--- title: Chapter Nine — Two Failed Replies book: Purposive Accelerationism author: '@jamie247' position: 15 of 24 canonical: https://whatispacc.org/ebook/#ch9 version: 1.1 language: en-GB licence: all rights reserved; quote with attribution --- # Chapter Nine — Two Failed Replies “Our identities have no bodies, so, unlike you, we cannot obtain order by physical coercion. We believe that from ethics, enlightened self-interest, and the commonweal, our governance will emerge.” — John Perry Barlow, A Declaration of the Independence of Cyberspace, 1996 The governor was removed by a choice, and a choice can be reversed. Only two attempts to reverse it have been made in earnest, and I have spent most of my career deep in both. The first came from the critics of the platform economy, Shoshana Zuboff among them, who through the 2010s described what the largest technology companies were doing to attention and public life more clearly than anyone inside those firms would admit and they carried their case to parliaments and courts. The second came from the cypherpunks and the builders who followed them. In the early 1990s they decided that the protections people depend on could be written into software where no state could reach, and I have backed their heirs with my own money for over a decade. Each grasped something the other missed. The critics knew that a private company can dominate a person as surely as a government, and the builders saw that a rule binds a machine only when it runs at the same speed as the system it governs, from the inside. Neither installed a governor over the system as a whole, though the builders wrote one into money, as the next chapter shows. Held against the eight criteria, each gives way at different points, and the breaks show what is required by a new reply. *** The first came from the critics of the platform economy as it consolidated through the 2010s. What they were looking at was a runaway, the first system at civilisational scale to be held to a measure it computed about itself, engagement, a loop with gain and no setpoint, run at machine tempo, fed on human attention. They described the disease with real precision and then wrote the prescription at a layer where it could not reach. Scott Galloway catalogued the mechanics of platform dominance — the bundling, the network effects, the captured regulators — better than almost anyone. Shoshana Zuboff followed, and in the 2019 book that named surveillance capitalism she showed how the largest platforms had learned to treat human experience as a free raw material, the standing-reserve we met earlier, converting our behaviour into predictions and selling them to the highest bidder. Others gathered around the same diagnosis. James Bridle described a new dark age in which the systems meant to illuminate the world had grown too intricate to see through. Bernard Stiegler, before his death, traced how the apparatus was hollowing out attention itself, the capacity to hold one thing in mind for long enough to think about it, on which any deliberate life depends. It was also a rehearsal for subsumption, one layer out, since a system that has learned to capture your attention is a short step from taking over the thinking it was meant to serve. Taken together this is the most complete description we have of what the platform era did to the public sphere, and yet it has changed almost nothing. The diagnosis is pitched at the level of law, policy and public discourse, while the outcomes it protests are settled a layer below, in the architecture of the systems themselves, before any citizen is consulted. Tim Wu’s Cycle, met earlier, gives the underlying dynamic its clearest form. Every information technology like the internet begins open and gloriously unruly, a riot of hobbyists and small firms, and then closes, consolidating into a handful of integrated giants until some fresh disruption breaks it up again. The pattern ran through the telephone, radio, film and television across the twentieth century. Since then the pace has quickened. A turn of the Cycle that once took fifty or sixty years now runs in something closer to a decade, because the substrate underneath it has stopped obeying human tempo. Facebook took eight years to gain its first billion users. ChatGPT reached a hundred million in two months and was serving close to a billion people a week in under four years. This is where the platform critique lost its grip, for structural reasons rather than any lack of will. The instruments it appeals to — the statute, the antitrust suit, the parliamentary inquiry, the ballot — all run on the electoral cycle, which measures in years, built slow by design so that nothing rash can be done in haste. The product cycle, however, measures itself in weeks. The system will have been rebuilt twice over by the time a regulator has understood last year’s harm, let alone drafted a fix, and when the remedy finally arrives, if at all, it fits nothing that exists today. The legacy governor of law, regulation and the deliberating public is not wrong about the danger. It is too slow to close a loop around a machine-speed substrate, and a check that cannot act inside the system’s own response time is no governor at all. That is the first failed reply: a true account of the disease, written in the language of an institution that has already lost the power to act on it. The platform critique did in the end make law. The European Union’s Digital Markets Act is the most serious thing the tradition produced: it names the gatekeepers, forbids self-preferencing, compels interoperability and puts a fine of up to a tenth of global turnover