# The Nature of the Club

> Source: <https://demonstrandom.com/essays/posts/theory_of_the_club/>
> Published: 2026-07-19 04:00:00+00:00

In previous essays I considered how value will change in a [post-AGI society](https://demonstrandom.com/essays/posts/human_value_post_ai/index.html), and also thought about what happens when [demand becomes a constraint on value](https://demonstrandom.com/essays/posts/thoughts_on_demand/index.html) rather than supply. I argued that if AI makes the supply side vastly more efficient, there will be both immediate term and long term constraints in how much demand is available.

While it is difficult to imagine a total limit to human desire, but nevertheless there are several mechanisms that can constrain demand locally or temporarily. First of all, in the short term supply of new goods can simply grow faster than people can discover, evaluate, finance, and absorb it. Similarly, in the ultimate term, we might expect that there is some asymptote beyond which the value of a marginal good for a human diminishes to a very low value.

More concretely, we can consider “satiable goods”. By analogy with rivalrous goods (whose available supply can be exhausted through consumption) a satiable good has a stock of demand that can eventually be exhausted through production. Even if human wants remain infinite in the abstract, demand for particular goods need not be. That is, for rivalrous goods supply is finite, so buying a rivalrous good is a zero-sum game; for satiable goods, demand is finite, so selling a satiable good is a zero-sum game.

If the bottlenecks in value creation migrate from the production of goods to the production, discovery, aggregation and organization of demand, then institutions built around the supply and manufacture of goods should lose relative importance1. In their steads, we would there expect new organizations to form around the new scarce demand-side activities, such as identifying what people want, coordinating their choices, and turning those choices into commitments that producers can act on.

The characteristic institution of industrial capitalism is the firm. A firm brings together land, labor, capital, machinery, and knowledge in order to produce outputs that consumers value. This essay will argue that the dominant institution of the post-AGI economy will instead be the “club”, an exclusive group of people that coordinates on desire and then commissions the necessary corresponding production.

Why do firms exist in the first place? Why not simply engage the market directly for all transactions?

Coase famously argued that firms formed due to transaction costs2. To produce goods, firms need to organize agents

cooperation

Many firms already form in the reverse order.

If every step in production was conducted through a separate contract, someone would have to search for suppliers, discover prices, negotiate terms, monitor performance, and coordinate all the resulting pieces. A firm internalizes some of these transactions under a common authority when doing so is cheaper than repeatedly buying them on the market.

The characteristic institution of industrial capitalism is the firm. A firm brings together labor, capital, machinery, and knowledge in order to produce something. Once the productive apparatus exists, the firm goes looking for customers. But we can imagine an institution that runs in the other direction: it begins with a stable group of people, figures out what they want, and then commissions the necessary production.

I will call this institution a **club**.

I do not mean a country club, or even necessarily a group formed for socializing. A club, in the sense used here, is a persistent association that organizes some portion of its members’ demand. Members give it enough information, trust, and authority to decide what should be purchased or produced on their behalf. The club may own factories, but it does not have to. Its durable asset is not a particular productive technology, but the group of people it represents.

The basic claim of this essay is that, as production becomes cheaper and more interchangeable, institutions organized around demand become more important relative to institutions organized around supply. This is speculative, and I do not expect firms to disappear. Chips, energy, models, logistics, and many physical goods may become even more centralized. The narrower claim is that the institution with the most durable relationship to a population may increasingly determine what gets produced.

This was especially useful in an industrial economy. Factories require large fixed investments in machinery, energy, buildings, management, and specialized labor. Once a factory exists, it needs to produce a large number of goods in order to spread out those fixed costs. Large-scale production therefore favored standardization: standardized parts, standardized worker roles, standardized schedules, and standardized final products.

The factory also pulled other institutions into the same shape. Schools produced comparable workers. Credentials made strangers legible to employers. National markets standardized money, measurements, contracts, and product categories. Mass media synchronized attention around a relatively small number of goods. The job converted a complicated human being into a more standardized productive role.

Once the firm had assembled this productive capacity, it still had to sell the output. The rough sequence was:

Advertising, branding, department stores, consumer credit, and broadcast media all helped complete the last step. This does not mean that firms simply invented arbitrary desires. People already wanted transportation, food, entertainment, status, and comfort. But firms had an incentive to translate these heterogeneous wants into demand for whatever their particular machinery could produce profitably.

Industrial society was therefore organized around a supply-side bottleneck. People wanted more food, clothing, housing, transportation, and manufactured goods than the economy could easily produce. The difficult problem was assembling the means.

