The solution to civilisation's problems is not more civilisation. It is less -- organised differently.
I want to begin with a colony that splits itself in half. The Hutterites are a communal Anabaptist group who have farmed the northern Great Plains of North America since the 1870s, living in colonies where property is held in common, meals are eaten together, and decisions are made by consensus. They have been doing this, with minor variations, for nearly five centuries -- longer than most nation-states have existed. And they have discovered, through trial and failure rather than through theory, that a colony works until it reaches approximately one hundred and fifty people. After that, something changes. The social fabric loosens. Free-riders appear. Conflicts that were once resolved over dinner require formal adjudication. The colony, as the Hutterites describe it, becomes "unmanageable." Their solution is not to impose hierarchy, not to appoint a police force, not to write a constitution. Their solution is to split. When the colony approaches one hundred and fifty members, they divide it in two. New land is acquired. Half the families move. Two colonies of seventy-five replace one colony of one hundred and fifty, and the mechanism that was failing -- trust maintained through personal knowledge of every other member -- begins to function again.
Does this number sound familiar? It should. We have been circling it since Chapter 1.
Robin Dunbar, the British anthropologist and evolutionary psychologist who has spent three decades studying the relationship between primate brain size and social group size, would not have been surprised by the Hutterites' threshold. In 1993, Dunbar published a paper extrapolating the correlation between neocortex ratio and mean group size in non-human primates to predict the natural group size for humans. The human neocortex, plugged into the primate regression equation, produced a number: approximately one hundred and fifty. Dunbar's number, as it came to be known, represents the cognitive limit on the number of people with whom a human can maintain a stable social relationship -- a relationship involving personal knowledge of who each person is and how they relate to every other person in the group. Below one hundred and fifty, you can manage the group through what Dunbar calls "social grooming" -- gossip, reputation, personal familiarity. Above one hundred and fifty, you cannot. You need rules, hierarchies, bureaucracies, enforcement mechanisms. You need, in other words, the institutional architecture described in every chapter of Part Two of this book.
The number appears with a regularity that is either deeply significant or suspiciously convenient, depending on your disposition. The Roman century -- the basic tactical unit of the Roman army -- comprised eighty to one hundred soldiers plus their officers, and a cohort of six centuries approximated five hundred to six hundred, with the maniple of two centuries sitting squarely at one hundred and sixty. Modern military companies, across nations and eras, average around one hundred and fifty personnel. Anglo-Saxon villages in the Domesday Book cluster around this figure. Neolithic farming communities associated with stone circles in Britain average one hundred and fifty to one hundred and sixty individuals, based on the estimated number of dwellings. Christmas card distribution lists -- a proxy for active social networks in pre-digital Britain -- peaked at an average of one hundred and fifty-four recipients. Mobile phone calling data, analysed across millions of users, reveal that the number of distinct contacts with whom an individual maintains regular communication averages, once more, approximately one hundred and fifty. The number is not magic. It is a constraint -- a ceiling imposed by the size of the human neocortex and the time costs of maintaining relationships. It is the answer to a simple question: how many other humans can one human actually know?
Dunbar's work also revealed that the one hundred and fifty is not a homogeneous mass. It is layered. Nested. Within the one hundred and fifty sits a sympathy group of approximately fifty -- people whose death would be genuinely distressing to you. Within that, a close support group of approximately fifteen -- the people you turn to in a crisis, the people you speak to at least once a month. Within that, an intimate circle of approximately five -- the people who constitute your core emotional world, to whom you devote roughly forty percent of your available social time. The structure is fractal: 5, 15, 50, 150, and it continues outward to 500 (acquaintances whose names you know) and 1,500 (faces you can recognise). Each layer requires a different intensity of investment. Each layer serves a different function. And the layers do not exist by choice. They exist because the primate brain can only process so much social information, and it allocates that processing power according to a hierarchy of emotional closeness that has been conserved, with minor variations, across the entire primate order. How many people are in your five? Can you name them without hesitation?
This is the number that every system described in Part Two has ignored. Cities of millions. Corporations of tens of thousands. Schools of fifteen hundred. Electoral constituencies of seventy thousand. Health systems serving populations of millions. Every one of these institutions operates above the scale at which the human animal can maintain trust through personal knowledge, and every one of them compensates for that deficit with the same set of surrogates: contracts, regulations, surveillance, enforcement, bureaucracy, hierarchy. The surrogates work, after a fashion. They produce functional societies. But they produce them at a cost that Chapter 11 described in detail: the erosion of trust, the replacement of relationships with transactions, the conversion of every human interaction into a problem of compliance rather than a practice of mutual knowledge. The question that Part Four must answer is not "How do we abolish civilisation?" -- that question is adolescent. The question is: how do we organise human life so that the fundamental unit of social organisation respects the animal's cognitive and emotional architecture, while still maintaining the benefits of large-scale cooperation?
