--- title: Chapter Eleven — On Sovereignty: The Trinity book: Purposive Accelerationism author: '@jamie247' position: 17 of 24 canonical: https://whatispacc.org/ebook/#ch11 version: 1.1 language: en-GB licence: all rights reserved; quote with attribution --- # Chapter Eleven — On Sovereignty: The Trinity “So this is the formalist manifesto: that the US is just a corporation. It is not a mystic trust consigned to us by the generations.” — Curtis Yarvin, “A Formalist Manifesto”, 2007 When Curtis Yarvin above calls the United States “just a corporation”, he is making an argument about sovereignty, and that word has been doing quiet work throughout the book, carrying four conflicting meanings often read as one. The most cited belongs to the creed of the sovereign individual, for which it is a personal exit: the capacity of someone cognitively and economically able, and mobile enough, to take themselves and their assets beyond a jurisdiction’s claims. That is a private good, held against everyone and available in proportion to one’s means. For Land it is economic, or catallactic, a term he borrowed from Hayek for a kind of order that comes out of buying and selling, and on his account no person really holds it. It sits wherever capital happens to be flowing, in the service of an intelligence that arises from markets and machines that keeps making more of itself, and even the sovereign corporations of his patchwork only have it on loan. For Yarvin it cannot be split or shared, because someone always has the final say, and any system that claims to spread authority around is only hiding who that really is. He concludes that a state should be run like a company, under a proprietor who admits as much, and that divided government is just a pretence. This book instead uses the word politically, for a people’s standing to set the terms they live under and to change them, shared among them by design and never held by any one party. The other three sit closer to one another than any does to this one, because each treats sovereignty as property and differs only over whose name is on the deed, whether a person over their own affairs, the market or a ruler. On that view the only remedies are the ones any owner has, to sell, refuse to buy or leave, and nothing in that arrangement lets one holder compel an account from another. That is where the sovereign-individual tradition ends up, at Land’s formula of no voice and free exit. Its thinkers were not indifferent to being heard, but a conception built on ownership has nowhere to put an objection, so leaving is the only move it allows. Shared sovereignty still needs something to anchor it. Bitcoin held one invariant perfectly: scarcity, re-imposed on a substrate that would otherwise have dissolved it. That is the job of a governor, to pin a system to a chosen setting, and here it was elegantly the number twenty-one million. Nobody had to tell the protocol why. Its makers, working in the cypherpunk tradition, had already decided that a currency no state could debase was a form of liberty, and the cap carried that judgement. Elsewhere there is no such shortcut, no figure that settles the question for you. Scarcity protects money because a currency keeps its worth only while its supply is limited, but no cap shields a person’s identity, attention or thought, and subsumption is wearing all three down at once. So every new instrument needs someone to choose its target. Letting each domain pick its own does no harm in itself, and the aims may even balance one another, but only while somebody watches what they add up to. An identity scheme built for security demands that everyone be verified, while a payments network designed for compliance makes each purchase traceable, and each is reasonable on its own terms. Combined and unwatched, they would hand one power a dangerously complete picture of a life. That oversight calls for a standard to judge by, so before the method can travel, the people using it should decide what, in the end, their tools are all collectively for — the one value the governance stack exists to protect. Here I agree with Marc Andreessen that freedom comes first, since nothing else can be chosen, revised or answered for without it. I do not mean it as he usually does, though, asking only that nobody interferes with you, but rather in the sense of not living at the mercy of someone else’s will, what republican political theorists call non-domination. It is what the slave with a kind master lacked, as we saw when we mapped the three roads to totalisation earlier, since he could be left alone for years and still never be free. This is the spine of all that follows, and every governor proposed later answers to it. *** Everything from here rests on one assumption: that a person’s standing to weigh, choose and answer for their life is worth keeping for its own sake. Almost nobody this book addresses is likely to deny it. The priesthood certainly doesn’t. Andreessen’s own manifesto celebrates technology as the crowning work of human ambition, and when Zuckerberg answered Anthropic’s pacing proposal he rested his case on people refusing any agent that did not serve them. Whilst some strands of posthumanists may think our species is a net cost to the planet and would welcome its retirement for everyone else, the open question is what keeps agency alive. So why freedom? Health and prosperity are pretty important to most people, as is the absence of pain. A capable enough system might deliver all three in quantities no free society has yet sufficiently managed to achieve. If the trade-off is a longer, richer and easier existence in exchange for a life that is less free, most reasonable people would be inclined to give it serious consideration. The answer is that almost all the ways we run our lives together depend on some form of freedom, and losing it does more damage than first appears. We make promises and hold each other to them. We enter contracts. We consent. We praise, blame, forgive, apologise, resign, prosecute and pardon. Every one of these assumes that somebody made a choice and is there to formally answer for it. That connection between freedom and moral responsibility is why cognitive subsumption, once it puts us at the mercy of whoever runs the system, is more serious than the loss of a job or a skill. What it outsources is the act of weighing a decision itself, the faculty by which a person arrives at a view they can be held to. A country can shed a great many capabilities and still remain a society of authors, but it cannot lose this, because every way it can hold people to account relies on it. The trade-off we began with therefore undoes itself. Consent is on that