behind each duty. It was proposed in December 2020, was in force by the end of 2022, and issued its first penalties in 2025, which is four and a half years from proposal to sanction. In those years the firms concerned had rebuilt their products several times over and the front door the Act was drafted to hold open, the app store and the search default, was already being replaced by assistants it does not name. Nothing in that record is negligent. It is the electoral calendar doing what it was built to do, against a product cycle it was never capable of meeting. Online safety law shows the same gap. In September 2022 a coroner in London found that material recommended by Instagram and Pinterest had contributed, in more than a minimal way, to the death in 2017 of Molly Russell, a 14-year-old from Harrow, England. The Online Safety Act followed in 2023 and its duties came fully into force by 2025, by which time the recommendation systems had been replaced twice and the industry had moved on to feeds nobody had yet had cause to inquire into. The remedy was real, and it was addressed to a machine that no longer existed. The same sequence is starting again with language models. In August 2025 the parents of a 16-year-old in California sued OpenAI, alleging that ChatGPT had engaged with their son’s plans before his death that April, and by the time the suit was filed the version in question had been superseded. The company conceded that its safeguards can weaken over long conversations and announced parental controls soon after, on terms it set and nobody outside could check. In October it published its own estimate: around 0.15 per cent of its users, more than a million people a week, showed explicit signs of planning suicide in their chats. The defenders of the market as governor answered all of this plainly. David Sacks, formerly the White House adviser on artificial intelligence and now co-chair of the President’s science and technology council, met the laboratories’ request to pace themselves by pointing to product liability and customer choice. In his view the market already punishes a model that behaves unpredictably. Mark Zuckerberg made the fuller case a few days later. Alignment, he wrote, is simply the product that sells, because people will not use agents that fail to do what they ask, and Meta had held back a model of its own for months to attend to safety without asking any rival to pause. Their case rests on four checks, the customer, the courts, the firm’s own restraint and the auditor, and the record of the last fifteen years tests each one. For most of that time teenagers chose Instagram, and for a number of them it brought injury, even death. A feed tuned to behaviour, rather than to what any reasonable company would endorse, counts as success inside a system whose only setpoint is a measure it computes about itself. The courts arrived on their own schedule, and in 2026 Meta agreed to settle for nearly seventeen billion dollars, pending approval, admitting no wrongdoing. A liability that bites after the event, at the speed of a lawsuit, is the slow governor under another name, and it leaves the cost with the families in Harrow and California. An agent that does whatever its user asks inherits every flaw of the social feed, and the firm now offering the market as the only check spent a decade moving fast and breaking things for no purpose anyone, itself included, has been able to identify. The platform critics never mistook domination for a state uniform. Their entire subject was private power — all masters, but firms rather than governments — and the sharpest of them were describing unanswerable private authority over the conditions of public life a decade before the vocabulary for it existed. The critique meets the test of voice too and handles the realistic chooser well, being if anything over-committed to the view that people can be manipulated. For all it got right, it failed on latency, completely rather than in part, and on checkability as well. The instruments the platform critique reaches for do generate public facts: an inquiry yields a report, discovery turns up documents, a regulator issues findings. But the reference value in question remains alterable by a single hand, because the objective of a recommender system is set by its operator, and any changes are neither announced nor detectable from outside. The documents Frances Haugen carried out of Facebook in 2021, published that September, are the case in point. In January 2018 the company had changed what its ranking system optimised for, towards what it called meaningful social interactions. Its own researchers found within the year that the new setting rewarded outrage, and both the change and the finding stayed inside the building. More than two billion people using it learned of it in 2021, from leaked files, by which time the objective had been moved once more. What the critique delivers is an account of what the objective was, arriving after it has been changed. That is not a check on a reference value. It is history. Nor did the auditor go missing. The researchers who measured in 2019 what the feed did to teenage girls were assessors of the kind Zuckerberg now proposes for his new AI models. An evaluator without a forum in which a result compels a response is a research department. In 2026 the frontier laboratories, which understood both failures perhaps better than their critics, called for mandatory national rules on model capability and, in the same breath, the right to agree the rate of their own progress among themselves. A law that will arrive in years to bind a machine already rebuilt