The answer is the cluster.
A cluster is not a commune. It is not a cult. It is not a return to the village, or a utopian experiment, or a gated community for people who have read too many books about intentional living. A cluster is a self-governing group of people -- one hundred and fifty or fewer -- who share enough of their daily lives to maintain trust through social visibility, who rotate service roles rather than electing permanent leaders, who train their members in conflict resolution, and who are networked with other clusters to provide the specialisation, infrastructure, and emergency capacity that no group of one hundred and fifty can supply alone.
I need to be precise about what I mean by "self-governing," because the word carries baggage. Self-governing does not mean isolated. It does not mean sovereign in the nation-state sense. It does not mean free from external accountability. It means that the decisions that most directly affect the daily lives of the cluster's members -- how shared spaces are maintained, how disputes are mediated, how service roles are allocated, how resources are distributed within the group -- are made by the group, within the group, through processes that every member can see and participate in. Decisions that affect multiple clusters -- infrastructure, emergency response, resource allocation at scale -- are made at the network level, through mechanisms I will describe later in this chapter. The principle is subsidiarity: decisions are made at the smallest scale at which they can be competently made. A cluster decides its own meal rota. The network decides the water supply.
This is not theoretical. There are working models, running now, in multiple countries, at multiple scales. None of them is perfect. All of them are instructive. And all of them suggest that we are closer to this than most people think.
In Denmark, the practice of bofaellesskab -- literally "living community" -- has been producing functional clusters since the late 1960s. A cohousing community typically comprises twenty to forty households sharing a common house with a large kitchen, dining hall, laundry, workshops, guest rooms, and sometimes childcare facilities. Each household maintains a private dwelling -- a flat or a terraced house -- with its own kitchen, bathroom, and living space. The common house is not a replacement for private life. It is an addition to it. A threshold between solitude and society that the members cross voluntarily, daily, in both directions.
The numbers are instructive. The Danish cohousing database maintained by the association Bofaellesskab.dk lists approximately three hundred existing communities, though researchers at Aalborg University have noted that the actual number is likely higher, as communities are not systematically registered by any public or private entity. A qualified estimate from the housing studies literature places the number of "traditional" intergenerational cohousing communities -- that is, communities maintaining a balance between common and private life, not reserved for particular age groups -- at around one hundred and fifty, with fifteen to twenty new communities established each year, a rate that has accelerated since the 2010s. The model has spread. Cohousing communities based on the Danish template now exist in the Netherlands, Sweden, Germany, the United Kingdom, the United States, Canada, and Australia.
What matters, for the purposes of this chapter, is not the architecture but the social mechanics. In a Danish cohousing community, meals are shared several times a week, with cooking duties rotated among all adult members. Maintenance of common spaces -- gardens, the common house, laundry facilities -- is divided through a system of working groups, with participation expected from all households. Decisions about community life are made by consensus at regular meetings, typically monthly, at which every household has a voice. There is no elected leader. There is no board of directors. There is no manager. There is a rotating facilitator, a set of working groups, and a culture of direct participation that is maintained not by rules but by the simple fact that everyone knows everyone. You do not free-ride in a community of thirty-five households because the people you would be free-riding on are the people who cooked your dinner last Tuesday.
This is Dunbar's mechanism in action: social visibility maintaining cooperation. The community is small enough that reputation works. Gossip works. Knowing who did what and who did not do what works. The enforcement mechanism is not a rule book or a security camera or a human resources department. It is the raised eyebrow of a neighbour who noticed that you have not signed up for a cooking shift in three weeks. It is the conversation at the dinner table. It is the accumulated social knowledge that every primate brain is exquisitely designed to track. We already know how to do this. Our brains are built for it. We just need to give them a group small enough to work with.
In 1956, a Catholic priest named Jose Maria Arizmendiarrieta, who had been working with young people in the town of Mondragon in Spain's Basque Country since 1941, helped five of his former students establish a small factory manufacturing paraffin heaters. The factory was organised as a cooperative: the workers owned it, the workers governed it, and the profits belonged to the workers. The enterprise was called ULGOR -- derived from the first letters of the five founders' surnames. Sixty-eight years later, the Mondragon Corporation is the largest cooperative enterprise in the world. It comprises approximately ninety-six cooperatives organised into four areas -- finance, industry, retail, and knowledge -- employing over seventy thousand worker-owners. It operates its own university, its own bank, its own social security system, and its own research centres. Its annual revenue exceeds twelve billion euros.