list, so whoever bargains away the power to refuse has signed the last agreement open to them, and the children who will live with the outcome were never asked. A system that can predict your next choice can also make it for you, and it never has to interfere with you to do so. A recommendation engine knows what you will watch. A model drafts the reply you were about to write, and a pricing algorithm has worked out what you will pay. None of them touches you, yet each narrows what you do. That is why freedom here means something modest and practical. The shape of your life, for better or worse, should trace back to your own decisions, even though much of it is inherited, constrained or luck. At its core is the ability to weigh a judgement, to refuse it and, above all, to do what nobody saw coming. Progress depends on that last one, from the innovator who builds what the market never knew it needed to the scientist whose breakthrough no one planned. Friedrich Hayek believed a society that can anticipate and pre-empt its members’ choices is poorer in liberty and, for the same reason, in invention. Hannah Arendt’s two conditions, met earlier, describe what keeps it alive: plurality, the many who see the world from different places, and natality, each newcomer’s power to begin something unforeseen. Where both are in place, a society does not have to decide how anyone should live, because everyone stays free to hold, change and argue over their own idea of a good life. So all we need to ask of the governance stack is to keep choosing open, and in many hands. So if we agree freedom is a reasonable minimal reference value for society, the next question is which kind. Political philosophy uses the word for at least three things, and the priesthood often quietly trades on the ambiguity between them. The first is negative liberty, Berlin’s freedom from interference, met at the start of this book. Its virtue is that it is clear, testable and hard to abuse, because it can be easily felt and monitored. Its weakness is the one the republican tradition pressed: it only counts acts, so it cannot see a standing power not currently being abused. By its measure the slave with a kind master is free, as are the tenants of a benevolent landlord and the West Virginia miner of a century ago, paid in scrip he could spend only at the company store. By that logic, so are all of us whose reasoning AI quietly steers, since it has never once told us what to think. Every one of them, us included, lives on sufferance, yet because nobody is interfering, negative liberty calls them free. The modern version is quieter. A developer who has spent five years building a business on one laboratory’s model is a fair example. The lab has not cut off his access, raised his price or launched a product to compete with his. He is, by the negative test of freedom, free, and yet he must arrange his life around the knowledge that any of the three could come within a week with no recourse. Dependence of this kind is, in outward form, indistinguishable from liberty, which is what makes it hard to legislate against and easy to live inside without noticing, the way Ellul’s fish is unaware of the sea. The second is positive liberty, freedom as capacity, which the starving man on Berlin’s empty plain lacks and which Amartya Sen taught development economists to measure, as we saw at the start of this book. It is the sense in which two centuries of material progress widened human lives. Its weakness is the one Berlin warned against, with the Soviet state in mind: once freedom means having certain capacities, someone has to choose which ones matter, and whoever decides gets to force people towards their own particular form of liberation. Every regime that has claimed to know its subjects’ true interests better than they did, from the Jacobins to the Bolsheviks, used this argument, so any governor we propose has to be built so that it can never be bent to the same end. The third is not a philosopher’s account, but the one the priesthood’s doctrine runs on: freedom as optionality. More choices equals more freedom, and a system that multiplies what is on offer enlarges liberty by that fact alone. It has the great advantage of being measurable and yet the fatal defect of being silent on what matters most, because it never asks who drew up the list in front of you, to what end, or whether you could even find out. Our preferences are formed by what we have been offered, and nobody craves a fruit they have never tasted, so the hand that sets the options shapes the wants as well. By this test, twenty choices picked by someone whose interests run against yours, in a way you cannot see, count as more freedom than five you found for yourself. *** If freedom matters for these reasons, what would it take to secure it in practice? Here again the republican tradition supplies the test, since a master’s forbearance can be revoked at any moment and answers to nobody. An AI model that decides what you see, read and consider never lays a hand on you, but interference, in the republican account, has always meant more than just violence. Pettit lists manipulation among the forms it might take: fixing what a person gets to consider, or shaping their beliefs and desires by means that bypass their own reasoning. That is very nearly what an AI does as a matter of course when it stands between a population and what it reads, searches, drafts and decides. Choosing which considerations reach you is fixing the agenda. Shaping what you believe without ever appealing to your judgement is the manipulation Pettit meant, however good the intentions or accurate the answers. Neither requires a lie, or a desire to harm. But both still count as interference on Pettit’s terms, because the objective behind them sits inside a model you cannot inspect or contest, and so in this sense they are arbitrary. That is domination by definition rather than by analogy. The anarcho-capitalist reading treats negative liberty as if the interfering party were always the state, though Mill, Berlin and Pettit, as we saw with the cypherpunks, never did. Private power that answers to nobody is a failure of freedom even on negative liberty’s own terms, and it falls with full weight on the handful of firms that now hold the frontier of cognition, whose forbearance is the modern master’s. There is, however, also a failure that does not look like tyranny at all. A polity can keep every formal instrument of contestation and yet lose the ability to change the outcome. The vote still happens, the courts still sit, the press is not censored and the right of appeal survives. The consultation still runs and the responses are published. By