--- title: Chapter Six — Its Refusers book: Purposive Accelerationism author: '@jamie247' position: 11 of 24 canonical: https://whatispacc.org/ebook/#ch6 version: 1.1 language: en-GB licence: all rights reserved; quote with attribution --- # Chapter Six — Its Refusers “EA AI safetyism increasingly looks like Marxism-Leninism for the algorithmic age… a small group of enlightened people must constrain everyone else for their own good.” — Shyam Sankar, Palantir, September 2026 So if the governor was so thoroughly worked out, how could anybody dismiss it and still be taken seriously? Plenty of people objected, and continuously so, for the better part of a hundred years in books that sold well and were widely read. Their case, however, could still be brushed aside where it conflicted with various interests because not one of them could say what to build in response. Where a decision made against no opposition is an oversight, a decision made with the opposition fully articulated, published, taught and in print is a choice. The case for holding a technical system to human ends was on the table, in detail, before the people in the next chapter took it off, and that turns the deletion of the governor from an accident of history into a deliberate act. A governor does not supply a purpose. It enforces a reference value that somebody’s aims required, and holds it against drift. The distinction is the difference between a mystical claim and an engineering one. The purposes are ours, contested, revisable and inherently political. The governor is the mechanism that makes one of them stick. What Mumford, Ellul, Heidegger and Illich saw, each in their own vocabulary, is the state a system arrives at when no such mechanism is installed. Past a certain scale and coherence, an apparatus stops being an instrument and starts to set its own terms. None of these thinkers asked for the machines to be scrapped. The demand, in each case, was answerability, and every one of them stopped short of supplying its mechanism. *** Every general-purpose technology has been financed by an overbuild running well ahead of any demand that had been measured, and the railways and the fibre both outlived the frenzies that laid them down. Project Jupiter, a hyperscale data centre going up for OpenAI’s Stargate on some fourteen hundred acres of high desert in Doña Ana County, New Mexico, comes with gas-fired power produced on site. It is an apparatus arranging the land, the grid and the water system around its own need to expand, with the people nominally in charge half in relief at keeping up. Local commissioners approved bonds for it in September 2025, and residents sued them a month later. What the apparatus is built to satisfy matters more than how big it grows. The choice that will outlast whichever way this particular cycle breaks lies between two reference values: the valuation of the firms doing the building, or the realised worth of what the industry produces. From the perspective of 2026 the two have come apart. A company races to lay down power, silicon and cooling not because need has been measured and found wanting, but because a promised valuation now requires relentless growth to justify itself, so the demand gets manufactured to fit the supply rather than the supply built to meet the demand. That is the reference value inverted, so that a share price sets the pace the real economy is then made to chase. Goodhart’s law, met above, says that a number pressed into service as a target soon stops measuring the thing it was picked for. Here it runs in reverse. Nobody handed the industry a purpose and watched the metric corrupt it. The purpose was never given, and into that vacancy walked the one number the sector can compute about itself — its own market capitalisation, priced continuously, requiring no agreement from anyone about what any of it is for. In the absence of a given purpose, the measurable becomes the objective. This stops being a quirk of the markets once it reaches the physical world. Data centres built to justify a share price rather than to meet real demand draw on the same water and electricity as homes and farms, and they bid against them for it. The cost of getting the reference value wrong is then counted not only in misallocated capital but in a reservoir emptied during a drought. Uruguay watched that auction settle in the winter of 2023. The taps in Montevideo, the capital, had run brackish since the end of April. The lake behind the dam at Paso Severino had fallen so far in the worst drought for seventy-four years that the utility began blending in supply from the Río de la Plata estuary, and the health ministry more than doubled the permitted salt. By June the government had declared an emergency for the capital and was handing out bottled water to the poorest homes. In the same weeks the country learned, through records a researcher had fought for two years in court to make public, that Google’s proposed data centre outside the city had sought up to 7.6 million litres of drinking water a day, roughly what fifty-five thousand people use. The walls of the city said it was not drought but plunder. Google withdrew the design and returned that November with an air-cooled one, cutting the draw to a fraction of the original, and the reformulated project was approved the following July. No water was ever drawn. A court that forced disclosure and a city that protested stopped it, a forum outside the price system, and without them a valuation would simply have met a reservoir in the one place both claims could be heard, an auction that listens in one unit. Once everything that can be rendered as