The numbers that matter, for the argument of this chapter, are not the revenue figures. They are the ratio figures. In Mondragon's cooperatives, the pay ratio between the highest-paid and lowest-paid worker is capped -- by democratic vote of the worker-owners themselves -- at a ratio that ranges from 3:1 to 9:1 across different cooperatives, with an average of approximately 5:1. The general manager of an average Mondragon cooperative earns no more than five times what the lowest-paid worker earns. In the United States, the average ratio between CEO compensation and median worker pay was 344:1 in 2023, according to the Economic Policy Institute. In the United Kingdom, the High Pay Centre reported a ratio of approximately 109:1 for FTSE 100 companies. Mondragon's ratio is not an aspiration or a policy proposal. It is a functioning reality, sustained across nearly seven decades, in a competitive global market, producing appliances, automotive components, machine tools, and retail goods that compete with -- and frequently outperform -- conventionally organised firms. What does it tell us that a cooperative in the Basque Country can maintain a 5:1 pay ratio while producing globally competitive products?
How does Mondragon maintain this? Through a governance structure that, stripped of its Basque terminology, is essentially the cluster principle applied to economic life. Each cooperative is governed by a General Assembly of all worker-owners, which meets at least annually and has supreme authority over the cooperative's direction. The General Assembly elects a Governing Council -- analogous to a board of directors -- from among the worker-owners. The Governing Council appoints a General Manager, but the General Manager serves at the pleasure of the worker-owners and can be removed by the General Assembly. A Social Council handles workplace conditions, compensation, and social matters -- a parallel governance body ensuring that the economic decisions of the Governing Council are checked by the social needs of the membership. Profit distribution is decided cooperatively: roughly sixty percent is reinvested in the business, thirty percent goes to worker-owners as capital in their individual accounts, and ten percent is directed to community development.
The system is not without failures, and the failures are instructive. In 2013, Fagor Electrodomesticos -- one of Mondragon's founding cooperatives and its largest industrial enterprise, with over five thousand workers -- collapsed under the combined weight of the European debt crisis, falling consumer demand, and overexpansion into markets it could not sustain. The failure was real. Jobs were lost. The cooperative model was tested. But what happened next reveals the difference between a cluster and an isolated firm. Mondragon's internal insurance organisation, Lagun Aro, paid eighty percent of the displaced workers' salaries for two years while the network worked to relocate them into other cooperatives. Of the approximately 1,800 worker-owners affected, the majority were reabsorbed into the Mondragon system. The organism lost a cell. It did not lose the tissue. The network absorbed the shock because the network was designed to absorb shocks -- not through a government bailout or a charitable intervention, but through the structural solidarity of interconnected cooperatives that had been contributing to a common fund precisely for this purpose.
This is the cell-to-organism relationship that I want to hold in view for the remainder of this chapter. A single cluster -- whether it is a Danish cohousing community, a Mondragon cooperative, or a Hutterite colony -- is a cell. It has its own membrane, its own internal processes, its own self-regulation. But it cannot survive alone. It lacks the specialisation, the scale, and the redundancy that only a network of cells can provide. The cluster is the fundamental unit. The network is the organism. And the organism is only as healthy as the connections between its cells.
If the Danish model demonstrates the cluster in domestic life, and Mondragon demonstrates it in economic life, then Buurtzorg demonstrates it in professional life -- and demonstrates it with an elegance that makes the conventional alternative look not merely inefficient but absurd.
Buurtzorg -- the name means "neighbourhood care" in Dutch -- was founded in 2006 by Jos de Blok, a former nurse who had watched the Dutch home-care system progressively fragment nursing work into discrete, timed tasks assigned by managers who had never met the patients. The system was efficient on paper. In practice, it meant that an elderly patient recovering from a hip replacement might see fifteen different nurses in a week, each performing a different task -- wound care, medication, bathing -- according to a schedule determined by a planning department. No nurse knew the patient. No nurse knew the other nurses. The system had achieved administrative legibility at the cost of every quality that makes nursing effective: continuity, relationship, autonomy, judgement. Does this sound familiar? It is the Dunbar problem again -- the system grew past the scale at which the humans within it could know each other, and institutions replaced relationships.
De Blok's alternative was radical in its simplicity. He formed a team of four nurses. They managed themselves. They decided which patients to see, how often, for how long, and what care to provide. There was no manager. There was no planning department. There was a small back-office providing IT support and administrative services, and beyond that, the nurses governed their own practice.
