Chapter 4 of 13

Organising the Scattered

About 26 minutes to read

Labour Day in Nairobi

On 1 May 2023, International Workers' Day, somewhere between one hundred and fifty and two hundred people met in Nairobi and voted to form a trade union. Most of them were young. Many were Kenyan; others had come to Nairobi from elsewhere in Africa, and between them they worked in some fourteen African languages. None of them worked in a factory. What they had in common was a job that most of the world did not know existed.

They were content moderators. Employed by outsourcing firms such as Sama and Majorel, they spent their working days looking at what users of Facebook, TikTok and other platforms had posted — beheadings, child abuse, suicides, hate speech — and deciding, in seconds, what had to be removed. Some had also labelled the violent and abusive text on which OpenAI trained ChatGPT to recognise what it should refuse to say. Chapter 13 of the first book described this work as the Turk in the machine: the hidden people who make the automated systems of the data farm look automated. In February 2022 the journalist Billy Perrigo reported in Time that some of the Facebook moderators in Nairobi were paid as little as about one and a half dollars an hour. Many of them described lasting psychological harm.

The meeting had a history. In 2019 a South African moderator at Sama's Nairobi office, Daniel Motaung, had tried to organise his colleagues to press for better pay and mental health care, and was dismissed. In 2022 he sued Sama and Meta, Facebook's parent company, in Kenya. Meta argued that Kenyan courts had no jurisdiction over a company with no office there; in February 2023 Kenya's Employment and Labour Relations Court ruled that it could be sued. By then Sama had ended its moderation work for Facebook, some two hundred and sixty moderators had lost their jobs, and nearly two hundred of them had gone to court as well. The Labour Day vote came out of those cases.

Two years later, in April 2025, moderators from nine countries — among them Kenya, Ghana, Turkey, Poland, Colombia, Portugal, Morocco, Tunisia and the Philippines — met in Nairobi again and launched what the global union federation UNI described as the first global trade union alliance of content moderators. Its aim was to bargain not with the outsourcing firms that employed the moderators but with the technology companies at the top of the chain.

It is far too soon to say whether these moderators will win much. Their work can be moved from one country to another with a change of contract, and the companies that profit from it are among the richest in history. But their story is a compact version of the question this chapter asks. The first book found that organisation was the force that most clearly distinguished the breakouts that worked from those that did not. The 99 of the twenty-first century are scattered: across contracts, across workplaces, across borders, and increasingly managed by software rather than by people. Can they be organised? And what have the new forms of organisation actually won?

What the factory gave

The first book told the story of the factory as one of cruelty. It also recorded something the factory gave the 99 almost by accident. By gathering thousands of workers under one roof, on one timetable, under one owner, it made them visible to one another. They could see that their grievances were shared, and they could stop the machines together. The mill-owners spent a century trying to undo that accident by dividing the workforce, as Chapter 9 of the first book described. But the concentration of workers was the precondition of everything the industrial breakouts achieved: the unions, the shorter working day, the vote and, eventually, the welfare state.

Over the past half-century, that concentration has been taken apart. The economist David Weil gave the process a name in The Fissured Workplace (2014). Large companies, he showed, had steadily pushed work out of their own payrolls and into chains of subcontractors, franchisees, staffing agencies and self-employed contractors. The cleaner in a hotel, the guard at a bank, the driver delivering a parcel and the worker in a warehouse might all wear the brand of a famous company while being employed, legally, by someone else — or by no one at all. Each link was squeezed by the one above it, and the worker at the bottom had no employer with the money or power to bargain.

The result can be seen in the numbers. According to the OECD, the share of workers in its member countries who belonged to a trade union fell from around thirty per cent in 1985 to around sixteen per cent in the late 2010s. In the United States, the share of wage and salary workers who were union members fell from 20.1 per cent in 1983, the first year for which comparable figures exist, to 9.9 per cent in 2024. In the private sector it was 5.9 per cent. As the first book noted, the rise of American unions explains a significant part of the great compression of inequality in the 1930s and 1940s, and their decline part of the reversal.

There is a puzzle in these figures. Unions have rarely been more popular. In Gallup's annual survey, American approval of labour unions has stood between roughly two-thirds and seven in ten every year since 2021; in 2022 it reached seventy-one per cent, the highest since 1965. The 99 have not stopped wanting to organise. Something has made it much harder to do.

Three kinds of scattering

The workers of the twenty-first century are scattered in at least three ways, and each poses a different problem for organisation.