Successive technologies reduced the cost of production and distribution. Industrial machinery reduced the cost of physical repetition. Global logistics expanded the market over which fixed costs could be spread. Software reduced the cost of calculation and administration. The internet reduced the cost of distributing information.

The internet also shifted power toward the institutions that controlled access to demand. Ben Thompson’s [Aggregation Theory](https://stratechery.com/2015/aggregation-theory/) describes how a company with a direct relationship to users can gain power over modular suppliers.3 Google does not write every webpage, and Facebook does not produce every post. They control the interface through which users discover supply, and suppliers must compete for access to those users.

AI may push this process another step. The internet made it cheap to distribute existing supply. AI makes many forms of supply cheaper to create, modify, and coordinate. It lowers the cost of software, design, analysis, administration, translation, instruction, research, customer service, and customization. Combined with robotics and flexible manufacturing, some of the same logic may eventually extend further into the physical economy.

The important change is not merely that AI can produce more stuff. It can handle more variation.

Historically, customization was expensive because a different contract, lesson, design, or service for every person required separate human attention. Standardization conserved cognition. AI makes some of this attention reproducible, allowing a common system to produce different outputs for different people without requiring an entirely separate organization for each one.

Imagine that an AI system can generate a thousand plausible curricula, insurance contracts, software packages, or housing plans. The expensive part may no longer be producing the options. The expensive part is deciding which option is trustworthy, coordinating it with other people, and committing enough resources to make it real. Possible supply scales computationally. Human attention, purchasing power, and commitment do not.

When this happens, the production sequence can run in reverse:

This is not entirely new. Preorders, insurance pools, labor unions, purchasing cooperatives, and consumer cooperatives already organize demand before production. The argument is that these forms become more general when productive capacity is abundant and easy to substitute.

A firm bundles means around an objective. A club bundles objectives into a purchasing mandate.

The firm asks: how should labor, capital, and expertise be combined to produce this thing?

The club asks: what do these people want to accomplish together, which parts should be purchased collectively, and who should produce them?

Individuals already pay costs on the demand side of the market. They must figure out what is possible, understand their own requirements, find suppliers, compare products, negotiate prices, monitor quality, manage risk, and sometimes coordinate with other buyers. This can be trivial for a bag of flour and extremely difficult for healthcare, education, housing, insurance, or a complicated piece of software.

A club internalizes some of these decisions. Its boundary is determined by the set of choices that members prefer to delegate rather than make independently. It grows when shared knowledge, bargaining power, risk pooling, common infrastructure, and stable commitment are worth more than the accompanying disagreement, bureaucracy, and governance costs.

Consider a professional club. It might provide health insurance, legal services, credentials, continuing education, software procurement, income smoothing, and access to projects. Members may still buy their own food and entertainment because the club has no special advantage there.

Or consider a parenting club. The same knowledge of its members could be useful for childcare, education, healthcare, housing, transportation, and recreation. From the perspective of a conventional firm, these are unrelated industries. From the perspective of the parents, they are all parts of the same life.

This is one reason a club might have strange economies of scope. A firm’s common asset is usually some productive capability. A club’s common asset is the population it serves.

There are several nearby ideas that are easy to confuse with this one.

James Buchanan’s economic theory of clubs deals with goods that sit between private and public goods.4 A swimming pool can be shared by multiple people, allowing them to divide the fixed cost, but too many members create congestion. The problem is to find the optimal number of people sharing a substantially predetermined good.

The club imagined here is not formed around one predetermined club good. It is a persistent institution that can express demand across multiple goods and repeatedly decide what should exist.

Henry Hansmann’s theory of enterprise ownership comes closer. Hansmann asks when an enterprise should be owned by its customers, workers, investors, or some other class of patrons. Customer ownership can reduce monopoly power, lock-in, and informational exploitation, but collective ownership creates its own monitoring and decision costs. A consumer cooperative makes sense when the savings in contracting costs exceed the costs of governance.

But the object in Hansmann’s analysis is still usually a particular enterprise: a customer-owned insurer, utility, store, bank, or apartment building. The club does not necessarily own the supplier. It may instead own the specification, the member relationship, the relevant preference data, and the right to replace the supplier.

The distinction is roughly:

Historical consumer cooperatives are the closest ancestors. Consumer movements already imagined federated buyers moving backward from retail into wholesaling, production, finance, and other services.5 The new question is whether AI relaxes one of the major constraints on these organizations.

Traditional demand aggregation usually required many people to want approximately the same thing. A purchasing cooperative could negotiate a lower price because thousands of members bought the same flour, fuel, insurance plan, or other standardized product. Cooperative ownership was easiest when the transaction was simple and members had homogeneous interests.