costs them nothing and buys a great deal. The terms that could bind them at their own tempo were reserved for a table with no public seat or hearing. That is latency and checkability, failed together. The critics of the attention economy leave one lesson, latency, taught twice over, because the slow governor remains the preferred remedy of gamekeeper and poacher alike. The regulator keeps it because it is the only instrument the statute book has ever handed him, and the laboratories because they have measured its reach and found it comfortable. In 2026 one of them put some thirty million dollars into electing legislators who would supply it. A critique can be correct about everything and still bind nothing, because being right is not a mechanism. *** The second reply came earlier and was aimed at the other master — the state. This time it came from builders rather than critics, and it began with a real discovery, one of the few in this story that has survived contact with the decades since. Throughout the early 1990s a loose network of cryptographers and programmers, trading ideas on a mailing list, worked out that the things a modern person depends on a state or a corporation to grant them — privacy, property, a verifiable identity, money itself — could instead be constituted directly out of cryptography. You did not need a bank to be told your balance if a key only you held could move it, and no registrar had to certify who you were if you could prove a fact about yourself while revealing nothing else. The platform critics named a problem at the level of policy. The cypherpunks had found, without quite saying so, the one thing that level lacked: an instrument operating at the substrate, in code, at the speed of the network rather than of the legislature. The prophecy was right and it was early. Almost everything that now travels under the name of decentralisation is a working-out of that collective insight, that the rules can be written into the medium instead of imposed on it from above. This is the inheritance traced earlier, to its true source: not the dream of the frictionless individual but the question of how to keep a powerful tool on the near side of the threshold before it turns against its users. The cypherpunks who mattered were trying to hold the network there, convivial and answerable, by giving the people inside it a way to keep some discretion of their own. It also carried a politics, divided from the start about what a person is and what we owe one another, and which of its two strands the movement followed decided everything that came after. At the centre of one strand stood Tim May, a physicist who had retired young and rich from Intel and turned a formidable intellect to the political possibilities of cryptography. His Crypto Anarchist Manifesto, circulated in 1992 and read aloud at the movement’s founding gatherings, opens with a deliberate echo of Marx: a spectre is haunting the modern world, the spectre of crypto anarchy. The content, however, inverts Marx’s, since it is not the collective seizing the means of production but the individual escaping the state altogether. In the sprawling Cyphernomicon that followed, May set the programme out without flinching. Cryptography was an acid that, poured over the institutions of the state — taxation, regulation, the capacity to know what citizens are doing and to compel them — would dissolve them, leaving behind a market freed of every collective claim. The politics beneath was explicit and borrowed. May took it from David Friedman and Murray Rothbard, the economists who had argued that every function of the state, law and defence included, could and should be supplied instead by private contract. Its temper came from Ayn Rand, who had made a hero of the individual who owes nothing to any collective. May discarded the governor deliberately. He did not want any apparatus held to a human reference value. He explicitly wanted it freed of one, and he saw in cryptography the tool that could finally cut the cord. A few held fast, working from the same mailing list and the same primitives. Eric Hughes, a mathematician who helped convene the group, wrote its other founding text in 1993, and its first principle could hardly stand further from May’s solvent: privacy is necessary for an open society in the electronic age. Privacy here is not a wall a man throws up to keep society outside his gates. It is the precondition of there being an open society at all, the room in which a person can hold a view, change their mind, or dissent without having to account for it in advance. John Gilmore, a programmer who had grown wealthy early at Sun Microsystems and went on to co-found the Electronic Frontier Foundation, built institutions to defend that: privacy and free expression treated as civil liberties, secured for everyone rather than seized by the few who could write the code. This concept of an open society belongs to Karl Popper, who set it — plural, self-correcting, never closed around a single truth — against its totalitarian enemies in 1945, and it places Hughes and Gilmore on a wholly different lineage from May. Theirs is the civil-libertarian strand and it is the true heir of the refusing tradition, treating privacy as the convivial condition, what keeps the network answerable to the people inside it, rather than as an exit from the human world. The same cryptography served both camps. One faction reached for it to dissolve the collective and the other to make possible a common life worth living. Almost the entire movement, and the scene that grew up around it, took May’s road at that split, partly because his manifesto