The first is scattering by contract. A growing share of the world's workers are not, in law, employees at all. The International Labour Organization counted the number of digital labour platforms — the companies that match workers to tasks through an app or a website — and found that it had grown from around one hundred and forty in 2010 to more than seven hundred and seventy in 2020. In 2023 the World Bank estimated that there were between 154 and 435 million online gig workers in the world, between about four and twelve per cent of the global labour force, with demand growing fastest in poorer countries. The estimates are wide and disputed, but they describe a workforce that is told it consists of entrepreneurs. A self-employed contractor, in most legal systems, has no right to a minimum wage or paid leave and — crucially for this chapter — no protected right to bargain collectively. Under competition law, self-employed workers who agree on a common price can even be treated as a cartel.

The second is scattering by place. Much of the work that, as Chapter 3 showed, stands behind the wall of what machines cannot do is done in places where workers rarely meet. Care workers work in private homes. Drivers and delivery riders spend their shifts alone on the road. Online freelancers work from their bedrooms. The International Labour Organization estimated that there were about seventy-six million domestic workers in the world in 2019 — cooks, cleaners, nannies, carers — three-quarters of them women and more than four in five informally employed. Each of them works for a different household. There is no factory gate at which to hand out leaflets and no canteen in which to hold a meeting.

The third is scattering by border. The content moderators of Nairobi are one example among many. The garment workers who sew clothes for European and American brands work in Bangladesh, Vietnam and Ethiopia; the data labellers who train artificial intelligence systems work in Kenya, the Philippines and Venezuela; the care workers of the rich world come from the Philippines, India, Ukraine and Latin America, often on visas that tie them to a single employer. Capital moves across borders with a keystroke. Workers are held back by passports. A strike in one country can be met by moving the work to another — a threat that is often more effective than any actual move.

The factory owners of the first book mixed nationalities on the shop floor so that workers would find it harder to act together. The global economy does this on a planetary scale, without anyone needing to plan it.

The foreman in the phone

There is a fourth feature of the new work, which is not so much a scattering as a replacement. The workers of the gig economy, and increasingly of warehouses, call centres and offices, are managed by software.

The sociologist Alex Rosenblat, who spent four years studying Uber's drivers for Uberland (2018), described how the app did what a foreman once did: it assigned work, set prices, nudged drivers towards particular places and times, rated them, and could deactivate them — in effect, fire them — without a conversation with anyone. Drivers were told they were their own bosses. In practice they were managed more closely than most employees, by a boss they could not see or argue with. Researchers call this "algorithmic management".

In Amazon's warehouses, as the first book described, the same principle governs the pace of physical work: scanners and software measure each worker's "rate" and "time off task". In 2021 California's AB 701 required large warehouse employers to tell workers their quotas and forbade quotas that prevent breaks or trips to the toilet — one of the first laws anywhere to regulate a quota set by an algorithm.

Some of the most revealing evidence about algorithmic management has come from court cases in the Netherlands, where Uber's international operations are based. Drivers from Britain and Portugal, supported by a small organisation called Worker Info Exchange and the App Drivers and Couriers Union, used the European Union's data protection law to demand that Uber and the Indian ride-hailing company Ola explain how automated systems had made decisions about them, including decisions to deactivate drivers for suspected fraud. In April 2023 the Amsterdam Court of Appeal largely found in the drivers' favour. Uber had argued that a human being reviewed the automated decisions before drivers were removed. The court found that this review was "not much more than a purely symbolic act", so the decisions were in substance made by the machine, and the drivers were entitled to an explanation of how.

That phrase matters for this book, and it will return in Chapter 10. The third element of the Federation hypothesis is that AI should replace the shepherds, so that no managing class can grow into an owning class. Algorithmic management is the first large-scale experiment in AI as a shepherd. On the evidence so far, the experiment has not freed the herd. It has given the owners a shepherd that never sleeps, never sympathises with the workers, cannot be persuaded or embarrassed, and cannot be held to account — and it has often kept a human in the loop only as a symbol. A machine that replaces the managers does not by itself remove the managing function. It moves it to whoever controls the machine.

Winning the vote, losing the contract

The scattering has not stopped workers from organising. The question is what the new organising has won, and the record is mixed in revealing ways.