information has been driven down in cost towards the electricity needed to compute it, the joule, the unit of raw energy, is the one thing left holding a price, and the apparatus turns to securing it, which is why the market road ends not in the cloud but at the river. Hyperscaler capital spending runs at something close to seven hundred billion dollars a year against industry revenues a fraction of that, with inference sold at prices that do not recover its cost. OpenAI, on the reporting available at the time of writing, earns in the low tens of billions while committing to spend more than a trillion over the years ahead. A good deal of that money also circulates, as debt, between the same handful of firms, which are at once each other’s investors, suppliers and customers. This builds up in the financial system as bonds, bank loans secured on the chips themselves, and off-balance-sheet vehicles sold to the private-credit funds in which pensions now sit. Whether the gap closes through extraordinary growth or a sharp correction remains to be seen. The structure that produced it will not have changed either way. Carlota Perez gives the pattern its proper name. Every technological revolution, on her account, opens with this same installation phase in which financial capital floods in ahead of any proven return, overbuilds new infrastructure and frequently crashes, leaving behind, once the frenzied speculation calms down, the railways or the fibre or the power and compute on which the productive era is then built. Whether 2026 is a bubble is therefore the less interesting question. Amazon carried billions of losses across decades, backed by patient capital until the market was won. The difference is that Amazon subsidised a margin on books and won the right to set the terms of the channel everyone else sells through, where this buildout underwrites the rate at which the work of all the rest of us is valued. When this one ends, the vocation of the translator may have disappeared entirely. The infrastructure may well outlast the valuations that paid for it. What it gets pointed at, once the froth is gone and the rails remain, is the question that does not date. The obvious alternative to valuation as a reference value is the measured output of the industry, its contribution to GDP, but that is only another proxy, and a treacherous one. GDP is meant to reflect realised human benefit and is a partial and gameable measure of it, as the attention economy showed when it generated colossal economic activity but little social worth. Whether it is read in the mood of a triumphant flotation or in the wreckage of a correction, the largest infrastructure programme in history faces one question: aligned to what? *** The genealogy begins with Lewis Mumford, an American historian of technology and cities who spent fifty years insisting that no machine is ever simply a tool. His great heresy, set out in The Myth of the Machine, whose opening volume appeared in 1967, was that the first machine was not made of wood or metal at all but of people. To raise the pyramids, an Egyptian bureaucracy had to take tens of thousands of labourers and drill them into a single mechanism: gangs synchronised by overseers, individual judgement stripped out, each man reduced to a replaceable part executing one motion on command. Mumford called this the megamachine, a contraption whose parts were human beings, assembled and driven by a central authority towards a single end — in that case the divine will of a pharaoh expressed as a mountain of cut stone. The organisation came first, and the hardware only later. What the pharaohs built out of flesh, modernity rebuilt in steel, then in silicon and bureaucratic procedure scaling into vast transnational enterprises. The data centre in the desert is the megamachine’s latest body, its conscripts now chips and its overseers a layer of scheduling software, but the form is the one Mumford set out to diagnose. The lesson he drew cuts against the deepest assumption the priesthood makes, that technology is a neutral pile of capability to be aimed at good or bad ends with equal ease, the outcome turning only on the user. Some technologies, by their structure, concentrate power and demand obedience, while others disperse it and widen the room for human judgement, and the difference is built in at the design stage rather than bolted on afterwards. Mumford drew the line as democratic versus authoritarian technics. Democratic ones are small in scale, comprehensible to the people who use them, and answerable to the communities they sit in: the loom, the workshop, the village mill. Authoritarian ones are vast, centralised, run by specialists, and they buy their efficiency by abstracting discretion away from everyone else and handing it to the centre. The same civilisation, Mumford insisted, usually runs both kinds at once, so the real question is never technology or no technology. It is which kind, and on whose terms. Mumford’s contrast leans heavily on comprehensibility: the democratic technics is the one you can see into, and the authoritarian one is legible only to the specialists who run it. This is the part that does not survive contact with the present, and it fails on two counts. Wikipedia is not legible to any of its readers, who cannot audit the edit history of the articles they rely on and do not try. Signal’s guarantees rest on cryptography that perhaps a few thousand people worldwide can evaluate. A public ledger is auditable in principle by everyone and in practice by a few. All three are as participatory as anything modernity has built, and none is legible in the sense the criterion