every institutional measure, voice is intact. But if the options from which a people choose have already been narrowed by a power they cannot inspect, then what survives is the form without the substance. You are still free to object. There is simply nothing left to contest, because the question was settled before you were ever asked, by a process you had no way of seeing and no standing to challenge. Call this a hollow voice, a failure distinct from censorship, capture or apathy: the quiet replacement of the real thing with an imitation of it. Republican thought has never been satisfied by a remedy that lives only on paper. It requires that power can be called to account in practice. Two conditions follow from this for our work. The first is political, answerability, a channel through which power can be challenged. The second, capacity, is developmental, a population still able to form a view of its own and to imagine options nobody put in front of it. That is different from poverty, where the choices are few and everyone can see it. In a hollow voice they exist, or could, but someone else has taken them out of view, which makes it a question of power rather than of means. Neither of the two is sufficient alone, so they have to be held together. If keeping people able to think for themselves is the constitution’s job, then the charter needs a real measure of that ability, one precise enough that a country could fail it, and the seven tests proposed later in this book are built to be that. Answerability depends on something it cannot supply for itself. To contest what a system is optimising for you have to be able to evaluate it, and that ability is the one this book has shown being worn down. So the remedy presupposes the capacity whose loss it is meant to prevent. It works only as long as it is not needed, and stops the moment it becomes necessary. A right you can no longer exercise is a formality, worse than nothing because it makes an arrangement look constitutional while leaving it unconstrained. You can lay on a bus to take addicts to work, but you cannot give them the capacity to get on it, and a simpler complaints procedure, on its own, only lowers the bar while the skill to clear it keeps falling. That is why the two conditions stand together rather than one inside the other. Answerability rests on non-domination, which was never simply being left alone. A tax authority interferes with you constantly, yet where it can be questioned, appealed and removed you are not at its mercy. The point is not that a model should stay out of your thinking, since helping you reason is what it is for, but that whoever set its objective can be found and made to respond when you contest it. Capacity covers what answerability cannot reach. You can consent to a system, inspect it, challenge its objective and get a reply, every right intact, while it quietly makes you less able to do any of those things. Nobody is your master and no hand holds power over you, yet the faculty all of that depends on is being drawn down by a process you agreed to, at a rate first measured in 2025, and then only as the by-product of a study designed for something else. That is a failure of development rather than domination, and it needs its own condition because answerability cannot see it. Opting out does not help. The harm is positional, in the sense met with the refusers, since what your judgement is worth depends on everyone else’s, so someone who declines the tools while colleagues, rivals and children all use them simply falls behind. Because the damage is structural, the remedy should be constitutional rather than advice to the user. It should be a property of the system, testable from outside, and what it measures is dependency, in the sense the criteria gave it. Non-domination is a standing you hold in relation to power, and is maintained by being used. A right nobody exercises decays into a permission, because the party it constrains gradually discovers that nothing happens when it misbehaves, and the lesson compounds. Contestation is the system’s error signal. Where nobody objects, the party holding power stops learning where it has gone wrong, and its conduct drifts. None of this can be seen from inside, because the signal that would have revealed it has stopped running. A governance stack that guarantees rights to a population that never uses them has only automated the appearance of freedom, a piece of theatre crypto has already staged. Much of what has passed for on-chain democracy worked this way, with proposals put to token holders at short notice, turnout often in the low single digits and outcomes settled in advance by a handful of large wallets, so that everyone held a ballot and almost no one had a say. The lesson for those designing such systems is unglamorous: contestation has to be made cheap, ordinary and habitual rather than heroic. If challenging a decision requires a lawyer, several months and a high tolerance for personal cost, the right exists and the standing does not. If it requires a fork, a migration and the abandonment of everything you have accumulated, likewise. Most of what a good constitution does is lower the cost of using rights rather than declare them. A captured system rarely removes them. It raises that cost until they go unexercised and, in time, unremembered. *** Perhaps the atrophy of our own judgement, as we hand more of our thinking to machines, need not be terminal. A community that stepped back from these tools, as some religious groups have held apart from the modern world before, could keep the faculties the rest of us lose and return them to us later, as custodians, just as Latin Europe recovered the Greek inheritance from Arabic scholars who had preserved it. That holds on only one condition: that such a group can survive next to a far mightier neighbour, as the United States eventually came to tolerate the Mormons of Utah. But this requires goodwill, and indifference lasts only until the enclave stops seeming a curiosity and starts to look like a competitor or inconvenience. The dominant party need not be a state. It could be a superintelligence, and Elon Musk, introducing xAI in July 2023, mused that to such a mind humanity would be at best a curiosity, so our hope as a species lies in being interesting enough to keep around. Either way the enclave lives on the forbearance of a power it cannot contest, which is the position of the slave with a kind master again, who is left alone for now but could be interfered with at any moment without recourse. The tolerated community, whether the Mormons or humanity as a whole, finds itself in the same predicament. The enclave fails on its own terms as