Begin with the most traditional form: a union of employees in a workplace, winning an election. In April 2022, as the first book recorded, the workers at Amazon's JFK8 warehouse on Staten Island voted by 2,654 to 2,131 to be represented by the Amazon Labor Union, a new and independent union led by Christian Smalls, a former Amazon assistant manager who had been dismissed — for breaking quarantine rules, the company said — after organising a walkout over safety in the first weeks of the pandemic. It was the first successful union election at an Amazon facility in the United States, and it was won not by an established union but by the warehouse's own workers, organising at a bus stop outside the gate.

What followed is a lesson in the difference between winning recognition and winning a contract. Amazon challenged the result. The National Labor Relations Board certified the union in January 2023, but the company refused to bargain and challenged the constitutionality of the Board itself in the courts. In June 2024 the Amazon Labor Union affiliated with the much larger International Brotherhood of Teamsters. In December 2024, Teamsters at several Amazon facilities, including JFK8, went on strike in the busy weeks before Christmas, in what the union called the largest strike against the company in American history; the company said deliveries were not affected. In April 2026 — four years after the vote — the Board ordered Amazon to recognise and bargain with the union. As this chapter was written, there was still no first contract at JFK8. The workers had won the vote. They had not yet won the thing a union exists to win.

Amazon is not unusual. In the United States a newly organised workplace commonly waits well over a year for its first contract, and many never get one. Labour law built in the 1930s for factories with a stable workforce is easily obstructed by a company with high turnover, deep pockets and every reason to wait.

The history of American home care offers an instructive contrast. Through the 1990s, the Service Employees International Union tried to organise the people who looked after elderly and disabled people in their own homes in Los Angeles County. The workers were paid by public programmes, but the courts had treated them as having no employer with whom a union could bargain, because each worked for a different client. The union's answer was to change the structure: it campaigned for California law and the county to create a public body that would act as the workers' employer for the purposes of bargaining. In February 1999, after an eleven-year campaign, some seventy-four thousand home care workers voted by an overwhelming majority to join the union. It was described at the time as the largest single union victory in the United States since the auto workers organised General Motors in 1937. The workers had been scattered across tens of thousands of homes. The union did not try to gather them together. It created an employer for them to bargain with.

What the law has given

The second route has been through courts and legislatures, and it is here that the scattered have made their largest gains.

In February 2021 the Supreme Court of the United Kingdom ruled unanimously, in Uber BV v Aslam, that Uber's drivers were "workers" with rights to the minimum wage and paid holiday. The case had been brought by two drivers, Yaseen Aslam and James Farrar, with the backing of the GMB union. In May 2021 Uber recognised the GMB to represent up to seventy thousand drivers in Britain — the first time the company had recognised a union anywhere in the world. The agreement was narrower than its headlines: Uber would meet the union regularly, but would not bargain with it over pay. A legal ruling had bought the drivers a seat at the table. It had not bought them a contract.

In 2021 Spain went further. Its "rider law" presumed that delivery riders working through apps were employees unless the company could prove otherwise, and it gave workers' representatives in any company a right to be told the parameters, rules and instructions on which the company's algorithms based decisions affecting working conditions. It was the first law anywhere to give workers a collective right to see inside the algorithm. The largest Spanish delivery company, Glovo, resisted for more than three years, continuing to treat most of its riders as self-employed. It paid a heavy price: fines and claims for unpaid social security contributions reported at more than two hundred million euros, and a criminal investigation in Barcelona into whether it had denied workers their rights. In December 2024, the day before its co-founder was due to give evidence to the investigating judge, Glovo announced that it would hire its riders in Spain as employees.

The European Union then adopted the approach as a whole. The Platform Work Directive, adopted in 2024, requires member states to establish a legal presumption that people working through digital platforms are employees where the platform directs and controls their work, and lays down the first Europe-wide rules on algorithmic management: platforms must tell workers how automated systems make decisions about them, must have those systems monitored by people, and must give workers a right to a human review of significant decisions. Member states must put it into national law by 2 December 2026, a deadline that falls a few weeks after this chapter was written. How seriously they will do so, and how the platforms will respond, is not yet known.

Elsewhere, the scattered have won rights without being reclassified as employees. In New York City, after a campaign by the delivery workers' organisation Los Deliveristas Unidos, many of them migrants from Mexico, Guatemala and West Africa, the city introduced a minimum pay rate for app-based restaurant delivery workers, which reached $21.44 an hour before tips in April 2025. In November 2024 voters in Massachusetts approved, by about fifty-four per cent to forty-six, a ballot measure giving app-based ride-hail drivers the right to form a union and bargain collectively, the first such law in an American state. In India, Rajasthan set up a welfare board for platform workers in 2023, funded by a fee on transactions, and the national Code on Social Security, in force from November 2025, recognised "gig workers" and "platform workers" in law for the first time and required platform companies to pay between one and two per cent of their turnover into a fund for their social protection.