By 2015, Buurtzorg employed over ten thousand nurses and nurse assistants, organised into more than eight hundred and fifty self-managing teams of ten to twelve people. Each team serves a neighbourhood, caring for fifty to sixty patients. The teams have no manager. They have no hierarchy. They make decisions by consensus, using a structured meeting process. They recruit their own members, manage their own schedules, and handle their own finances. A small coaching staff -- roughly twenty coaches for the entire organisation -- is available when teams encounter problems they cannot resolve internally, but the coaches have no authority over the teams. They advise. They do not direct.
The results are not ambiguous. A 2009 Ernst and Young study found that Buurtzorg's patients required care for less time and regained autonomy more quickly than patients served by conventional home-care organisations. Buurtzorg used forty percent of the authorised patient care hours, compared with an industry average of seventy percent -- not because it provided less care, but because it provided better care, earlier, in the context of a relationship, by nurses who knew the patient and could exercise professional judgement about what that patient actually needed. A KPMG study found that Buurtzorg's overhead costs were eight percent, compared with an industry average of twenty-five percent -- because the layer of management that conventional organisations require to coordinate fragmented tasks simply did not exist. Buurtzorg has been named the best employer in the Netherlands multiple times. Patient satisfaction ratings are the highest in the sector.
The mechanism is, once again, Dunbar's. A team of twelve is well within the intimate circle -- the five-to-fifteen layer where trust is deepest and collaboration most natural. Every team member knows every other team member. Every nurse knows every patient. The social visibility that maintains cooperation in a Hutterite colony or a Danish cohousing community maintains cooperation in a Buurtzorg nursing team. No surveillance is needed because the group is small enough that everyone can see everything. No manager is needed because the group is small enough that decisions can be made by the people who have to live with the consequences.
De Blok, when asked how he scaled a self-managing organisation to over ten thousand people, gave an answer that should be engraved above the entrance to every business school on the planet: "We didn't scale. We copied." Each team is a new cell. Each cell is autonomous. The organism grows by adding cells, not by making cells larger. What if we applied this principle beyond nursing -- to our schools, our workplaces, our neighbourhoods?
The cluster solves the trust problem. But it creates a governance problem, because somebody has to make decisions, and the history of small self-governing groups is littered with the carcasses of communities that were destroyed not by external threat but by internal dysfunction: the charismatic founder who became a tyrant, the consensus process that was captured by the loudest voice, the egalitarian ideal that concealed a rigid informal hierarchy. We have all seen this -- in families, in workplaces, in any group where one person's force of personality gradually becomes indistinguishable from authority.
The zoological lens offers a diagnosis. In every primate group studied in the wild, the individual who seeks dominance most aggressively is rarely the individual who governs most effectively. Frans de Waal, whose work on chimpanzee social behaviour at the Arnhem Zoo colony I described in Chapter 1, documented this pattern across decades of observation. The alpha male who maintained his position through aggression alone was unstable -- constantly challenged, constantly vigilant, generating stress across the entire group. The alpha who maintained his position through coalition-building, conflict resolution, and strategic generosity was stable and, critically, produced a calmer, more cooperative group. But de Waal also observed that the drive toward dominance was itself a consistent feature of chimpanzee psychology: some individuals wanted the position, worked for it, manoeuvred for it. The question was never whether hierarchy would emerge. The question was how to prevent the hierarchy from being captured by the individual whose desire for power was strongest.
The Athenians solved this problem twenty-five hundred years ago, and their solution is so simple that it sounds like a joke: they used a lottery. In the Athenian democracy of the fifth and fourth centuries BCE, most public offices were filled not by election but by sortition -- random selection from among eligible citizens. The mechanism was a device called the kleroterion: a stone slab with rows of slots into which citizens placed their identification tokens. A tube attached to the side of the slab dispensed coloured dice -- black or white -- and each die determined the fate of one row: black eliminated, white selected. Because each row contained exactly one representative from each of the ten tribes of Athens, the procedure ensured both randomness and proportional representation. The Athenians did not consider election to be democratic. They considered it aristocratic -- a mechanism by which the wealthy, the well-known, and the well-connected could secure power. Sortition, by contrast, gave every citizen an equal probability of serving. It was, in the Athenian view, the only truly democratic method of selecting officials. What would our politics look like if we selected our decision-makers the way we select our juries?
The insight beneath the method was not mathematical but psychological: the person who seeks office is precisely the person most likely to abuse it. Sortition bypasses the self-selection problem entirely. The farmer, the potter, the merchant, the teacher -- any of them might find themselves serving on the Boule, the five-hundred-member council that set the legislative agenda, for a single year. Service was temporary, rotational, supported, and expected. It was not a career. It was a duty, like jury service -- which, in most modern democracies, remains the last surviving artefact of the Athenian sortition principle.