well, because the reservoir was never meant to be a library. Manuscripts are the easy part, since hard drives survive fires, empires and centuries of neglect. What the reservoir has to carry is agency itself, which is not an object to be stored and handed on later but a standing in relation to power. A community that lives by permission may keep every practice, craft, language and book intact and still lack the one thing recovery needs: the ability to set its own ends against a force that would rather it did not. The Arabic-speaking world did not preserve Aristotle at Latin Europe’s sufferance. It preserved him as the working property of civilisations setting their own course, which is why what came back was a living tradition of argument rather than a crate of documents. A dominated reservoir can pass on the memory of agency, the texts and rituals of self-government, but not the power to act on them, because it has none to give. That is terminal subsumption again: everything remains except the ability to use it, so any recovery would happen on terms someone else dictates. Such a community could stand undominated beside a far stronger neighbour only if both were bound by a rule neither could change alone, which is the work of a governor. Then there is the future in which no one is ruled at all, only no longer needed. There is no overseer, no boot on the neck and no command to obey. People are simply surplus to requirements, routed around by the market, priced out and made economically irrelevant by systems that bear them no ill will and give them no thought. Being unneeded is not the same as being mastered, but it becomes domination the moment your livelihood depends on a party you cannot contest, which is the position of the tolerated enclave. The harm lies less in losing the job than in having no recourse when it happens. Someone laid off by a firm in a functioning labour market — with savings, a union, benefits paid by rule rather than favour, an appeal to a body obliged to answer and other employers ready to hire — has suffered a serious loss but has not been dominated. Take those away, make the decision-maker unreachable, the review a form nobody reads and the support conditional on an authority beyond question, and the economic fact is unchanged while the political one is not. Gig workers know the difference well. In April 2021 a court in Amsterdam, the Netherlands, ordered Uber to reinstate six drivers — five in Britain and one in Portugal — whom it had dismissed by an automated process, and to pay them compensation. None had been told why, the notice said it was final, and asking again returned an identical message. Losing the work was an ordinary misfortune. What turned it into domination was that no one at home could reverse it, and the drivers had to look abroad for a court that could overrule Uber. The worker and the enclave have nothing else in common, but each depends on a power it cannot contest, and one account covers them both. There is a way of deciding between the worlds this book has described. The American philosopher John Rawls asked what principles people would choose for a society before knowing who they would be in it. Behind what he called a veil of ignorance, no one can tell whether they will be born rich or poor, gifted or ordinary, healthy or sick, so nobody can rig the rules in their own favour, and self-interest turns into fairness. Rawls hid the generation too, and drew from it a duty to save for those who come after. Stretch the horizon and you might arrive in the generation that builds the substrate, the one that inherits it, or the one three hundred years downstream living inside whatever it became. The answer that survives is thinner than equality and harder to argue with. Whoever you turn out to be, and whenever you arrive, there should be a power you can contest. It need not treat you well, which cannot be specified in advance. But it has to answer, to somebody, by a route that does not depend on its own goodwill, and that is what none of the three roads supplies. The present may bind the future, as every constitution does, in that each generation lives under inherited laws until it changes them by an agreed procedure, but nothing should be placed beyond amendment except the capacity condition, without which no later revision is possible. Whatever is set up now must leave those who come after able to challenge it, and a doctrine that writes them off in advance, on the strength of a view of deep time they had no part in choosing, fails before it starts. The laboratories already apply this test to one another, since none would accept a rival it had no way to check. Each eyes the others with suspicion over which might capture the state first, and Musk spent years in court with OpenAI over who should control the company, until a jury in May 2026 ruled that he had sued too late. Gambling on the average result rather than insuring against disaster makes sense only where the damage can be recovered, and here it cannot, because the worst outcome is the permanent loss of the standing to change it, and the faculty by which compensation would be claimed is itself gone. That turns a question of philosophy into one of engineering. If no party may ever hold a power beyond contest, the arrangement has to be constructed so that none can dominate. *** The cypherpunks failed not because their cryptography was wrong but because their politics shipped one part of the answer and called it the whole. They built tools that defended the individual against the state and left it exposed to everything else, to the private concentrations of the whale and the protocol owner. A single force, run on an ungoverned substrate, gathers power back into a few hands, and crypto’s first decade is all the proof needed. So the reference value cannot be the rule of any one force. It has to be an order in which no party is sovereign in itself, held in balance by the governor. Their technology went further than their politics. The forces that make a society free are old, and they have always pulled against one another. Wherever innovation has dictated the terms, the market has usually come out on top, which is the road to techno-capitalism this book has followed. What the stack the cypherpunks set in motion adds is the means to hold that tension rather than settle it in favour of one. For the first time each force can work without a gatekeeper able to capture it, and be limited by rules written into the systems it runs on. Self-custody removes the custodian, while a shared ledger binds every account, however large, to rules nobody can rewrite alone. Permissionless exchange lets anyone enter the field, while open code and the