There have also been defeats, and the largest was a lesson in the farm's resources. In 2019 California passed a law, AB5, that would have made most app-based drivers employees. Uber, Lyft, DoorDash and their allies spent more than two hundred million dollars on Proposition 22, a ballot measure to exempt their drivers from it — at the time the most expensive ballot campaign in the state's history. The unions opposing it spent a small fraction of that. In November 2020 it passed with 58.6 per cent of the vote. In July 2024 the California Supreme Court unanimously upheld it.

Organising the top of the chain

The third route is perhaps the most inventive. If scattered workers cannot find an employer with the power to meet their demands, they can go over the employer's head, to the company at the top of the chain that sets the prices everyone below it must accept.

The Coalition of Immokalee Workers, an organisation of mostly migrant farmworkers in southern Florida, showed how. The tomato pickers of Immokalee worked for growers who were themselves squeezed by the fast-food chains and supermarkets that bought their crop. In 2001 the Coalition launched a boycott of Taco Bell, demanding that the company pay one penny more for every pound of tomatoes it bought, and that the extra money be passed to the pickers. In 2005 Taco Bell agreed. Over the following decade McDonald's, Burger King, Subway, Whole Foods and, in 2014, Walmart followed. The Fair Food Program that grew out of the campaign does more than pass on the penny: it enforces a code of conduct on participating farms, with independent auditors, a complaints line workers can use without fear, and a commitment by the buyers to stop purchasing from farms that break the code.

The garment industry produced a still larger example, after a catastrophe. On 24 April 2013, the Rana Plaza building outside Dhaka in Bangladesh, which housed several garment factories, collapsed, killing more than eleven hundred people, most of them young women sewing clothes for Western brands. Within weeks, two global union federations, IndustriALL and UNI, together with Bangladeshi unions, persuaded dozens of brands to sign the Accord on Fire and Building Safety in Bangladesh. Unlike the voluntary codes of conduct that had preceded it, it was legally binding: brands committed themselves to independent inspections of the factories that supplied them, to public reporting, to paying for repairs, and to arbitration if they failed to comply. More than two hundred brands eventually signed. In 2021 it became an International Accord, which began work in Pakistan in 2023.

Both examples share a structure. The workers were scattered and poorly paid, and their direct employers were weak. But the companies at the top of the chain were few, rich, visible and sensitive to their reputations. By turning the pressure of consumers and the public on them, the workers created a counterpart with which an enforceable agreement could be made: the logic of the Los Angeles home care campaign, applied to the global supply chain.

The creators strike back

One group of workers deserves particular attention in this book, because they work on the far side of the wall that Chapter 3 described. They are the creators.

On 2 May 2023, the Writers Guild of America went on strike against the Hollywood studios and streaming companies. Its demands included higher pay and a share of streaming revenue, but one demand was new: limits on the use of artificial intelligence. The strike lasted 148 days. The agreement that ended it, in September 2023, provided that AI could not write or rewrite "literary material", that material produced by AI could not count as source material that would reduce a writer's credit or pay, and that companies could not require writers to use AI, though writers could choose to use it themselves. The question of whether studios could train AI systems on writers' work was left open, with the union reserving its right to challenge it. The actors' union, SAG-AFTRA, struck for 118 days and won a requirement that studios obtain performers' informed consent before creating or using a "digital replica" of them, and pay for its use.

These agreements are modest, and cover a small, comparatively privileged workforce. But they matter for this book's design for two reasons. The first is that they show the creativity wall being defended not by the market but by organisation. The writers did not rely on AI being unable to write a script; they bargained for a rule that the human creator would remain the author of record. The second is that they offer the beginning of an answer to the danger Chapter 3 raised: that in an economy where creativity is scarce, the people who own the rights to creation — studios, platforms, the owners of models trained on human work — will become the next aristocracy. The writers organised as workers, not as owners. Their union bargained for the many writers who earn modest incomes, not only for the few stars. If creativity is not to be enclosed by those who own the machines, this is one way it might be prevented.

Across borders

Organising across borders is the hardest problem of all, and the record of success is thinnest.

