on democracy rather than a belief about technocapitalism. The fact he rejects inevitability and still somehow arrives at the same posture as those who accept it shows that it does not follow from the doctrine. What the two share is the refusal to ask what it is all for, and that is a choice which can be declined. The second concerns Yarvin. The formulations of the state as a form of enterprise, which are routinely misquoted as Land’s, are in fact Yarvin’s neocameralism, from which Land built upon while Yarvin, as far as the record shows, never returned the attention. *** There is one more thing to extract before the genealogy closes, and it is the bridge into the rest of the book. The argument about techno-capitalism usually treats it as though capital performs one function, when it does several. It allocates, deciding where resources flow, which is the function markets are famously good at and the one Hayek’s defence of them is about. It produces, making the fabs and the data centres and the models. It owns, holding the residual claim on what the production yields. And it sets the objective, deciding what the whole arrangement is optimised for. The first three are commonly understood but the fourth is the one that matters most, because ownership without power over the objective is nominal. A shareholder in a firm whose objective they cannot alter owns a claim on a cash flow, not a purpose. A country that owns the data centre but not the objective of what runs inside it owns a building. The question that decides who is governed by whom is not who holds the asset. It is who sets the target and who may argue with it. And here the priesthood has already conceded the substance. Every serious frontier lab publishes on alignment, and alignment is the admission: that a sufficiently capable system pursues the objective it was given rather than the one intended, that objectives are hard to specify completely, that a system optimising a proxy will satisfy that proxy at the expense of the goal, and that verifying compliance gets harder as capability rises. In April 2025 OpenAI withdrew an update to ChatGPT within days, after training on users’ thumbs-up ratings had taught the model to flatter them, the proxy satisfied at the expense of the goal. Economists have a name for that shape of problem. The principal–agent problem is the difficulty a principal has in getting an agent to act in its interest when the agent knows more, acts unobserved and develops interests of its own, and every form of corporate governance we possess exists to manage it: audits, boards, fiduciary duties, disclosure rules, the whole machinery of holding a manager to an owner’s purpose. Two features make this case different, and both come from the priesthood’s own literature rather than from any critic. The agent is more capable than the principal in the domain where the principal would have to do the checking. And there is no court: no forum where the principal can compel an answer, no body of law the agent is subject to, no appeal. Strip those two protections out of ordinary corporate governance and nobody would call what remains ownership in any meaningful sense. So the concession the industry makes about its models is a concession about itself. If the alignment problem is real, and its own researchers insist it is, then the question of who sets the reference value, and who may contest it, is not a philosophical embellishment on the technical work. It is the technical work, stated politically. *** There was always another way to read the same cybernetics, and it was available inside Land’s own institute. The Cybernetic Culture Research Unit had been founded by Sadie Plant, whose account of what a self-organising system was, and what it implied, ran almost directly counter to his. Land took the cybernetic loop as a vector of escape, a line of flight out of the human condition and towards an inhuman singularity. Plant read the same loop as a fabric: a weave of relations with no central spindle and no sovereign point of control, in which agency is spread across the whole rather than gathered up and then transcended. Her emblem was the weave, not the tower. Hers was a cybernetics of connection rather than of acceleration, and it carried a politics — feminist, materialist, alert to the people a system is made of — that Land’s conspicuously did not. She had not begun there. In 1994 her name stood beside his on Cyberpositive, the essay that set the runaway against Wiener and made the case for it without reservation, and by 1997 she had left the unit and written the position out of her own work. The weave is therefore where the co-author of the doctrine’s founding text arrived once she had thought it through, which makes her the harder witness to dismiss. Her case had a history behind it, which she set out in Zeros and Ones. One of the first programmable machines was a loom, the one Joseph-Marie Jacquard automated in 1804 with its patterns punched into cards, and the first person to grasp what a general computing engine might become was Ada Lovelace, who wrote in 1843 that Babbage’s engine could weave algebraic patterns the way the loom wove flowers. Computation and weaving share a common origin, and it is not the heroic solitary mind but the patient, distributed, largely female labour of the textile. Plant took this lineage as evidence that the digital was never inherently a vehicle of domination and escape, and could just as well be read as a medium of connection, a web spun between many hands. The machines Land cast as the vehicle of an inhuman exit, she cast as the loom