right to fork stop any platform from fencing it off. A decentralised autonomous organisation, or DAO, lets people pool funds across borders, while the freedom to leave, at a known price, means it cannot trap them. Earlier restraints, from the separated branches of a constitution to the competition authority, rested on institutions the powers they policed could co-opt or outlast. These primitives write the limit into the protocol, so the order no longer depends on the goodwill of the strongest. I will call these three forces — the Individual, the Market and the Collective — and the pull each exerts on the other two, the Cypherpunk Trinity. Each is generative in its own right. The Individual is the person with keys, able to hold property, identity and speech that no authority grants and none can revoke, with an exit always in their back pocket, and from that standing comes initiative, what Arendt called natality, the power to begin something nobody foresaw. The Market is the open field of voluntary exchange, where strangers transact directly, without an intermediary but under rules no single trader can rewrite, and it yields discovery, the knowledge no planner holds that Hayek saw prices gather. The Collective is the polity and the commons, the body that acts together and keeps what belongs to all of it, and it makes possible provision and voice, the things no one can secure alone. A society that kept a single force would not be purer, only poorer, and it would hand that survivor the power to decide what the rest were worth. Like Aristotle’s virtues, each is excellent in measure and destructive in excess. Left to itself every force presses outward until it totalises, singly or fused with another, and becomes one of the roads to the unfree world traced at the start of this book. *** Of the three, the Collective is the force most in need of definition, because it is the one the cypherpunks never articulated and the one nothing in this argument can do without. It runs as a spectrum, and the axis that orders it is not size or formality but the cost of leaving. A DAO written entirely in code has a constitution more elaborate than most statutes, and yet nobody mistakes it for a government. Walking out of a book club costs nothing. A union, a church or a DAO prices the exit — a notice period, a social cost, a forfeited stake — and the toll is known upfront and finite. Priced fairly, that toll does useful work, keeping members at the table until a quarrel is settled, provided the door stays open. A network state, a DAO that has bought land and seeks recognition, is cheap to join, but each acre, passport and treaty it gains makes the door heavier. A platform charges more than most people will knowingly pay, holding a livelihood hostage to network effects and accumulated data. And the nation-state demands the most a person has: borders, exit taxes, renunciation fees, the body itself the last thing to be carried out. On this axis the state is simply an association taken to its limit, the collective whose exits closed, and which then acquired territory and the monopoly on violence because nobody could get out to contest it. I have never been against the state as such. The enemy instead is the total state, sitting at the top of the exit-cost curve, where membership stops being a choice and becomes a condition. Its fully automated form is where the state road ends, and any association or platform can climb towards it. The firms met on the market road were collectives too, and climbed that gradient faster than any government, with no public accountability attached to the power they gathered on the way up. Cyberspace’s job on this axis is to drag the state back down the curve, so that wherever leaving is possible again, it has to compete, on taxes, liberties and law, like any other collective that wants to keep its members. The Individual runs along a spectrum too, and it is the harder one to see, because its axis is not the cost of leaving but how far a person’s power outruns what they are required to answer for. At one end stands the citizen, whose property and speech are theirs and whose choices touch mostly their own life, a position worth defending for its own sake. Next comes the sovereign individual, rich enough to carry wealth, skills and family anywhere. Here the problem is less a technique than the creed met on the road to exit. It rests on a fallacy this book has already taken apart. If the advance of technology obeys a law of physics, the system that delivers it is natural too, success within it is deserved, and those who achieve it become agents of progress and therefore of freedom itself. Those who hold this view do not claim that such a person is morally better than anyone else, only that the order of things has certified what they won, so nobody has standing to question how they use it. Marc Andreessen said in a March 2026 interview that he practises introspection as little as possible. The creed explains why he can afford not to: if present success is natural, future decisions need no examining, and whoever makes them stands above ordinary politics and the voice of the people they affect. Further along stands the oligarch, one of a few whose wealth and indispensability buy a private channel to ministers, though he still answers to rivals and needs the state he has captured. Taken to its extreme, that creed crowns the autarch. The autarch met earlier — whose name should not be confused with autarky, the self-sufficiency of a closed economy — would sign the sentence Peter Thiel wrote in 2009: “I no longer believe that freedom and democracy are compatible.” He treats the state as a supplier he may drop, and moves between two postures without settling in either. In the first he leaves to found a polity in his own image. Thiel funded the Seasteading Institute in 2008 and became a New Zealand citizen after twelve days in the country. Starlink’s terms of service declared Mars a free planet in 2020 beyond the authority of any Earth-based government, and Musk has floated loans for settlers who cannot afford the fare, to be repaid by work on arrival. In May 2025 the launch site at Boca Chica, in Texas, voted itself into a city called Starbase, its electorate largely SpaceX employees. In the second posture he stays and takes hold of the state from within. The office Musk ran in 2025 reached into federal payrolls, contracts and data systems, among them agencies that oversaw his own companies. The alternation is not inconsistency, because the possibility of leaving is what makes his presence count. What marks that end is not success, which this book is all for, but the loss of anyone able to say no. Command over others at any