AI could weaken this constraint.

Consider a school organized by a group of parents. The parents may want to share facilities, accreditation, child-protection rules, teachers, social activities, and an educational philosophy. But they do not necessarily want every child to receive the same curriculum. An AI system could help translate the common layer into different lessons, schedules, exercises, and pacing for each student.

The same pattern applies elsewhere:

The common layer is aggregated while the final layer is individualized. We can call this **aggregation without homogenization**.

This would make the club more general than the traditional purchasing cooperative. The institution could be organized around a group of people rather than a single transaction, because the shared information about those people could be reused across many transactions.

There is an important limit here. AI may reduce the cost of translating a common objective into individualized outputs, but it does not make disagreement disappear. If the parents disagree about what education is for, or the patients disagree about how medical risk should be shared, then personalization does not solve the underlying governance problem. AI helps with variation inside an agreement; it does not automatically produce the agreement.

Another way to see the value of the club is to classify economic transactions along two dimensions. Does the particular seller matter? And does the particular buyer matter?

By “matter,” I mean that replacing the party would change the value or specification of the final output. The seller matters when the output depends on a particular brand, reputation, skill, or productive capability. The buyer matters when the output depends on the buyer’s identity, circumstances, preferences, or intended use.

This gives us a simple table:

| Buyer is interchangeable | Buyer is specific | |
|---|---|---|
Seller is specific |
Product: a novel, branded car, or designer object |
Service: therapy, tutoring, consulting, or commissioned architecture |
Seller is interchangeable |
Commodity: wheat, electricity, or a standardized fastener |
Fabrication: a print run, CNC part, mixed paint color, or AI-generated output |

These are ideal types, and real transactions can move between them. A famous chef turns a meal into a product by making the seller’s identity important. A standardized tax-preparation service pushes what was once a personal service toward a commodity. Branding can make an otherwise interchangeable seller appear specific, while technical standardization can make a previously specific seller replaceable.

I will use **fabrication** as the name for the fourth category. The buyer supplies the differentiating specification, while the seller supplies a generalized productive capability. A print shop does not decide which book should exist; it executes the publisher’s file. A CNC shop does not decide the shape of the part; it executes the customer’s design. The paint mixer does not care which wall will be painted; it reproduces the selected formula. The seller can be replaced without fundamentally changing the output as long as the specification survives.

Fabrication does not have to be physical. An AI system that produces a contract, curriculum, software tool, image, or research report from a sufficiently detailed specification occupies the same economic position. The model provides a general capability, but the particular output is defined by information coming from the buyer.

AI may greatly expand this quadrant. Ordinarily, making something specific to a buyer requires a specific human seller, which is why personalization takes the form of a service. A tutor learns about a student, a lawyer learns about a client, and an architect learns about a household. If AI makes this knowledge portable between producers, the buyer can remain specific while the seller becomes increasingly interchangeable. The transaction moves from service to fabrication.

Similarly, cheaper production does not necessarily turn every differentiated product into a commodity. If AI makes customization cheap, seller differentiation can decline at the same time buyer differentiation increases. The product moves diagonally into fabrication instead. We still receive different curricula, contracts, software, media, and physical objects, but the differences increasingly come from the demand side rather than from the productive organization.

This is the strongest reason to think that the differentiating value created by AI will accrue to clubs. The model or manufacturer may still capture infrastructure rents, especially when compute or physical production is concentrated. But the value that makes one output more suitable than another comes from knowing the members, forming a specification, resolving conflicts between their wants, and committing enough demand to act. Those assets belong to the club.

If the producer raises its price or falls behind technically, the club can take its specification and member relationship to another producer. If the club disappears, the organized demand, preference knowledge, and shared specification disappear with it. AI therefore makes the productive layer more replaceable while making the organized buyer layer more valuable.

Suppose a club does more than recommend products. Members give it permission to negotiate or direct spending in particular categories. The club now controls something economically valuable: recurring demand.

One possible metric would be **annual member spending under direction**. A club with one million members and authority over $3,000 of annual spending per member would direct $3 billion in demand without necessarily owning a single factory. Suppliers would compete for preferred access to that pool.

The competition need not only take the form of lower prices. Suppliers could offer custom features, stronger warranties, service guarantees, rebates, revenue shares, or even equity. If the club owns the relationship to the buyers and can switch suppliers, then the supplier is replaceable in a way that the club is not.

Some of the resulting demand rent could flow back to members through discounts, dividends, shared assets, or subsidized services. The club’s economically important assets would look less like factories and more like dues, deposits, subscriptions, purchasing mandates, long-term commitments, and member preorders.