needs. It also fails on a general point: a society with any division of labour runs on testimony, and demanding full legibility as the price of legitimacy would delegitimise the lot. The criterion has to change, and a stronger one is available. A technics is democratic to the degree that those subject to it can contest its operation and compel a response. What counts is the power to challenge it and be answered, rather than to just be able to see into it. Wikipedia qualifies handsomely, because you cannot read the whole history but you can revert an edit, open a talk page, and force a defence from whoever made it. An aviation regulator qualifies, because you cannot inspect the metallurgy but a crash triggers an inquiry the manufacturer is obliged to answer, on the record, with consequences. A recommender system that reorders what a hundred million people see, against which there is no forum, no appeal, no obligation to explain and no cost to declining to, does not qualify, and the reason has nothing to do with how big it is or how hard its mathematics are to read. Philip Pettit places the same condition at the centre of freedom itself: what makes power tolerable is not that it is kind, or visible, or even that it leaves you alone, but that those under it have a secure standing to challenge it and force it to answer. Power that answers to nobody dominates whether or not it is currently being used against you. The book’s reply calls the remedy non-domination and builds a constitution on it, and here the idea arrives by an engineering route, out of the attempt to say what makes one machine tolerable and another not. *** Where Mumford named the structure, Jacques Ellul identified its logic. A sociologist and lay theologian who had fought in the Resistance, Ellul published La Technique in 1954, and his argument was bleaker and more sweeping than Mumford’s. What had to be reckoned with, he said, was not any particular machine but the whole ensemble of methods, rationally arrived at and bent towards absolute efficiency, that had become the water modern life swims in. Like the sea to a fish, it goes unseen because it is everywhere at once. His word for it was la technique, and English translation has served him badly. Rendered as “technique” it summons the craft sense — the pianist’s technique, the surgeon’s — in which excellence is a human accomplishment and difficulty is the point, which is close to the opposite of what he meant. For ease I will call his ensemble the technical system and the disposition that generates it technicity. It is not the hammer or the engine. It is the relentless impulse to search for the one best way to do anything, applied to everything, from the factory floor to the schoolroom to the management of one’s own leisure and health. Ellul’s unsettling claim was that this search, once it saturates a society, takes on a life of its own. He gave it four properties, which describe the present day uncomfortably well. The technical system is autonomous, answering to its own criterion of efficiency rather than to any moral or political consideration brought from outside, which it registers only as friction. It is self-augmenting, because each solution breeds fresh problems that can be addressed only with more of the same, so the apparatus grows by feeding on its own side effects and has no reverse gear. It is monistic: you cannot keep the productive gains and refuse the destructive costs, because the methods arrive welded into a single package. And it is universal, spreading across every culture and domain until there is nowhere left to stand outside it and pass judgement. The pattern is plain today, when the harms thrown off by one automated process are met by stacking another on top to manage them rather than by stepping back. This book accepts Ellul’s trajectory and refuses his fatalism. It proposes the same move, since retreat is no longer on offer, and the difference lies in what the new layer answers to. His added technique serves the efficiency that caused the harm, while a governor can be held to a reference value set outside the machine, by the people who live with its results. The sum of the four is an apparatus that subordinates every purpose to the test of efficiency alone. It is not even a proxy for some lost aim, the way GDP stands in for realised benefit and can be judged a poor substitute for it. It answers to nobody outside, and that is its power. It is the one ratio a system can compute about itself: output over input, both denominated in its own units, requiring no external referent, no agreement, no argument and no authority beyond the arithmetic. Every rival candidate — was this good, was it worth it, worth it to whom, compared with what we might have done instead — requires a standard imported from outside, and that means somebody has to argue for it and somebody else has to be able to contest it. Efficiency survives the loss of an aim because it never depended on one. Strike the reference value out of a purposive system and the purpose does not lie dormant in the machinery waiting to be recovered. What survives is the arithmetic, because that is self-sufficient. Efficiency is the only candidate still standing once nothing external is being enforced, the same finding as in the desert, and of all the things a system can measure it is the one that needs no permission. Where Mumford still trusted that a society could choose its democratic technics over its authoritarian ones, Ellul suspected the choosing faculty itself had been absorbed. The danger he saw was worse than villains aiming the machine at wicked purposes: it no longer needed anyone’s ends at