meaningful scale always needed other hands, staff to run it, soldiers to enforce it, workers to make it, and each of them could down tools, which capped the climb. Subsumption removes that limit, since a single owner directing swarms of agents needs nobody’s consent to act like a ministry. Here the kind master arrives in its most complete form: a ruler who may well be generous, answering to nobody, over people who can neither leave nor refuse. Hirschman explains why the alternation works. A credible threat to leave makes any protest louder, and its force turns on two quantities: how cheaply a person can get out, and how much the state loses by his going. For the richest both run high. Departure costs them a flight and a residency application. The treasury, meanwhile, forfeits a tax base, an employer and often a capability nobody else supplies, so the warning rarely needs saying aloud. Scholars after Charles Lindblom call this the structural power of capital, the discipline a government accepts from the mere possibility of disinvestment. For a teacher in Ohio or a farmer near Nairobi, in Kenya, the arithmetic runs the other way. Moving would cost either of them family, language and licence, the government would barely register the loss, and a grievance counts for little however just it is. Call the difference the exit premium, the influence a man is paid for keeping his door ajar. It holds only while somewhere will have him. Russia’s richest men bought the state’s best assets in the loans-for-shares auctions of 1995 and 1996, were told after 2000 to keep their fortunes and stay out of politics, and found in 2022 that sanctions had shut the London door from the far side. *** The balance is held pairwise, each force checking the next, on the principle James Madison set down in 1788 in The Federalist No. 51, of giving each power “the necessary constitutional means and personal motives to resist encroachments of the others”. Give the Individual a hard floor the Collective cannot breach — keys no authority can confiscate, a wallet no government can freeze, speech that nothing silences at the source — and you have the classical guard, the one the cypherpunks got right, against the totalitarian state and any majority or stronger party that would override a person. But a person cut off from exchange is not freer, only more isolated, which is why the Market checks the Individual: to transact is to need a counterparty and to accept the trade-offs that brings, so the openness that lets you exit also obliges you to bargain. That is the guard against autarky. It breaks down where one party’s bargaining power dwarfs the other’s, as a crypto whale’s does where trading is thin, and there the Collective has to step in with rules no single holder can rewrite. The Individual, in turn, checks the Market by being able to refuse it. A person who holds their own keys, records and identity can take their custom elsewhere, so no venue can price them as a captive. The Collective checks the Market by being a register that counts people in something other than price. In a commons or a polity a person’s standing comes from membership, contribution or the vote, so a cleaner weighs as much as their employer at the ballot box. A good held there is off the auction block not because a law forbids the sale but because the ledger it sits in has no column for a bid. That is what stops price becoming the measure of everything, and it is why the check has to be built into the commons itself, in its rules and its code, rather than decreed, since passing a law is the state road arriving to do the Collective’s work. A profession is such a register, keeping standards of conduct no balance sheet is meant to override, and the last decade has shown what happens when it gives way. The documents Frances Haugen took from Meta in 2021 showed its own researchers reporting that Instagram made body-image problems worse for one teenage girl in three, yet the product went on answering to the engagement figures it was designed to raise. The researchers had done what their discipline asked of them, and the company’s metric overruled it, which is the Market reaching into a register it was meant to stay out of. Co-operatives have kept such ledgers for nearly two centuries. The rules the Rochdale Pioneers adopted in England in 1844 gave each member one vote whatever they had put in, and paid capital a limited return rather than control, so that money could fund the society without owning it. Token-weighted governance is the opposite arrangement, and it is the one the first generation of crypto built: it handed people keys and trading rails and left the commons empty, because standing that can be bought is only the Market dressed as the Collective. This check fails worst when the Collective itself starts to measure its members in monetary value alone, since they cannot take their business elsewhere. A customer has one advantage a citizen lacks, since he can walk, and so the price stays tolerable. Treat a citizen as a customer and he gets the pricing without the ability to leave. In the Netherlands, between 2013 and 2019, the tax authority’s risk model for childcare benefit took a second nationality as an indicator of fraud, and some twenty-six thousand parents were ordered to repay everything they had received, in many cases more than they could ever raise, with no officer to answer for the score. Families lost their homes, more than a thousand children were taken into care, the data regulator found the processing unlawful and discriminatory in 2020, and the cabinet resigned over it in January 2021. None of those parents had been treated as a customer. Each had been priced as one, by the one supplier they could not switch. A newer arrangement adds a second party behind the counter: a government that answers its citizens through an AI licensed from a lab speaks in a voice it does not own, and can correct what it says only with the lab’s cooperation. Albania took that step in September 2025, naming Diella, a system built with Microsoft on OpenAI’s models, its minister for public procurement. What a market cannot see when it is handed a polity is the person who fetches nothing: the old, the disabled, the child, anyone in the year after a diagnosis. Citizenship covered them on the same terms as everyone else, because it was never a price. The state was the last register in which a person was owed something rather than quoted a figure, and when it reports in the market’s unit too, there is nowhere left to stand, and nothing to fork. What keeps the Collective in check is that people can walk away from it. An association stays legitimate only while exit costs something known and finite