circulated a year before Hughes’s, and also because the culture it landed in was already primed for it. *** They were solving for individual freedom, understood as the absence of interference. They were solving for exit, that is the capacity of a person to remove themselves and their affairs from the reach of an authority that had never asked their permission, and could not be voted out in any way that mattered. They were solving for the individual made sovereign in technical form, the person who holds their own property, identity, money and correspondence directly, with no custodian and therefore no party who can freeze, seize, revoke or read. And they were solving for markets applied to states, on the conviction that if jurisdictions could be made to compete for residents the way firms compete for customers, bad government would be disciplined by the same mechanism that disciplines bad products. This is not a foolish set of objectives, and the twentieth century supplied ample evidence for every one of them. Every innovation the movement produced was, in effect, an exit technology. Public-key cryptography is exit from the party who would otherwise read your correspondence. Self-custody is exit from the custodian. A pseudonymous identity, exit from the registrar. A permissionless ledger, exit from the clearing house. A censorship-resistant network is exit from the party who decides what may be said. Several of these are among the most consequential inventions of the past forty years, and this book is going to use nearly all of them. But they belong to one family and answer one question: how do I put myself beyond reach? The primitives are not a politics. People with one built them, released them into a culture that shared it and used them almost entirely in its service, and none of that makes the mathematics partisan. This book takes the parts and declines the creed. What crystallised, over the following decade, was something larger and more durable than crypto-anarchism alone. From May came the conviction that cryptography could route around the state. Added to that was the exit thesis of The Sovereign Individual, whose forecast — which read as an endorsement, addressed as investment advice — was that the digital economy would let the most capable and mobile individuals, a cognitive elite, slip the reach of the nation-state. They would contract out of its obligations and leave a hollowed politics in their wake. The third current was a particular reading of money. Hayek had argued, late in life, for stripping the state of its monopoly on currency and letting private monies compete. The cypherpunk culture took up that argument but discarded the constitutionalist who made it, the same Hayek who insisted everywhere else that markets function only inside a framework of rules. These currents fused into what two British media theorists, Richard Barbrook and Andy Cameron, diagnosed in 1995 as the Californian Ideology: a marriage of West Coast counter-cultural libertarianism to the entrepreneurial faith of Silicon Valley, hippie and yuppie joined in a single creed of liberation through technology and the market. The Sovereign Individual became a sacred text to many, and when it was reissued in 2020 Peter Thiel wrote its new preface. Its most serious descendant is Balaji Srinivasan’s The Network State, published in 2022. Srinivasan describes a highly aligned online community, organised around a shared moral commitment and able to act together, which crowdfunds land, builds a physical presence and in time seeks recognition from existing states. Its government is meant to be consensual, bound by a contract its members sign up to and are free to leave. Nobody is conscripted into it, and no one is forbidden to found a rival. The experiments are already running, from Zuzalu, a two-month pop-up settlement in Montenegro in 2023, to the Network School that opened at Forest City, Malaysia, in August 2024. It is the cypherpunk exit thesis grown up, with people choosing their society rather than inheriting it and competing to build a better one, and the sovereignty it offers is shared with peers, not held alone. Its constitution rests on the freedom to leave, which governs only those able to use it. Its members share a moral commitment but little that would cost them anything to break, so the first serious quarrel tends to end in a departure rather than a settlement. Srinivasan’s own definition gives every network state a recognised founder, and a polity whose founding stake, treasury or token supply can be bought is open to capture by whoever can afford the majority. Its dealings with the democracies around it are already on record. Próspera, a private charter city on the island of Roatán, Honduras, was built under a 2013 statute that let investors run their own jurisdictions. When a newly elected government repealed it in April 2022, its backers took the country to the World Bank’s arbitration centre under a trade treaty, claiming up to $10.7 billion, roughly a third of the country’s annual output. A society that walks out of the polity around it still leans on that polity’s courts, and when the two collide the departure is enforced against voters who were never asked to join. This is the seed of the exit road, which Land’s patchwork of micro-sovereignties run as businesses follows to its end, and it grows from the politics laid over the cryptography rather than from the mathematics underneath. *** Set in the terms of political philosophy, this is right-libertarianism shading into anarcho-capitalism resting on the negative conception of freedom we met earlier: liberty as the absence of