The modern revival of sortition is no longer hypothetical. In 2016, the Irish government, facing decades of political deadlock on the question of abortion, established a Citizens' Assembly composed of ninety-nine citizens randomly selected to be broadly representative of Irish society by age, gender, social class, and regional distribution. The Assembly deliberated over five weekend sessions, heard from twenty-five expert witnesses, reviewed hundreds of submissions, and reached a conclusion that the elected parliament had been unable to reach for thirty-five years: that the constitutional provision on abortion was unfit for purpose and should be repealed. The Assembly's recommendation was put to a national referendum on 25 May 2018 and approved by 66.4 percent of voters. Randomly selected citizens, with no political careers to protect and no lobbyists to satisfy, reached a consensus that professional politicians could not.
In 2019, President Macron of France, facing the Gilets Jaunes protests, convened the Convention Citoyenne pour le Climat -- one hundred and fifty citizens selected by sortition from a pool of 250,000 randomly generated telephone numbers, stratified to ensure demographic representativeness. The Convention deliberated over seven weekend sessions and produced one hundred and forty-nine policy recommendations for reducing France's carbon emissions by forty percent. The process was not without flaws -- Macron subsequently diluted or rejected many of the recommendations, prompting justified criticism -- but the deliberative quality of the Convention itself was remarkable. Citizens with no prior expertise in climate policy, given access to experts, time to deliberate, and freedom from electoral pressure, produced recommendations that were more ambitious, more coherent, and more publicly supported than anything the elected parliament had managed.
In 2019, the Parliament of the German-speaking Community of Belgium -- the small region of Ostbelgien -- went further than any government in history. It established a permanent Citizens' Council: twenty-four randomly selected citizens serving eighteen-month terms, empowered to convene Citizens' Assemblies of twenty-five to fifty randomly selected citizens on any topic the Council deems important. The Assemblies deliberate, produce recommendations, and present them to the Parliament, which is required to respond publicly. In the five years since its establishment, the Permanent Citizens' Council has organised six Citizens' Assemblies. The model has been recognised by the OECD and the Council of Europe. Cities including Aachen, Brussels, and Paris have adopted modified versions.
This is what governance looks like in a cluster. Not elected representatives serving multi-year terms, accumulating power, developing the professional politician's characteristic deformation -- the inability to say what they think for fear of losing what they have. Randomly selected citizens, serving temporary terms, supported with expert information, deliberating in good faith because they have nothing to gain from bad faith. Sortition does not produce perfect outcomes. No governance mechanism does. What it produces is a structural resistance to the capture of power by the individuals who most desire it -- which is, as de Waal's chimpanzees and every chapter of Part Two have demonstrated, the central vulnerability of every governance system the species has built. We can design around this vulnerability. The tools already exist.
Governance distributes power. But power is not the only source of conflict. People living in close proximity will disagree, misunderstand each other, hurt each other, and occasionally hate each other, regardless of how wisely power is distributed. The cluster needs a mechanism for repairing the social fabric after it tears. It needs mediators.
The zoological evidence for third-party conflict resolution is extensive and, to anyone who assumes that peacemaking is a uniquely human cultural achievement, humbling. De Waal's work at Arnhem, beginning in the 1970s, documented in meticulous detail the phenomenon he called reconciliation: after an aggressive conflict between two chimpanzees, the former opponents showed a statistically elevated rate of friendly contact -- approaching, touching, kissing, embracing -- in the minutes and hours following the fight. The rate exceeded baseline expectations. The behaviour was not random. It was directed: the opponents sought each other out. The aggression had created a social debt, and the reconciliation was the payment. But de Waal also documented something more subtle: consolation. After a fight, a third party -- typically a close associate of the victim -- would approach the distressed individual and offer physical contact: an arm around the shoulder, grooming, sitting in close proximity. The consolation behaviour reduced the victim's stress indicators. It was preferentially offered by individuals who shared a valuable relationship with the victim. And it was, in de Waal's assessment, the best behavioural marker of empathy observed in any non-human animal.
Consolation is not limited to chimpanzees. In 2010, Orlaith Fraser and Thomas Bugnyar, working with captive ravens at the Konrad Lorenz Research Station in Austria, published a study in PLOS ONE demonstrating that after aggressive conflicts, bystander ravens offered affiliative contact to the victims -- and that this bystander affiliation was more likely when the bystander shared a valuable relationship with the victim and when the conflict had been particularly intense. The ravens were not reconciling -- the opponents did not reliably seek each other out. But the bystanders were consoling, and they were doing so selectively, based on relationship quality and conflict severity. The behaviour required the bystander to assess the emotional state of the victim, identify a relationship worth maintaining, and intervene with contact that functioned to reduce distress. This is, in essence, what a mediator does.