rather than everything, since the moment leaving becomes impossible the climb towards the terminal case has begun. A state that shoots those who try to flee, as East Germany’s border guards did at the Berlin Wall, may last for decades, but on this account it has forfeited any claim to obedience. The right to leave at a known price forces collectives to compete for their members the way firms court customers. Belonging is more than custom, though, because it brings duties to the group as well as claims on it, which is why departure is only half of what keeps the arrangement honest. The rest is the voice of the members who stay, which this book treats as a right in its own name. Call it fluid collectivism: association strong enough to hold a commons and check a market, kept so liquid that it can never close its hand around its members. The warrant for that restraint comes from Mill, who drew the line of legitimate interference at harm to others. The harms that matter at this scale are the structural ones — the monopoly, the debased currency, the destabilised climate — too slow, too aggregate and too opaque for any single chooser to price, so a market of fallible individuals cannot be left to find them on its own. The mechanics of priced exit, such as the escrowed stake that keeps a member of a DAO from fleeing a judgment, or the notice period that prevents a run, are for founders to design and the market to test. The hardest case is a polity nobody chose, as the so-called accidental Americans know. Born in the United States to parents who soon left, they are told as adults to file yearly with the IRS and to pay a fee to renounce a citizenship acquired before they could speak. Two invariants are set here. Exit is priced, never closed, so the collective is disciplined by leaving and the market by belonging. And membership is chosen rather than inherited, so anyone born into a group may walk away freely on coming of age, whatever a returning adult might later pay. The people who feel this first are individuals, long before any state does, as the clock showed. What they need is not a replacement for the country they live in but a layer alongside it, reaching across borders and in some respects stronger. It should complement the physical order where it can and compete with it where it must, so that a person or a firm can move part of their life into it as a hedge against wherever they happen to be. Three cases from earlier show what it has to hold. The refugee at a frontier, facing a state that would confiscate everything, needs a key no ministry can take. The community whose land registry sits on a firm’s server must be able to carry its record away. The worker whose judgement has been priced out of the loop wants a channel to object that asks nobody’s permission. People will cross into it in desperation at first, and later because there is nowhere better to go. What emerges is a fuller form of sovereignty than the one the cypherpunks named, an equilibrium that exists only while the three contest each other, which no single force can hand out or own because it is born from the tension between all of them. Non-domination is the requirement, and the Trinity is where it is won or lost, since each side is at once a threat to it and the only check available on its rivals. If no party may wield an uncheckable power over another, then no force may be sovereign, and each has to be held by the others rather than by a referee standing over them, who would then become the unanswerable authority the whole design exists to prevent. What the balance finally protects is Arendt’s plurality, the room for more than one kind of life and for new beginnings. So every new protocol entering the stack answers to non-domination, which here means the standing of the Trinity: none sovereign, none dissolved. *** Knowing what a governor should protect, a balance in which no force is sovereign, still leaves the harder question of how to build one. Scarcity, the invariant Bitcoin held, is only one species, the one that happened to fit money. The genus is incentive engineering: the deliberate arrangement of cost and reward so that holding to the invariant is what pays and breaking it is what hurts. An elegant protocol is a narrow, substrate-native governor, a small structure of hard-coded incentives that steers conduct towards a value that subsumption would otherwise carry away. Answerability and Arendt’s two conditions are invariants of this kind, and each will want a governor cut to its own shape. Such governors have to be designed because physics referees only some loops, as the book’s opening argued, and the domains that matter now are populated by parties who push back, where feedback is slow, gamed and reflexive. Agents will probe every rule for weaknesses, and the remedy is to pay them for each one they find. Crypto is, in aggregate, the largest standing hacker bounty ever run: every flaw in code securing millions is a prize, so the urge to crack the network is converted into what hardens it, far more effectively than in centralised systems where the attack surface becomes the security budget. A protocol that has survived that gauntlet carries a kind of proof no audited, permissioned alternative can claim, and no failure it can hide. Economists have studied the same problem for decades under the name of mechanism design, the work of setting a game’s terms so that participants pursuing their own ends arrive at the outcome intended, and Leonid Hurwicz, Eric Maskin and Roger Myerson shared the Nobel Prize in 2007 for founding it. Unlike the nudge that Richard Thaler and Cass Sunstein made famous, which works on choices from behind the scenes, a mechanism’s incentives are published, apply to everyone alike and can be inspected by anyone they bind. Satoshi Nakamoto compressed the idea into one sentence of the Bitcoin paper, reasoning that an attacker strong enough to rewrite the ledger “ought to find it more profitable to play by the rules”. What the method also has to answer is Goodhart’s law, met earlier, under which a number made into a target stops measuring what it was chosen for, and few aims are as visible as a hard-coded reward. The escape is to pay only for what can be verified in itself rather than for a proxy standing in for it, the way proof-of-work earns a block for computation any node can check. Where a mechanism rewards a figure instead, it is gamed as Goodhart predicted, and TerraUSD, whose collapse in May 2022 we have already seen, held its dollar peg with a minting rule that defended the number long after nothing stood behind it. A governor built this way has a particular character. It is narrow, enforcing one invariant and refusing