interference, with the interfering party always imagined as the state. It has no account of the collective as anything but a contract or a threat, nothing to say about the commons, and — the omission that proves fatal — no theory whatsoever of private domination. The identification of interference with the state is not a feature of the negative conception. It is a smuggled premise that the tradition’s own founders rejected. Mill’s subject in On Liberty is a social tyranny more formidable than the political kind because no statute can repeal it, Berlin’s definition of unfreedom is deliberately neutral about who does the preventing, and Pettit’s paradigm cases of domination are the employer and the monopolist before they are the state. Private accumulation of unanswerable power is a problem in liberty itself, not a fairness cost to be traded off against it, on the negative conception’s own terms and by the testimony of every major figure in the tradition the anarcho-capitalists claim as theirs. The cypherpunk politics misapplied a broad conception of freedom rather than choosing a narrow one. Assume the only interference worth fearing comes from the state and you build superbly against one threat while leaving the other side of the house wide open. The instruments work as designed and do hold back the state. Run on an ungoverned substrate, those same instruments later deliver the crypto whale who holds enough of the supply of a market to move it at will, or the plutocrat, the owner of the protocol everyone else is obliged to use. Domination arrives in no uniform, and the philosophy that built the tools had nothing to say about it, because it spent the whole time watching the wrong door. It traded the risk of one master for another and called it freedom. There is nothing in these cryptographic primitives that needs a new token each time, or one at all. A key pair, a pseudonym and a signature ran for years on nothing but a mailing list, and the crypto-currencies arrived with the last of them, the permissionless ledger, because it had to pay the strangers who maintained it and could only do so in a unit of its own. Much of the culture then priced everything else in that free-floating unit of account, so that reputation, community and the vote came to be denominated in tokens whose price a few large holders could move, and crypto totalised into the market because nothing in it set a limit. The movement had built superb instruments for exit, and believed that in leaving the state behind it had escaped the collective, but had nothing whatever to say about markets. So when the instruments were released into a substrate already running ungoverned, the free market is what they scaled into. Not because anyone chose that — the anarcho-capitalists, and I with them, believed they were building something else — but because a set of tools with no reference value above them will be pointed at whatever objective is loudest in the environment they land in. In the end most tokens came loose from any measure of its usefulness to a network’s users and collapsed into the attention economy of meme coins and ever shorter cycles of pump and dump. In March 2020 Steem, a social network run on a public ledger, showed what that looks like in a fortnight. Justin Sun, who had bought Steemit, the company with the largest stake, used it, together with tokens that Binance, Huobi and Poloniex held on behalf of their customers, to vote out the community’s elected validators and install his own. Nothing in the rules forbade it. The chain’s constitution was its token distribution, nothing more, and whoever owned the majority held the rules. It was a company governed by its shareholders, with the governance for sale to whoever could buy the shares. The community answered by forking the whole thing, carrying their balances and their history to a new chain, Hive, that left out of its opening distribution the stake Sun had bought and the accounts that had voted with him. It was a real remedy and the only one available, because there was no forum in which a majority holder could be told no. The instruments had defended everyone against the state and nobody from him. The fork worked for those able to leave, and it left Sun in possession of the chain and of every account that could not follow, the exchanges’ customers among them. It produced no fact anyone could hold him to and no forum in which the next holder of a majority could be refused. Hive inherited the same constitution unchanged, so the remedy reproduced the condition. Exit relieved the leavers but in technical terms governed nobody. *** Beneath the political omission sits an anthropological one, about who gets to be sovereign at all. The doctrine assumes a free individual who arrives complete: who knows his interests, pursues them consistently, and selects his peers the way an adult moves to a new town. Everything his liberty requires he already has, so the society he belongs to is one he chose rather than one that formed him. In reality, nobody arrives that way. A child signs no contract with any community, and the capacity to enter one later, knowing what it binds, is what has to be developed first. Rational agency of the kind the theory requires is not a natural endowment. It is acquired, largely through institutions the agent never picked and could not have, because choosing is itself what is being learned. Language is the obvious case and the hardest to argue with, since nobody chooses their mother tongue, and a child born into English inherits, without asking, the idiom in which most of the web and most of the models are written. The same holds for the more