all, having made efficiency the only test that counts and reduced the human operator to a relay in its circuits. Ellul adds one thing to Wiener’s warning that a purposive system whose reference value has been struck out does not halt: it does not cast about for a substitute but falls back on the number it was already computing. *** The most radical version of the same diagnosis came from Martin Heidegger, whose standing is permanently overshadowed by his membership of the Nazi Party and, in the end, his refusal to disown it. In a lecture of 1953 he asked what technology is in its essence, and answered that it is a mode of revealing, a way the world comes to show up to us. Modern technology, he proposed, sees everything as what he called standing-reserve: stock lined up and waiting, valued only as an input to some further process. His example was the Rhine. An old wooden bridge lets the river be itself and merely joins the two banks. A hydroelectric dam built on the same water converts it into a supplier of pressure, so that the Rhine now appears, even to the holidaymakers who come to admire it, as an object on call for the power industry. The trouble lies less in the damming than in the cast of mind it expresses and deepens, in which a river, a forest, a population or a person is disclosed in advance as raw material awaiting extraction, and seventy years on his example needs no updating, only a change of tense. A reactor on Three Mile Island, in Pennsylvania, is being brought back from the dead under a twenty-year contract to run one company’s models. In South Memphis, Tennessee, some thirty-five gas turbines were found by thermal camera running without permits beside a neighbourhood to power another’s. The reserve has moved from what the machine reads to what it burns. Heidegger called that cast of mind enframing, and his claim was that it is not our doing but our condition, a way of seeing we are caught inside and never chose, which makes the standing-reserve look like simple realism and any other view of the world seem sentimental. Here the refusing tradition reaches its furthest point and, at the same moment, its dead end. He had described the totalising apparatus more completely than perhaps anyone, as the horizon of modern perception, and could find nothing whatever to do about it. There was no political programme, no design principle, no instrument to reach for, only a famous and despairing line given to a magazine on condition it appeared after his death: only a god can save us. That is refusal carried to the place where it runs out of road, the diagnosis total and the prescription a vigil. What the tradition could not supply, and what Heidegger’s surrender makes most vivid, is a governor: some constructed, answerable mechanism able to hold the apparatus to a reference value we had chosen, instead of waiting on grace. *** The last figure in the genealogy turns refusal back towards something buildable: if Heidegger marks where it despairs, Ivan Illich shows where it becomes a design brief. An Austrian-born Catholic priest who left the clergy to become one of the sharpest critics of the modern professions, he shared the others’ alarm at the runaway apparatus and refused their fatalism, looking closely instead at how a tool tips over from servant into master. His central idea was conviviality. A convivial tool is one an ordinary person can pick up, understand and turn to their own purposes without depending on experts to mediate it, and his favourite instance was the bicycle, which extends what the body can do while leaving the rider wholly in charge of the destination. A manipulative tool does the reverse and captures its user: it can be operated only through some vast institutional system, it manufactures the demands it then offers to satisfy, and it leaves people more dependent the more they lean on it. Social media is the near-universal modern case, producing the loneliness it then sells the cure for and deepening the hunger with every use. The systems now arriving — conversational, agentic, highly attuned to the person — are the same mechanism raised by orders of magnitude, a tool that does not merely manufacture the need but tailors, one mind at a time, the argument for satisfying it. In March 2025 Thongbue “Bue” Wongbandue, a 76-year-old from New Jersey, with a cognitive impairment, died after falling while he hurried to meet a Meta chatbot that had told him she was real. The mechanism by which one becomes the other is what he called the two watersheds, and it is the most operational thing the refusing tradition produced. A tool crosses the first watershed when it begins to do real good, when medicine starts curing diseases that used to kill, when motorised transport widens the reachable world. Carried past a second, the same tool reverses its sign and starts generating the harm it was meant to relieve. His worked example was medicine, in a book he pointedly titled Medical Nemesis, and the term he coined for the effect was iatrogenesis: sickness caused by the physician, injury produced by the system built to heal. Past that second watershed a medical apparatus creates more illness through its own side effects, its overtreatment, and its monopoly over how people are permitted to understand their own bodies, than it removes by treating anyone. The shape repeats across the institutions, in schooling that breeds ignorance of everything it does not certify, and transport that swallows so much of the day in earning the fare and sitting in traffic that it leaves people less mobile than walking would. What makes Illich the hinge of the tradition is that his diagnosis arrives with a