These commitments may also be financeable. Imagine that 20,000 members commit to buy an electric vehicle satisfying a shared specification within a given price range. The club can solicit designs and bids, choose a manufacturer, and use the commitments to reduce the risk of financing production. Instead of a company raising money, building a product, and hoping customers appear, the customers partially appear first.

Demand is not literally collateral in every legal or financial system, and commitments can be weak or unreliable. But a binding preorder from a known population is less speculative than a market forecast. As the commitments become more durable, the demand side begins to acquire a balance sheet.

The club may then integrate backward. At first it negotiates prices on existing products. Later it retains the product definition: the design, quality standard, software interface, educational protocol, clinical standard, member data, service history, or brand. Manufacturers compete periodically to produce to the club’s specification.

This is vertical integration in the opposite direction. An industrial firm begins with productive capacity and integrates forward toward distribution and customers. A club begins with customers and integrates backward toward specification, financing, and selected productive assets.

A parenting club that expands into education, childcare, insurance, housing, media, and healthcare would look like an incoherent conglomerate from the supply side. From the demand side, the activities are coherent because they all serve the same members. The organization is unified by a population rather than a production technology.

This overlaps with Aggregation Theory, but there are two possible endpoints.

In the conventional aggregator model, users remain separate. A platform owns the interface, observes their behavior, directs discovery, and makes suppliers compete for access. As Thompson points out in his distinction between [platforms and aggregators](https://stratechery.com/2019/shopify-and-the-power-of-platforms/), the aggregator intermediates the relationship between suppliers and users.

AI could make this model extraordinarily powerful. A single platform could understand each user, generate personalized supply, steer choices, and capture much of the resulting surplus. This would be Aggregation Theory extended from distribution into production.

The alternative is for users to aggregate themselves.

An aggregator says, “I control access to these buyers.”

A club says, “We are these buyers.”

The legal details matter more than the branding. A company is not a club merely because it calls customers “members.” In the stronger version, members govern the relevant data, purchasing mandate, product specifications, and supplier-switching rights. Rebates and other demand-side rents accrue to the members rather than to an outside intermediary.

The important contest may therefore be between platforms that aggregate demand from above and institutions through which people aggregate their own demand from below. AI increases the value of knowing and coordinating customers, but it does not determine who will own that relationship.

The modern employer provides much more than wages. It often provides insurance, training, credentials, reputation, identity, social contact, and a career path. This bundle made sense when a productive organization needed a stable body of workers.

If production becomes more modular and project-based, the employer may become less capable (or less willing) to serve as a person’s permanent economic home. A professional club could take over some of the persistent functions while members work temporarily for many different producers.

Such a club could operate on both sides of the market. It represents members as consumers when buying insurance, housing, or software, and as producers when assembling teams or bargaining for projects. A mature professional club would combine some functions of a guild, benefits provider, reputation network, purchasing cooperative, and talent market.

The pattern is **durable membership, temporary production**.

Small clubs would still lack the scale required for catastrophic insurance, major infrastructure, large research programs, or negotiations with concentrated suppliers. They could federate, pooling capital, legal systems, technical infrastructure, and risk while leaving ordinary culture and governance with the smaller groups. The resulting structure might look something like:

people → clubs → federations → generalized production infrastructure

The clubs may also acquire physical form. Once they enter housing, education, care, and culture, they will need apartment buildings, campuses, clinics, workshops, event spaces, and perhaps whole developments.

This creates the company town in reverse. A company town gathers workers around a productive asset. A club town would begin with a group of people who want some particular form of life and then commission the physical environment needed to support it.

At some point the distinction between an economic association and a political institution becomes blurry. If a club collects dues, pools risk, owns assets, distributes benefits, accredits members, regulates conduct, and resolves internal disputes, then dues start to resemble taxes, benefits resemble welfare, and arbitration starts to resemble law.

The club would not have territorial sovereignty, but it could become the institution through which members experience education, insurance, work, culture, reputation, mutual aid, and social status. In practical terms, this may matter more to daily life than many functions of the state.

This is also where the theory becomes less utopian.

Strong clubs would compound. Members with high incomes, useful skills, or low insurance risk make a club more attractive. Better services then attract still more desirable members. Elite memberships could become informally hereditary even if they cannot legally be inherited. Class position might depend not only on how much money someone owns, but on which collective purchasing and risk pool is willing to admit them.