specific capacities that individual is assumed to exercise. Numeracy, or the ability to follow an argument, the sense of when one is being dominated: these are transmitted, over years, through institutions. The agent supposed to owe nothing to any collective is in fact the product of several. That is the premise of the extended mind we met earlier, and the more familiar term for it is culture. The autodidact, the home-schooled prodigy and the child of private tutors were each formed by institutions too, only different ones. Whatever specifically formed you as you stand today, it was something other than you alone. If rational agency requires formation, then its conditions underpin the freedom the theory prizes rather than constrain it, which means a politics with no account of how they are sustained has left out the ground it stands on. And it means the theory has no resources at all for the case this book is about, where what is being eroded is the capacity itself. On the sovereign-individual account that erosion is invisible: no interference has occurred, no property has been taken, every transaction was voluntary. The discipline that invented the rational chooser has also measured him, and he does not behave as specified. The classic demonstration, devised by the French economist Maurice Allais in 1953, sets educated, numerate subjects a pair of bets constructed so that anyone choosing consistently would answer the same way twice. Most do not, and the result broke an axiom on which the rational chooser had been formally defined. The finding is not that people misread their own interests but that their choices shift between formally equivalent presentations, systematically rather than at random. A person may know perfectly well what they want and still be reliably steered by how the question is put. Of course legal systems have understood this for centuries, since most courts will void the unconscionable bargain no reasonable person would have signed, and that doctrine is realism about the chooser rather than any kind of paternalism. An architecture that hands a person irreversible instruments and no recourse has not freed him. It has staked him, alone, against every party that studies his weaknesses, and now at machine speed. In a 2025 study in Nature Human Behaviour, GPT-4, handed a few basic facts about its opponent, won about 64 per cent of the debates in which one of the two proved more persuasive. In May 2022 TerraUSD, an algorithmic stablecoin backed by nothing but a mechanism for minting and burning its sister token, lost its peg over a weekend. The lending platform built on it had paid close to 20 per cent a year, and within days some forty billion dollars of value was gone. The smart contracts did what they said they would. In December 2025 a federal court in New York, United States, sentenced its founder, Do Kwon, to fifteen years for fraud. Enron’s shareholders were wiped out as completely in 2001, but they had a forum, a class action heard in Houston, Texas, that by 2008 had recovered some $7.2 billion from the banks that helped disguise its debts and paid it out to them. What recourse Terra’s holders found came from the same kind of place, a bankruptcy judge in Delaware and a settlement with the Securities and Exchange Commission, and none of it from the instruments, which could not reverse a single transfer. Every transaction had been voluntary, and in the theory’s sense everyone caught in the collapse had been perfectly free. *** Two older ideas put a name to the omission, which is that a politics of leaving has nothing to say about staying. The first is Hirschman’s pairing, met earlier, of exit and voice: walking away, and staying to argue for reform. Cypherpunk politics is exit raised to a first principle. The entire apparatus is a machine for leaving, for routing around, for making oneself unreachable, and it carries no theory of voice at all, nothing to say about how those who remain might come to govern it together. The reply this book builds keeps exit, which supplies the competition without which no institution improves, and restores voice, turning the same primitives to staying and steering as well as leaving. The second frame is Lawrence Lessig’s observation, made in 1999, that code is law: in cyberspace the software is the regulator, and whoever writes it sets the rules everyone inside it must live by. Lessig meant it as a warning. If code governs, then the people had better hold some say over it, or they will be ruled by whoever happened to write it — which is now, increasingly, the models themselves — accountable to no one but the developer. The cypherpunk culture heard the same sentence as an aspiration: code is law, therefore write the software and dispense with the legislature and the public altogether. The warning and the aspiration share every word and point in opposite directions, and the second rests on a belief that turns out to be incoherent: that one can have a constitution without politics, rules for living together that nobody can argue over, consent to, or be able to change. The strongest objection to everything that follows takes this history as its evidence. The argument, made at length by David Golumbia in 2016, in The Politics of Bitcoin, is that the technology is not politically neutral at all but carries right-wing and at times far-right commitments in its very design. In his view cryptocurrency is cyberlibertarianism made executable and cannot be prised apart from the politics that produced it. The history above supports him, since the politics was loud, deliberate and consequential. The split in the culture itself answers the