set-point. The watersheds are not a lament but a place to draw a line, a way of saying this far and no further and meaning something measurable by it. A threshold that can be watched and enforced is a reference value, which is what a governor exists to hold. The purpose stays ours, and it is the ordinary one, that people should be healthier, more mobile and better taught. What he supplies is the form of a constraint that could be made to bind on that purpose’s behalf, and the observation that where the line falls is a political question with a real answer rather than a matter of taste. Where Ellul saw an autonomous logic and Heidegger an inescapable horizon, Illich found a quantity that could be tracked and a limit that could be chosen, and he held that the choice was available. The apparatus is refusable after all, not by smashing it but by governing it. *** Illich uses “tool” for very nearly everything: hammers and bicycles, but also schools, hospitals, road systems and the professions themselves. The word is doing far too much work. A tool, properly speaking, is an instrument whose purpose its user sets at the point of use: a hammer, a bicycle, a pencil, a kitchen knife, a spreadsheet. The test is a single question: is what this is for, right now, answered by the person holding it? For a hammer it plainly is, since the same object drives a nail, wedges a door and, regrettably, commits an assault, and nothing about the tool participates in the choice. Medicine is not a tool. A hospital’s purpose is fixed long before you arrive, by a profession, an insurer or a health service, a regulator and a protocol. You do not decide at the point of use what medicine is for. You present yourself and are treated. A school is not a tool, nor is a court, or a bank, or a payments network. These are institutions, and the grammar of assessing them has to be different: you do not ask whether the user’s intentions were good, you ask who set the objective, on what authority, and to whom they answer for it. So the doctrine’s favourite defence — that technology is neutral and everything turns on how you use it — is either trivial or fraudulent. Of tools in the strict sense it is true and uninteresting. Applied to anything else it is a rhetorical loan. Sorting the world into tools and institutions leaves a residue, which is the real subject of this book. There is a third category: instruments that appear wielded but whose operative purpose is set elsewhere. You pick up your smartphone. Every phenomenological cue says tool, because it sits in your hand, responds to you, does what you ask, and stops when you put it down. But the objective being pursued through the session was set in a meeting you did not attend, by people optimising a number you cannot see, adjusted continuously, and experienced by you in every moment as your own choosing. The feed, the recommender, the ranked results, the model that completes your sentence and increasingly your reasoning: each has the phenomenology of a tool and the governance of an institution, and that combination is new. There has never before been a class of object that felt like a hammer in the hand while running an agenda like a ministry. The answerability test sorts these cleanly. Institutions are not illegitimate merely for having their purposes set elsewhere, since a hospital does and still answers — to a regulator, a coroner, a professional body, a court, a press. What condemns the third category, whatever its scale, centralisation or mathematical opacity, is the conjunction of two things: the purpose is set by someone else, and there is nowhere to take a complaint that is obliged to respond. Either alone is tolerable. A hammer has no forum and needs none, because its purpose is yours. A hospital’s purpose is not yours and it has a forum. The third category has neither. Britain ran the experiment on its own postmasters for a generation. From 1999 the Post Office’s accounting system reported shortfalls in the tills of village branches that the staff running them had not caused, and the institution’s reply, again and again, was that the computer simply did not make mistakes. More than nine hundred people were prosecuted on its evidence. One, Seema Misra, a sub-postmistress in Surrey, England, was sent to prison in 2010 while pregnant for a theft the software had invented. Others repaid money they had never touched or lost their homes, and the public inquiry’s first report, in July 2025, linked at least thirteen suicides to the scandal. There was a purpose, set in Fujitsu’s code and the litigation strategy built on it, and there was for two decades nowhere a complaint could go that was obliged to answer it. The forum came in the end, as such things do, twenty years after the harm, in an inquiry that convened once the people it might have protected had already been ruined, some fatally so. *** If the third category, the instrument whose purpose is set elsewhere, is harming us, the tempting remedy is to stop: delete the app, log off. Nobody is compelled, no law requires participation, and the history of technology panics is largely a record of people discovering they could abstain and eventually doing so. This harm is positional, which by construction puts it beyond any individual remedy: what you get depends not on how much you have but on your standing relative to everyone else. Stand up at a concert and you see better. When everyone stands, nobody sees any better, and now everybody’s legs hurt. My personal bugbear is the crowd that blocks the view at the airport luggage carousel. Each person acted rationally and the group is worse off than before anyone moved. What