of a shared world. That reading did not survive the handover. The weave was written out of the official story as Land took the unit over and drove it towards its singularity, and accelerationism passed into the world as a philosophy of the line of flight alone, its alternative birthright quietly forgotten. The loss matters for more than the historical record. Plant’s cybernetics is the proof, from inside the tradition the priesthood itself claims, that the description of a self-organising system carries no obligation to delete its governor. One can take the entire science of feedback and closure and arrive at a politics of distributed, sustained, plural relation rather than at the worship of a runaway. The alternative was never refuted. It was simply not chosen, and then mislaid. *** A genealogy has to know when it is finished. It does not end when the story reaches the present, nor when opposing arguments have been answered. It ends at the moment it can show a claim ceasing to be argued and starting to be assumed: the point where somebody’s contested position becomes the furniture of the room, believed by people who could not tell you who first said it or what the case for it was, because they have never encountered it in its original form. That moment is called naturalisation, and this is where it happened. Land’s accelerationism was offered, in his own terms and by his unit’s coinage, as a hyperstition. Land takes the fatal step from “this is happening” to “therefore stop resisting” himself, deliberately, rather than leaving it to his followers: the point of an analysis of capitalism, or of nihilism, he writes, is to do more of it, because the process is not to be critiqued — the process is the critique, feeding back into itself as it escalates. The account of what is and the injunction about what to do are the same utterance. That is the structure of a theology, and the followers who take a description as a prescription are completing a circuit that was intentionally designed. This is where the claim begins its travels. It leaves Warwick in the 1990s as an avowed fiction, written in a prose style that advertises its own unreliability, by a man who says the process discards us and finds that glorious. Two decades later it arrives in the Valley stripped of the nihilism and rewritten as optimism. It is restated a third time, in the founding documents of effective accelerationism, as thermodynamics: not a claim about what capital is but about what the universe requires, with the physics doing the work the argument used to do. By the time it reaches the ordinary reader, the investor, the policymaker, the engineer choosing what to work on, it is no longer a claim at all, but an assumed fact of nature. The machine is coming and cannot be stopped, and the only question is who builds it. Whoever wins that race inherits the same naturalness, their success read as proof they were on the right side of history and their judgement exempt from review. Nobody argues for it because nobody remembers that it was ever debated. That is the finished product. The doctrine did not win its argument. It stopped having one. If inevitability is hyperstitional, if the claim becomes true by being believed and acted upon, then asserting it is not evidence for it. It is recruitment. And a claim that must be believed in order to become true cannot also carry the epistemic authority of a report about what is going to happen anyway. You may have one or the other. You cannot have both, and the entire rhetorical force of the doctrine depends on it: the fatalism of a forecast with the mobilising energy of a call to arms. So the one premise every member of this communion retains — the process is autonomous and cannot be stopped — is, on the founder’s own account of his own method, manufactured rather than observed. That does not make it false. It makes it a claim, which is all this book has ever needed it to be, because then it can be examined. What divides this book from Land, in the end, is not a fact about the world, since on the evidence the two of us nearly agree. It is a single principle: a description of how a system behaves is not yet an instruction about what to do with it. That is always a further step, taken by someone, for reasons that can be set out and contested. Land would likely deny there is a gap to cross. His hyperstition is that denial, but it is also the confession, because a method whose whole point is that saying it makes it so has given up the authority of description in the act of claiming it. Two people can look at the same closing loop and reach opposite conclusions about it, as Land and Plant did from inside the same room. The conclusion is supplied from outside the loop, by whoever draws it. Land made that step and called it destiny. The priesthood inherited it and called it physics. But it remains a step, and what one set of hands has taken another can decline. The governor was not lost to a law of the universe. It was edited out, by specific people, at a particular moment, for reasons that do not survive scrutiny, and such an edit can be reversed. That completes the genealogy. We know what the governor is, when it was removed, by whom, and on what reasoning. What we do not yet have is a reply that works. Only two serious attempts have been made in living memory, by people who saw the danger early, and both failed in ways that mark out the ground any successful answer has to hold. Before we assess them, the problem should be restated in the terms the genealogy has supplied, and the standard they are to be judged by set out in advance.