There have been striking moments. On 8 May 2019, the day before Uber's shares went on sale to the public, drivers in a number of American cities, in London and other British cities, and in Australia switched off their apps in a coordinated protest against falling pay. It was timed to the company's stock market launch, and it showed that scattered workers could coordinate globally using the same digital tools that managed them. It did not change Uber's pay. The Global Trade Union Alliance of Content Moderators is a more durable attempt, but it is young.

The labour movement has older international institutions: global union federations such as IndustriALL and UNI, which negotiate framework agreements with multinational companies, and the International Labour Organization, which sets standards — in 2011, for example, a convention on decent work for domestic workers. But they depend on governments to enforce those standards and on companies to keep their word, and their power over the owners of a global economy has always been limited.

The difficulty is structural. The owners can choose where to produce, where to book their profits and where to hold their wealth; the workers, for the most part, cannot choose where to live. Every gain in this chapter was won within the jurisdiction of a particular state — Britain, Spain, the European Union, New York, Massachusetts, India — or through the reputation of a particular company. That is not nothing. But the gains are only as secure as the state that enforces them, and the owners can always try to move to one that does not. Chapter 6 will consider how gains can be locked into law, and Chapter 12 whether a borderless society would close this exit or simply move it.

Division, again

The first book's deepest lesson about organisation was that its oldest enemy is division. That lesson applies to the scattered with special force, because some of the divisions among them are real.

The most obvious is the division between those who want to be employees and those who value the freedom of working for themselves. Many gig workers genuinely value flexibility: to work when they choose, to combine several jobs, to fit work around care or study. Surveys of drivers consistently find this, and the campaign for Proposition 22 made it the centre of its message. Some of those surveys were paid for by the platforms, and the freedom is often less than it appears: a driver who can choose her hours but not her price, and who can be deactivated by an algorithm, is free in a limited sense. But organisers who dismiss the desire for flexibility as false consciousness tend to lose. The Danish approach shows that the choice need not be stark. In 2018 the Danish union 3F signed what was described as the first collective agreement with a platform company anywhere, with Hilfr, a cleaning platform: cleaners could remain freelancers, but after a hundred hours of work on the platform they became employees, with a minimum wage, holiday pay, sick pay and a pension, unless they chose not to. In 2021 3F and the Danish employers' association agreed a sector-wide agreement for food delivery riders, which the platform Just Eat was the first to sign.

The second division is between migrant and native workers. Many of the scattered workers of the rich world are migrants, and in many rich countries political movements promise to protect native workers by keeping migrants out. The first book recorded how industrial employers used one wave of immigrants against another, and how some unions grew by excluding newcomers rather than organising them. The workers who have won most in the new economy have often been migrants organising themselves — the deliveristas of New York, the tomato pickers of Immokalee, the home care workers of Los Angeles, the content moderators of Nairobi. The movements that have organised across the line between migrant and native have generally done better than those that have tried to defend one side of it.

The third division is newer and in some ways the most important for this book: the line between the herd and the shepherds. The first book described the 1% as the managing class that runs the farm on behalf of the 0.01%, and Chapter 3 described how AI may erode the lower rungs of that class. There are early signs that some of the shepherds are beginning to see themselves as workers. Workers at Google's parent company, Alphabet, formed a union in 2021, open to contractors as well as employees; journalists, academics, museum curators and video game developers have organised in growing numbers. These are, in the terms of the first book, disaffected shepherds. Many of history's breakouts were led by such people. Whether they organise alongside the scattered or in defence of their own remaining privileges will matter a great deal.

What the record shows

Looking across these cases, a pattern emerges, and it is consistent with the first book's account of what made organisation work.

The sociologist Erik Olin Wright distinguished between two kinds of power that workers can have. Structural power comes from their position in the economy: from being scarce in the labour market, or from occupying a point in production where a stoppage does great damage. Associational power comes from organisation: unions, parties, associations, cooperatives. The historian Beverly Silver, in Forces of Labor (2003), used this distinction to trace the history of workers' movements across the twentieth century, and found that workers won most where they had both.

The scattered have structural power in some places and almost none in others. Chapter 3 suggested that ageing and the wall will make care workers, skilled manual workers and creators scarcer. The Hollywood writers had structural power: the studios could not make new shows without them. The online freelancer competing with millions of others across the world has very little.

Associational power, for the scattered, has had to take new forms. The successful cases in this chapter rarely followed the classic model of a union organising one workplace and bargaining with its owner. They found or created a counterpart with the power to agree: a public authority as employer of record, an employers' association, a brand at the top of a supply chain, a city government, a court. The organiser Jane McAlevey, in No Shortcuts (2016), argued that movements win by building deep organisation among the workers themselves rather than mobilising support from outside. The cases here suggest both are needed: the workers who won built real organisations, then used them to reach whoever above them could change the rules.