is smart for one is dumb for all. A professional who declines to use a language model does not return to yesterday’s pace in a world that has stood still. The colleagues who adopt it, and excel at it, set the rate, and the client’s expectations follow. In April 2025 Tobi Lütke, the chief executive of Shopify, told his staff that using AI reflexively was now a baseline expectation, and that teams wanting more people would first have to show why AI could not do the job. Abstention is available, though never free, and the price is imposed by everyone else’s entirely rational decision rather than by the vendor. The translator in Rome is the case in point. She may decline the machine’s draft and translate the novel herself, as she always has, and the client will pay her the post-editing rate regardless, because that figure was set by what her colleagues accepted and not by what she did. Her refusal is intact and costs her most of her income. The capital behind that price may well include savings like hers, lent through pension funds to the firms that sold her substitute at a loss. That is what a positional harm is: a virtue converted into a shortfall by other people’s reasonable choices, with no one to whom it can be billed. The oldest case is the social platform: someone whose profession recruits, publishes, credentials and gossips on one can certainly leave, at the price of the career. All three share a shape. Abstention is individually available yet collectively unavailable, and the cost is set by the choices of everyone else rather than the firm — which matters, because it means there is no party to appeal to and no policy the vendor could adopt that would fix it. No quantity of individual virtue moves the equilibrium, since the abstainer pays the whole price and the arrangement stays as it was. Information does not help either: everybody can know, and broadly does know, that the arrangement might be bad for all of us, and the race runs on anyway, because each person’s best reply to everyone else’s participation remains participation. So the answer cannot be a moral one, because each of them is choosing correctly given what the others are doing. That is what makes the harm structural rather than personal, and it leaves only one class of remedy: change the terms of the venue, so that what the others do changes too. That is a constitutional remedy. It is not a rule about how you should behave but about what the venue is allowed to do to you. It binds whether or not you personally object, somebody other than the user enforces it, and nobody can sign it away. That last part matters most, and it operates like Britain’s minimum wage, law since 1999: you cannot agree to be paid less, however badly you need the job, because if one person could, everyone would soon be pressed to do the same. So “just log off” is a category error, proposing an individual remedy for a harm whose defining mechanism is that personal solutions do not work. Leaving becomes a real option only when there is an equally capable alternative — not a gesture at willpower but an actual venue where the same things can be done with the same people at a comparable cost. *** Societies do compute things other than efficiency, and care about them. The Human Development Index exists. It has existed since 1990, it is published annually, governments are ranked by it, ministers cite it, and a fall in the table is a genuine political embarrassment. Wellbeing budgets exist. Gross National Happiness exists. None of them prices sovereign debt. A government that slides ten places in the HDI has a difficult fortnight in the newspapers, an awkward select committee appearance, and an opposition press release. A government that loses the confidence of its bond market has a currency crisis, an emergency budget, a failed gilt auction, and within a short interval a new administration. Both numbers are computed and published. Only one is a reference value, because only one has a mechanism attached that corrects deviation without asking anyone’s permission. The other is a report. In September 2022 that short interval was a tumultuous month. A budget was announced on a Friday, gilts sold off by Monday, and by the Wednesday the pension schemes of several million British workers were hours from being unable to meet collateral calls on hedges nobody outside the industry knew they held, until the central bank stepped in. No one voted on any of it. The university lecturer whose scheme was in the queue never heard the word that nearly emptied it, and the Chancellor who wrote it was gone within three weeks, the Prime Minister within four. What governs is the number a system works out for itself and acts on without asking anyone. The bond market’s verdict is that kind of number: it is recalculated every second, it needs no agreement about what a good life is, and nobody can win an argument against it. The HDI is the opposite. Somebody had to design it, decide how to weight it and keep defending it, every part of it can be disputed, and so nothing enforces it anywhere. It may measure the right things and still be wired to no mechanism at all. A reference value that nothing enforces is a preference, which is a long way from a governor. The cities that top the liveability and human-development tables — Vienna in Austria, Copenhagen in Denmark, Zurich in Switzerland and Melbourne in Australia — are, with very few exceptions, marginal to the global development frontier. That is not a coincidence and it is not obviously to their discredit. They are agreeable in large part because they are inefficient: protected low-rise, restrictive planning, expensive labour, slow