to take on a second, because a protocol that reaches for every goal secures none and offers a wide surface to attack. It is composable, built to lock together with other governors into a stack rather than stand alone, though parts that are each sound can still fail in combination, so the whole needs a watcher of its own. And it is opt-in, joined because it is useful rather than mandated, which is what lets it run in a medium no authority can command. Narrowness is the source of durability, since the smaller and more specific a protocol’s task, the fewer the weak points an attacker can exploit and the longer it keeps its shape under pressure. Most governors of this kind turn out to be a family of proofs — proof-of-work was the first, and the pattern generalises into proofs of personhood, provenance, reserve and screening, each making one fact expensive to fake and therefore cheap to trust. What that cheapness is worth can be read off what trust costs today. Measuring a century of American national accounts, John Wallis and Douglass North found the transaction sector — the share of an economy spent not on making things but on verifying, enforcing and intermediating the promises around them — had grown from about a quarter of the output itself in 1870 to more than half by 1970. The toll steepens at every border. Sending a modest sum home still costs migrant workers more than 6 per cent on average, the unmet demand for trade finance runs into the trillions, and corporates surrender something on the order of a hundred billion dollars a year simply moving money across borders. A proof attacks the largest line item a modern economy carries. Take a nurse from Zimbabwe employed in Johannesburg, in South Africa, who sends a third of her wages home. That 6 per cent costs her roughly half a working day a month, for the length of her career. The line item carries the names of several hundred million people like her. None of these is made from nothing, and the work is one of assembly rather than discovery. Bitcoin itself invented none of its parts. Proof-of-work, the Merkle tree that lets a vast ledger be checked from a small fingerprint, public-key signatures, the gossip by which nodes pass messages to their neighbours — all of it lay around for years, used for other things, before somebody put the pieces together so that they held a chosen scarcity. The innovation was the arrangement, and that is the division of labour behind this response. I name the invariant a domain needs and point to the primitives, already lying around, that could be fused to secure it. The engineering, which pieces in which order at which parameters, belongs to founders and the market, because these are expensive problems and competition is how they get solved, which also spares the argument here any designs that will date. What keeps the naming disciplined is narrowness, applied as a test of elegance: each governor must enforce a single invariant. Any aim broad enough to need several has to be decomposed into narrow primitives that secure one each, never fused into a single protocol asked to carry them all. Each governor built to that test has three layers, and the first working example shows them apart. The purpose is a choice people made, when the cypherpunks decided that money no state could debase was a form of liberty. The governor is what holds the system to it. In Bitcoin that is the difficulty adjustment, which keeps a block arriving about every ten minutes and issuance on its schedule towards twenty-one million coins. The protocol is the software and rules thousands of nodes run to carry that out, and it could be written another way, as rival implementations show, without changing the purpose or the governor. The design leans towards the Individual and the Market and is unapologetic about it. That is allowed. No single governor has to keep the Trinity in balance, only the stack they form together, and watching over the whole is the meta-governor’s job. The method has already moved past currency. Ethereum extended the governor from issuing money to enforcing arbitrary agreements, smart contracts, written as code that executes itself when its conditions are met, so the constraint now reaches past what a thing is worth to what parties have promised each other. A general-purpose machine for contracts can cover far more ground than a single-purpose ledger. Built on top of it, the protocols loosely grouped as decentralised finance already run composite arrangements of these primitives, unsupervised, holding collateral and settling positions with no firm or clearing house behind them. Money, then contracts, then composite finance built from both: a real gradient of demonstrated capability rather than a promise of one. It is also an economy that anyone with an internet connection, anywhere, can partly exit into, joining cyberspace as a participant rather than a spectator. The governor, then, is a method that can scale. Incentive engineering should be able to reach any invariant that subsumption dissolves. Proof-of-personhood establishes that a real human stands behind an account without revealing their identity, and the primitive is foundational, because rights have always been built outward from the body — you own yourself, therefore your labour, and by extension your agency. Every other entitlement is scaffolding on that claim, and it is a footing that grows contested the moment cognition is delegated to machines. Content provenance, of media and increasingly of cognitive work itself, makes origin expensive to forge. Both are governors aimed at invariants the priesthood’s substrate is actively dissolving: the line between a person and a bot, and the difference separating a record from a fabrication. Both are nascent, and neither yet runs at anything like Bitcoin’s assurance. But they will each earn their place by showing that the method reaches past finance, towards the invariants taken up in the chapters ahead. Beyond money, contracts and the few early instances that reach past them, what we have are not yet protocols at scale but frontiers, contested domains: identity, authorship, the embodied world, and the polity itself. In each, subsumption is dissolving an invariant that nothing yet holds, or the governors that exist tilt dangerously towards a single force. The World Orb, the iris scanner from Sam Altman’s identity venture, is the standing case, its code and even its circuit boards now published, while the decision over which devices may certify a human still rests with the company that makes them. Who will control those four, and on whose terms, is the question of who writes the rules, and rules of that kind are what a constitution is for.