There is one further pattern, which points to the next chapter. Almost every gain in this chapter came with help from somewhere among the powerful: a court willing to look past a company's contracts to its practices; a legislature willing to act; a city government in need of the workers' votes; a rival company that already paid its workers properly and wanted its competitors to do the same; a brand that feared its reputation more than it valued its suppliers' low prices. In Massachusetts, Uber and Lyft did not actively oppose the 2024 union measure, having already agreed a settlement with the state's attorney general earlier that year. Where the owners stood together, as in California in 2020, the scattered lost.

The tests

This chapter tests two parts of the Federation hypothesis: the third element, AI replacing the shepherds, which algorithmic management is already trying in practice; and the fourth, every person a worker and a shareholder — read here as the claim that the 99 can act as organised members of their society rather than as a managed herd.

Precedent. The industrial breakouts depended on workers being brought together; the scattering of the past half-century has coincided with the collapse of union membership and the rise of inequality. But the scattered have won before: the home care workers of Los Angeles, the tomato pickers of Immokalee and the Rana Plaza Accord all organised workers with no single workplace by creating a counterpart to bargain with. Algorithmic management, the first large precedent for AI as a shepherd, has so far concentrated control in the owners' hands.

Evidence. Union membership has fallen while approval of unions has risen, which suggests that the obstacle is structural and legal rather than a lack of will. Court rulings and laws in Britain, Spain, the European Union, several American cities and states, and India show that the classification of workers, the transparency of algorithms and the right to organise can all be changed by law. Evidence on the effects of the newest laws is still thin.

Capture. The farm has captured the new forms of work through the contract (calling workers entrepreneurs), through the algorithm (a manager that cannot be questioned, with a human review that can be "purely symbolic"), through money (Proposition 22) and through delay (four years without a contract at JFK8). Any design that relies on an AI shepherd must expect the same: whoever controls the machine will be tempted to use it as the gig platforms have.

Exit. Scattered workers have a strong right to exit — to log off, switch apps, quit — and a weak right to voice. Albert Hirschman argued in Exit, Voice, and Loyalty (1970) that the two work best together; exit without voice leaves people free to leave but powerless to change anything. The design must give both.

The ledger

What survived. The premise that the 99 can be organised survives, on the evidence of this chapter, though not in the forms of the industrial age. Workers with no common workplace, employer or nationality have organised and won enforceable gains, especially where they created a counterpart with the power to agree. The fourth element of the design — every person a worker and a shareholder — survives, with the understanding that being a member of a society is not the same as being organised within it.

What had to change. The third element changes most. AI as shepherd, as tested so far in the gig economy, has not removed the managing function; it has hidden it and handed it to the owners of the machine. If the Federation is to let AI replace the shepherds, the machine must be governed by the people it manages: its rules must be open to inspection, its significant decisions open to a real (not symbolic) human review, and those subject to it must have an organised voice in how it is set — a collective right, like the one in Spain's rider law, and not only an individual right to complain. Chapter 10 will have to build this. The design also needs a permanent place for organisation itself. Even where everyone is a shareholder, people will need associations of their own, independent of the machine and whoever runs it, through which to bargain, object and propose. A Federation without independent unions would be, in the first book's terms, a farm with a kinder shepherd. Finally, creators need to be organised as workers, as the Hollywood writers showed, if creativity is not to be enclosed by those who own the rights to it.

What remains unknown. Whether workers can organise across borders on anything like the scale at which capital operates; no case in this chapter has shown it. Whether the new laws in Europe, India and elsewhere will be enforced or evaded. Whether the shepherds displaced by AI will make common cause with the scattered. And whether the structural power Chapter 3 predicted for care workers and creators will become associational power before the owners find ways around it.

Bridge: when the owners split

Every gain in this chapter came with help from somewhere above. A court looked past Uber's contracts. A Spanish government chose to legislate. Taco Bell decided that its reputation was worth more than a penny a pound. Where the owners of the farm stood together — the gig companies in California, Amazon on Staten Island — the scattered lost or waited. The first book found that breakouts most often succeeded when the owners were divided among themselves: when barons fought kings, industrialists fought landlords, or one faction of the powerful needed the many against another. The next chapter asks which divisions among the owners of the coming decades — technology against finance, new money against old, nationalist against globalist — the 99 could use, and at what risk.