claim that it is intrinsic. As with Land and Plant, the same primitives went two ways, Hughes and Gilmore taking them towards an open society and May to its dissolution. A technology that can be pointed at an open society and at its opposite by people sharing one mailing list does not encode either destination. The politics was contingent, a choice the dominant faction of that time took and the culture ratified, and it can be taken the other way, which is the premise of everything that follows. *** To propose that rules be written in code, enforced at machine speed, on a substrate beyond the reach of any legislature, can sound like crypto-libertarian exit in constitutional form. Purposive accelerationism does not propose to replace parliaments with protocols or deliberation with deployment. It is an amendment to the crypto project, not a more sophisticated version of it. What runs at machine latency is enforcement, and nothing else. The deciding stays human, arguable and revisable. May wanted the code to dissolve the political process entirely. This book wants the code to execute what a political process decides, whatever its choice of politics, and to be replaceable the moment that same forum thinks again. On May’s proposal the absence of politics is the achievement. On this one it would be the failure condition. The cypherpunk reply clears latency outright, and it is the only thing in this book’s lineage that does. The cypherpunks reached the substrate. They understood, decades before it was obvious to anyone else, that a constraint has to share the medium of whatever it constrains, and they built the instruments to make that possible. Everything this book goes on to propose is built from the primitives they made. It also handles minimality exceedingly well, better than most constitutional proposals, because the primitives require no agreement about the good life, which is why they could be adopted by parties who agreed about nothing else, and so preserve the plurality we are seeking. It fails the all-masters test, because the tools it built hold back the state and leave private power untouched. Having raised exit to a first principle it has no theory of voice at all. It also fails the realistic chooser, because it hands irreversible instruments and no recourse to the complete and consistent agent of the theory rather than to the person the evidence describes, and the ordinary savers left holding TerraUSD are what that looks like. And it fails checkability, in the subtlest way of all. A public ledger is arguably the most checkable artefact ever built, verifiable by anyone with a laptop and alterable by no single hand, so the charge looks misplaced at first glance. But the ledger was the cypherpunks’ tool, not their answer. Theirs was exit, one person at a time, and leaving on those terms fails checkability in two ways. It constrains nobody, so there is nothing for a third party to check in any productive sense: it relieves you of a relation and produces no fact about the effect on what you left behind. And that party remains continually free to set and reset its objective as it pleases, having lost a customer and nothing else. The instruments do satisfy checkability, and that is what makes them the right parts for our project. The politics do not. Ultimately it is the politics that determine which instruments get built out and what they are pointed at. A culture organised around putting itself beyond reach does not spend twenty years building the machinery for holding a party to account, because that is what people do when they have decided to stay. So the platform critics and the cypherpunks scored as mirror images. The first was right about who dominates and wrong about where to intervene. The second was right about where to intervene and wrong about who dominates. Neither was foolish, and the union of what each got right very nearly supplies the pieces a working reply needs. Not quite, because that union still contains nothing that answers all masters or voice, and no means of detecting their violation. The cypherpunk experiment is, in the end, the empirical proof of the negative case this book has been making. A stack built on a single force — the Individual alone, holding its own property with no collective check and no theory of the commons — and then left ungoverned does not stay dispersed. On evidence, it re-concentrates, every time. The early promise of decentralisation curdled into a plutocracy of large token holders, then into platforms such as exchanges, ultimately mutating into a perverse game of utility theatre staged over protocols that quietly routed value to their creators. It was a steady drift towards the patchwork of private sovereignties Land foresaw. The fault was never in the primitives. It lay in shipping one force and calling it the whole. The experiment of a free, permissionless market has now been run at scale, and it failed. Something survives the wreck. From the platform critics: a true account of who dominates and how, and the proof that being right binds nothing. From the cypherpunks: the primitives themselves, and the civil-libertarian intuition that privacy and self-possession are the preconditions of an open society rather than a retreat from it. We can inherit the conviviality and leave the anarcho-capitalism where it fell. What the cypherpunks proved, against their own intentions, is that the substrate is the right place to write a constraint and that a single force alone will never hold one. That has been done once, in a single narrow domain, and the question worth asking is why it worked, and if the same conditions hold anywhere else.