In 2011, Inbal Ben-Ami Bartal, Jean Decety, and Peggy Mason at the University of Chicago published a study in Science that pushed the evidence further down the mammalian tree. A free rat, placed in an arena with a cagemate trapped in a restrainer, learned over several sessions to deliberately open the restrainer and liberate the cagemate. The rats did not open empty restrainers. They did not open restrainers containing inanimate objects. They opened the restrainer specifically to free a distressed conspecific. And when the experimenters placed a second restrainer containing chocolate alongside the restrainer containing the trapped rat, the free rats opened both -- and, in most trials, shared the chocolate. A rat, given a choice between helping a distressed companion and consuming a high-value food reward, chose both. Pro-social behaviour -- action motivated by another's distress -- is not a human invention layered on top of selfish biology. It is woven into the mammalian substrate. The capacity for peacemaking is not something we need to learn from scratch. It is something we need to stop suppressing.
What these studies reveal, taken together, is that conflict resolution through third-party intervention is not a cultural technology that humans invented and could, in principle, uninvent. It is a biological capacity -- present in chimpanzees, present in ravens, present in rats -- that humans have elaborated culturally but did not create from nothing. The peacemaker chimp, the consoling raven, the door-opening rat are all doing versions of the same thing: responding to social distress with pro-social action, mediated by relationship quality and emotional assessment. They are natural mediators. The capacity is in the hardware.
The cluster operationalises this capacity. Every member of a cluster is trained in basic conflict mediation -- not as a professional qualification but as a life skill, equivalent to cooking or first aid. The training is not esoteric. It involves learning to listen without interrupting, to paraphrase before responding, to identify underlying needs beneath surface positions, to acknowledge harm without assigning blame, and to facilitate a process by which the parties themselves generate a resolution. These are skills that can be taught in a weekend and refined over a lifetime. They are the cultural elaboration of the consolation instinct that de Waal observed in Arnhem: the capacity to approach a distressed social partner and do something that helps. In a cluster, conflict is not pathologised. It is expected, because it is inevitable in any group of social primates. What is not expected -- what the cluster treats as a design failure rather than a personal failure -- is unresolved conflict. The question is never "Why are you fighting?" The question is "What do you need, and how can we get there?"
I do not want to present the cluster as an untested ideal. It has been tested. It has succeeded and it has failed, and both the successes and the failures are essential to the design. We learn as much from the failures -- perhaps more.
The kibbutz movement offers the most comprehensive long-term dataset on communal living in the modern era. The first kibbutz, Degania, was established in 1910 in what was then Ottoman Palestine. By the mid-twentieth century, there were over two hundred and seventy kibbutzim in Israel, housing roughly 130,000 people -- approximately eight percent of the population -- and exercising an influence on Israeli political, military, and cultural life wildly disproportionate to their numbers. The early kibbutzim were radically communal: property was held collectively, income was shared equally regardless of role, children were raised in communal children's houses rather than in family homes, and decisions were made by general assembly. No member exercised personal property rights. No member earned a personal salary. The kibbutz provided housing, food, clothing, healthcare, education, and a small personal allowance.
The model worked, with notable success, for several decades. The kibbutzim were astonishingly productive -- pioneers in desert agriculture, early adopters of industrial technology, significant contributors to Israeli defence and cultural life. But from the 1970s onward, strains appeared that intensified into a crisis in the 1980s. Israel's hyperinflation and banking collapse hit the kibbutzim hard. Many had taken on debt during a period of expansion and now found themselves unable to service it. The government and banks eventually arranged bailouts, but the damage was psychological as much as financial: members who had entrusted their entire economic lives to the collective saw the collective fail. Personal savings, which had never been encouraged, suddenly felt necessary. The communal children's houses -- which had always been controversial among some parents -- began to close. Children returned to family homes. Differential salaries replaced equal pay. Private property replaced communal ownership. By the twenty-first century, the majority of kibbutzim had been "privatised" -- retaining some communal features but operating, in economic terms, much more like conventional communities with shared facilities.
What happened? The standard interpretation is that the kibbutz model was idealistic but unsustainable -- that human selfishness eventually reasserted itself over communal aspiration. I think this interpretation is wrong, or at least incomplete. The kibbutz model was not undone by selfishness. It was undone by rigidity. The early kibbutzim made three errors that the cluster design must avoid.