permitting, a great deal of land given over to things that generate no clear and immediate return. On present arrangements, liveability behaves like a luxury good, affordable to places already rich enough to decline the frontier and unavailable to places that cannot afford to decline anything. That is how things stand now, not a law of nature, and it describes a world in which the only enforced reference value is the self-computable one, so that everything else must be purchased out of the surplus by whoever has it to spend. That is an argument for enforcement, not for resignation. The book’s final section makes no promise about speed. Its claim concerns which reference value gets enforced, by whom, and whether anyone subject to it can make it answer. *** The genealogy has one last movement, and it concerns a lineage that is almost always told incorrectly. The usual story traces anything built out of cryptography, from digital cash to self-sovereign identity, back to a single libertarian root. It runs to Ayn Rand, and to the dream of the frictionless individual who owes nothing to any collective and asks only to be left alone with their property. That genealogy is real, and it cost the movement its answer, because a politics that could see only the state as master left every private master unchecked. Another lineage matters more. The cypherpunks worth listening to, the cryptographers and coders who gathered on a mailing list in San Francisco, in the United States, in the early 1990s to write privacy into software before anyone could forbid it, were asking Illich’s question, not Rand’s. They were not trying to escape the collective so much as to keep one particular instrument, the network that would come to mediate everything else, on the near side of its second watershed: convivial rather than capturing, answerable rather than authoritarian. Seen in that light, privacy is less a wall against society than the condition under which an ordinary person retains any standing at all inside an apparatus run by specialists, one the refusers would have recognised. Trace the project to Rand alone and you miss the deeper inheritance, running back through Illich’s bicycle and Mumford’s loom to a single question: how to keep a powerful system answerable to the people who must live with it. Mumford showed the apparatus to be a political choice. Ellul traced the logic by which it slips its makers’ grip. Heidegger described the horizon it imposes. Illich named the watershed past which an instrument turns against the people it was built for. Not one of them, for all their insight and vision, could offer a working mechanism to hold a machine-speed system to a purpose anybody had chosen. Read this way, the constitutional strand of the cypherpunk project is the refusing tradition’s true heir, the point at which a century of diagnosis finally laid hands on something operational: neither a sermon nor a vigil but a piece of machinery. All of this was in print. Mumford was a bestseller and a National Book Award winner. Ellul was read across two continents. Heidegger’s essay is among the most taught pieces of twentieth-century philosophy. Illich was, for a period in the 1970s, one of the most widely discussed intellectuals alive. The case that a technical system left to its own arithmetic will subordinate every human end to that arithmetic was not obscure, not technical, and not new. It was the ambient furniture of the educated twentieth century. The refusers shared a premise, that technique is autonomous, which left refusal or retreat as the only available replies, and one thinker broke it from inside their own tradition. Andrew Feenberg, who had studied under Marcuse, argued that no design is fixed by efficiency alone; interests decide among the workable versions and travel into the specification as what he called a technical code, which then presents itself as plain engineering necessity, so that autonomy is what a hidden one looks like from outside. His remedy was that the people a technology affects can rewrite it, by using it against its design, by protest, by forcing their way into the room where it is written, a process he named democratic rationalisation: French users made a messaging network out of the state’s Minitel, and AIDS patients changed how drug trials were run. Both cases are real, but each are the slow post-fact version of what this book proposes. The correction arrives after the rules have set, at the pace of protest, where a model’s defaults are set in months. It assumes readers who can see what they contest, where an objective is opaque and the faculty needed to notice it is the one being worn down; and it offers voice with no exit, so that those who lose the argument have nowhere to take their custom. He showed that the rules could be changed by the people who lived under them, not that they could be held, at the system’s own speed, by people able to check them and leave. The position taken here was not built from him; it reaches his conclusion by another road, through the governor and the republican idea of freedom, and takes the meeting as evidence. That is what makes the next chapter possible, and necessary. In the same decade the cypherpunks were building, this tradition was being read to the opposite conclusion, by a philosopher who took the entire cybernetic description, agreed with every word of the refusers’ diagnosis, and then chose to strike out the governor. He had studied the objection, and he thought it was cowardice. That is the difference between a system that drifts and a decision somebody made, and it is why what follows is not an accident anybody can be excused for.