I suspect art would provide a particularly clear example. In my [functional account of art](https://demonstrandom.com/essays/posts/functional_theories_of_art/index.html), artistic taste helps coordinate groups and distinguish their members from outsiders. If everyone can observe the same work, some of its value as a distinction disappears. An elite club may therefore use its combined purchasing power not only to acquire art, but to acquire exclusive observation rights. It could place paintings in members-only collections, commission performances that cannot be recorded, restrict access to archives, or keep site-specific works inside private buildings. The product being purchased is not merely the artwork, but the right to be among the few people allowed to experience it.

Complete secrecy would defeat the status function, since outsiders would not know what they were missing. The club would instead want controlled visibility: enough publicity to make the exclusion common knowledge, but not enough access for outsiders to share the experience. We might see catalogue entries without images, reviews of performances without recordings, or photographs of the guests at an exhibition that never show the art itself. Even if copies become cheap, the distinction can move to authenticated presence: seeing the original, attending the event, or participating in the private conversation around it. The rich would not merely own different things. They could increasingly inhabit a partially private culture.

Clubs could also become intrusive. An institution that knows a member’s medical history, purchases, work record, social network, and preferences has many opportunities for surveillance and control. Governance could be captured by administrators or a narrow faction. A club that provides housing, income, insurance, and reputation could make formal exit possible while making actual exit ruinous.

The central legal question is therefore:

Can a person leave without losing their livelihood, insurance, reputation, housing, and accumulated rights?

A club economy would need some body of membership law analogous to labor and corporate law. Data portability, benefit portability, nondiscrimination rules, fiduciary duties, democratic governance, and rights of exit would become central economic institutions. The state might move upward, maintaining the common substrate, governing externalities between clubs, redistributing basic purchasing power, and making it possible for people to move between associations.

This theory does not imply that all production decentralizes. The opposite may occur at the substrate. Chips, frontier models, energy systems, logistics networks, robotic fleets, payment rails, and manufacturing systems may have enormous economies of scale.

The possible structure is a barbell:

The middle gets squeezed. A conventional producer with neither irreplaceable infrastructure nor a durable group of customers becomes a contractor. Surplus flows toward ownership of the generalized means of production on one side and ownership or governance of organized demand on the other. The infrastructure owner captures the rent on generalized production, while the club captures the value of buyer specificity.

How would we know if this was actually happening?

Ordinary communities, cooperatives, subscription businesses, and buying groups already exhibit pieces of the theory. The stronger empirical signature would be organizations that:

No single feature is decisive. The distinct institution appears when several occur together: a durable population governs demand, carries its specifications between suppliers, and captures the resulting rent.

The theory could also fail. AI may strengthen existing firms and aggregators so much that self-organized demand never becomes competitive. Governance costs may remain larger than the savings from collective procurement. People may prefer the convenience of an outside platform to participation in a member-owned institution. Or demand may remain too unstable to support financing. If organizations acquire detailed preference data but never acquire meaningful purchasing mandates or supplier-switching power, then what emerges is probably just better marketing, not a club economy.

Industrial modernity placed productive organizations at the center of the economy. Firms assembled scarce machinery, capital, knowledge, and labor, then searched for enough customers to justify the resulting output. People entered the system as workers on one side and consumers on the other.

If AI makes productive capacity more abundant and interchangeable, the scarce activity shifts toward deciding what should exist and assembling people willing to support it. Many services and products become forms of fabrication: the seller becomes replaceable, while the buyer and the buyer’s specification continue to matter. The persistent institution may then be the group that holds the demand rather than the company that temporarily supplies it.

This would be a deep reversal. Industrial modernity organized means and then searched for ends. The club begins with the ends and rents the means.

The theory of the firm explains how society organizes scarce means. The theory of the club explains how society organizes scarce ends.

I used AI to help rewrite and edit this essay from my notes and original draft.

Assuming goods are not monopolized by a single supplier.↩︎

Ronald Coase, [“The Nature of the Firm”](https://onlinelibrary.wiley.com/doi/full/10.1111/j.1468-0335.1937.tb00002.x), *Economica* (1937).↩︎

Aggregation Theory applies most directly to digital markets where distribution costs approach zero. The broader demand-side logic also appears in retail, payments, marketplaces, insurance, and other institutions that control access to customers.↩︎

James Buchanan, [“An Economic Theory of Clubs”](https://aike.smu.edu.cn/pluginfile.php/103748/mod_resource/content/1/An%20Economic%20Theory%20of%20Clubs%20Buchanan.pdf), *Economica* (1965).↩︎

Charles Gide’s *Consumers’ Co-operative Societies* discusses both federations of consumer societies and production conducted on behalf of consumers. A digitized copy is available through [HathiTrust](https://catalog.hathitrust.org/Record/005056079).↩︎