First, they eliminated privacy. Communal child-rearing, communal property, communal meals with no private alternative -- these were not adjustments to the social architecture. They were the abolition of one of Dunbar's layers. The intimate circle of five -- the family unit, the core emotional world -- was dissolved into the community. Parents who wanted to raise their own children were told that the community's method was superior. The organism's need for a boundary between public and private life was treated as a bourgeois remnant to be overcome, rather than as a biological requirement to be respected. The Danish cohousing model, by contrast, maintains private dwellings alongside common spaces. The threshold between solitude and society exists, and the organism crosses it voluntarily. The cluster must have a membrane. The cell that has no membrane is not a cell. It is a spill.
Second, they enforced equality of outcome rather than equality of access. The commitment to identical compensation regardless of contribution produced a well-documented problem: members who worked harder, or who developed specialised skills, or who carried heavier responsibilities, received the same reward as members who did less. The ideal was beautiful -- from each according to ability, to each according to need. The biology was uncooperative. Humans, like all primates studied to date, have an acute sensitivity to perceived unfairness. De Waal demonstrated this experimentally with capuchin monkeys: a monkey who sees a neighbour rewarded with a grape while she receives a cucumber for the same task will reject the cucumber -- will, in some trials, throw it back at the experimenter. The sensitivity cuts in both directions: we object to receiving less than others for the same work, and we object -- less viscerally but measurably -- to seeing others receive the same as us for less work. Mondragon solved this with a capped ratio: the range between highest and lowest is bounded, but it is not zero. The cluster permits differentiation within limits. The limits are set collectively. The floor is high. The ceiling is low. But the floor and the ceiling are not the same number.
Third, and most critically, the kibbutzim were isolated. Each kibbutz was a world unto itself, connected to other kibbutzim through ideological federations but not through the kind of structural interdependence that Mondragon's cooperative network provides. When Fagor collapsed, the Mondragon network reabsorbed 1,800 workers because the network had been designed for mutual support -- common insurance funds, retraining programmes, reallocation mechanisms. When a kibbutz failed, its members had no comparable safety net within the movement. They fell into the general Israeli economy. The cluster, as a design unit, must not be self-sufficient. It must be networked. Independence without interdependence is fragility masquerading as strength.
The cluster is a cell. But a cell alone is not an organism. It is a bacterium -- capable of survival, but incapable of complexity. The transition from single-celled to multicellular life is, by the consensus of evolutionary biology, one of the half-dozen most significant transitions in the history of life on Earth. It has occurred independently at least twenty-five times. And in every case, the transition involved the same fundamental trade: individual cells surrendered some of their autonomy in exchange for the emergent properties that only a collective of specialised cells can produce. A liver cell cannot see. An eye cell cannot detoxify. But the organism that contains both can navigate the world and survive the poisons it encounters there. The properties of the organism -- vision, detoxification, locomotion, thought -- are not present in any individual cell. They emerge from the interaction of specialised cells connected through communication networks.
The cluster network operates on the same principle. A single cluster of one hundred and fifty people cannot perform surgery, cannot manufacture semiconductors, cannot maintain a water treatment plant, cannot produce the antibiotics that prevent a scratch from becoming a death sentence. These capacities require specialisation and scale that exceed the cluster's size. But a network of clusters -- ten, a hundred, a thousand -- can distribute specialisation across its components. One cluster's members include three surgeons. Another's include two water engineers. Another's include a team of teachers who have developed a particular expertise in early childhood development. The specialisations are not locked in. Members move between clusters. Knowledge transfers through the network. But at any given time, the network contains more expertise than any individual cluster, just as the organism contains more capability than any individual cell.
The network also provides redundancy. When Buurtzorg's model is examined closely, the most striking feature is not the autonomy of the individual teams but the minimal connective tissue that links them: a shared IT platform, a small back-office for invoicing and regulatory compliance, and a coaching staff that intervenes only when requested. This is not accidental. It is a design choice. The network does not manage the cells. It supports them. It provides the shared infrastructure -- the circulatory system, the nervous system, the immune system -- that the cells need but cannot produce individually. And when a cell fails, the network provides the redundancy that allows the organism to continue functioning while the damaged cell is repaired or replaced.
The Mondragon network is the most developed real-world example of this architecture. Its cooperatives share a common bank (Laboral Kutxa), a common university (Mondragon University), a common social security system (Lagun Aro), and a common research infrastructure (twelve technology centres). These shared institutions are governed cooperatively -- their boards are composed of representatives from the member cooperatives -- but they operate at a scale that no individual cooperative could achieve. The bank provides financing. The university provides training. Lagun Aro provides insurance. The technology centres provide innovation. Each cooperative is autonomous in its daily operations. The network provides the services that make autonomy sustainable.
This is not a hierarchy. It is an ecology. The relationship between the cluster and the network is not the relationship between the employee and the corporation, or the citizen and the state. It is the relationship between the cell and the organism -- a relationship of mutual dependence in which the cell maintains its internal integrity while contributing to and benefiting from the emergent properties of the whole. The cell does not take orders from the organism. The organism does not exist without the cell. The communication between them is continuous, bidirectional, and mediated by structures that are themselves governed by the components they serve.
I want to be honest about the limits of what I am proposing, because the history of utopian thought is a history of people who were not honest about limits, and the consequences were invariably borne by someone other than the thinker.
The cluster is not a solution to every problem. It is a design unit -- a way of organising the fundamental social unit of human life so that it respects the animal's cognitive architecture. It does not, by itself, fix the food supply, reform education, heal the mind-body split, or reconnect the indoor species with the outdoor world. Those are the subjects of the chapters that follow. The cluster is the container. It is the habitat within which the other reforms become possible. You cannot rewild an animal without first providing a habitat that supports wild behaviour. You cannot redesign civilisation without first establishing the scale at which the redesign is implementable.
The cluster is also not a rejection of modern life. I am not arguing for a return to the Neolithic. I am writing this chapter on a laptop, in a flat in a European city, using electricity generated by a power plant I have never visited and transmitted through infrastructure I do not understand. I do not wish to give up antibiotics, or central heating, or the ability to speak to my mother in England while sitting in the Netherlands. What I am arguing is that these benefits -- the genuine, life-saving, life-enhancing benefits of large-scale cooperation -- do not require the specific institutional arrangements that currently deliver them. They require networks. They do not require corporations of fifty thousand employees managed through seven layers of hierarchy. They require coordination. They do not require governments of career politicians serving consecutive terms in buildings most citizens will never enter. We can keep what works and redesign what does not.
The cluster is the habitat. It is the scale at which the animal can see, know, trust, and be accountable to the other animals in its group. It is the scale at which governance can be rotational and service-based rather than electoral and career-based. It is the scale at which conflict can be resolved by trained mediators who know both parties, rather than by courts that know neither. It is the scale at which the free-rider is visible and the contributor is recognised, without surveillance technology or performance metrics. It is, in short, the scale at which the social primate can do what social primates have been doing for sixty million years: live in a group small enough to know.
I sat, a few months ago, in the common house of a cohousing community on the outskirts of a Dutch city whose name I will withhold because the residents asked me to. It was a Tuesday evening. Thirty adults and perhaps a dozen children were eating a meal that three of them had spent the afternoon preparing -- a rota they have maintained, with modifications, for eleven years. The children moved between tables. Two teenagers were loading the dishwasher without being asked. A woman in her seventies was explaining something about the garden's irrigation system to a man in his thirties who had moved in eight months earlier. At the next table, a couple were having what appeared to be a tense conversation, and a third person -- a neighbour, not a therapist -- was sitting with them, listening, occasionally speaking. Nobody was performing community. Nobody was living out an ideology. They were eating dinner. They knew each other's names, each other's children, each other's habits, each other's vulnerabilities. The group was small enough that knowledge did the work that institutions do at scale. Trust did the work that contracts do. Proximity did the work that surveillance does.
It was not utopia. The woman who told me about the irrigation system also told me about a dispute over parking that had consumed three community meetings and left two households not speaking for a month. The community had, she said, "almost broken" over the question of whether dogs should be allowed in the common garden. Almost broken. Over dogs. This is what humans do. This is what social primates have always done -- fought over territory, status, resources, and the proximity of other primates' animals. The question has never been whether conflict will occur. The question is whether the group is structured so that conflict can be survived. The cohousing community survived the dog dispute. They survived the parking dispute. They survived because they were small enough to see each other, and because they had learned -- imperfectly, gradually, through practice -- how to repair what broke. I sat there that Tuesday evening and thought: this is not perfect. But it is recognisable. This is what the animal looks like when the scale is right.
The cluster is the habitat. It is the space within which the social primate can be a social primate -- cooperating, conflicting, reconciling, contributing, receiving, governing, and being governed, at a scale the brain evolved to handle. But the habitat is only the container. The question that remains -- the question that Chapter 18 must address -- is what happens inside. How does the animal learn? Not in a classroom, not from a curriculum, not through thirteen years of compulsory instruction followed by three years of debt. How does the animal actually learn?
The cluster provides the structure. But structure without learning is an empty enclosure -- a habitat with walls and no enrichment. Chapter 18 examines what it means to learn as an animal learns: by doing, by watching, by failing, by playing. The lion cub does not attend Hunting Academy.