The Miller’s Thumb

A framework that starts in the mind has to end, if it’s honest, at the machine. The earlier essays argued that a free order defends itself by staying diverse and keeping the exits real — that a rule anyone can walk away from can’t be captured. But being free to walk away from a rule is no freedom at all if the machine that runs it belongs to one owner, and you can’t walk away from the machine. This essay follows the arrow down to that machine — to the computing power, the energy, the money everything now rides on — and asks the oldest political question in its newest clothes. Who owns the mill?

A theory caught in the act

For most of its length this framework has talked about minds and meanings — how a self comes to care about things, how a group comes to know its world, how a diverse society keeps its shared center thin. On its face, it’s been a theory of ideas and the rules that carry them. But somewhere in the middle of defending that kind of society against its hardest objection, it let slip something it hadn’t quite meant to say. It said that the freedom to leave — the exit that keeps any shared center honest — has to be real, and real all the way down. Not just the freedom to say no to a rule, but the freedom to actually go and run a different one, on your own, without asking the owner of some machine for permission.

That last part isn’t a claim about ideas. It’s a claim about things — about who owns the presses and the wires and the engines that ideas get printed and run on. The instant the framework insisted on real exit, it quietly stopped being only a theory of rules and became a theory of political economy. It had been doing political economy in its sleep. This essay is the framework waking up and taking a look at its own hands.

The waking isn’t comfortable, because the world of machines doesn’t yield to a good argument the way a conversation does. It has weights and costs and owners. But the framework doesn’t need a new principle to meet it. It just needs to carry the principles it already has — diversity, exit, being able to undo, difference all the way down, the commons instead of the crown — one floor lower than it’s gone before. Down to the metal.

The mill

Start with a picture older than any factory. For a thousand years, the surest power in a European village wasn’t the castle on the hill — it was the mill down by the stream. The lord owned the mill, and the law of the manor forbade you to grind your own grain at home. You hauled your wheat to his mill, and you left a share of it behind as the price of grinding: the miller’s toll. There was even a proverb about it — an honest miller has a golden thumb — meaning that even the honest ones pressed a thumb on the scale, and nobody really believed theirs was honest. You couldn’t opt out. Bread had to be flour, flour had to pass through the one machine in the valley, and the machine had an owner. That was a quieter, deeper power than the sword ever was. The sword takes now and then. The mill takes a little from everybody, every day, forever, just by standing between the grain and the loaf.

This is the shape of the problem the framework now has to face — and it wears a modern name that hides how old it is. The earlier essays said: keep the center thin, keep the rules easy to copy and walk away from, and no one can seize the whole thing, because grabbing a rule anyone can copy just means watching everyone copy it and leave. True enough — of the rule. But a rule gets ground, like grain, on a machine. Today the machines that matter most are the giant engines of computing, the data that feeds them, and the power that runs them; tomorrow it’ll be something else; the point is general, and the mill is just its plainest face. You can be perfectly free to copy the rule — the model, the design, the recipe — and perfectly unable to run it, because running it takes a mill you don’t own and can’t build. And then your freedom to leave is the freedom to walk out of the valley into silence. A diverse order where every part depends, for the power to do anything at all, on one owner’s machine already has a lord. It’s just declined to call him one.

So the honest exit the framework demands can’t stop at the rule. It has to reach the mill. And that turns the whole question — from one about laws and rules into one about ownership, money, and power. Which is to say, into political economy, whether the framework likes it or not.

The law that grinds one way

Before we can ask how to hold the mill in common, we have to see why the ordinary way of holding it pulls so hard the other way — why the machines keep collecting into single hands as if by gravity. Part of the answer is cost, and we’ll get there. But part of it is written into law, and it’s worth saying plainly, because it’s the counter-dynamic — the framework’s word for coherence bought by narrowing, by shrinking the world you answer to until a single number can stand in for the good — wearing the most respectable suit it owns.

Think about what an ordinary company is legally for. Under the rule that governs most of them, the directors owe a duty to make the most money possible for the shareholders. That sounds neutral, even sensible. But look at its shape. It’s a legal order to serve one group, measured by one number, and to treat everything else — the workers, the neighbors, the air, the far future — as a cost to be trimmed or pushed off the books where it won’t get counted. It is, in plain words, a legal command to narrow: shrink the circle of concern down to one owner and one number, and shove the rest outside. The framework has a name for that move in the abstract, and warns against it everywhere from the cell to the whole civilization. Here it isn’t a temptation to resist. It’s a duty the courts enforce.

That’s why the hopeful advice — “just build better companies, run by better people” — doesn’t reach the problem. A good person inside that legal shape is still bound by it; the shape narrows whether the person wants to or not. If you want an institution that widens instead of narrows — that answers to more of the world rather than less — you can’t get there by pouring good intentions into a container built to pour them right back out. You have to change the container. And that’s the first shovelful of the real work: not better milling, but a different arrangement of who owns the mill and what it’s bound to do.

The false choice

When people first see this, they grab for one of two single answers — and each one fails in a way worth learning from.

The first is law: write a new kind of company charter, legally bound to widen — to weigh the neighbors and the air and the future, not just the owner and the number. Pass the statute, spell out the duty, and let the courts enforce widening the way they now enforce its opposite. The second is code: forget the law, which is slow and easily captured, and route around it completely — build the machines and the money on open, shared systems that no government and no owner controls, so that ownership itself dissolves into a network nobody can grab.

Each is half right, and each fails on its own in exactly the way the framework would predict. Pure code — the dream of routing around law with clever shared systems — is all edges and no center. We’ve watched it run. The last big wave of “unstoppable, ownerless” digital money re-collected, within a few years, into a handful of enormous exchanges and a few giant holders: a fresh set of lords who’d simply skipped the paperwork of lordship. A network with no thin center to keep the exits open doesn’t stay spread out. It drifts back into the same few hands, because nothing’s holding the door open. Getting rid of the law didn’t get rid of the power. It only got rid of the accountability.

Pure law fails from the other side. A single new charter, however wise, is all center and no edges: captured the day it’s written by whoever holds the pen, unable to keep up with machines that change faster than statutes, and — worst, by the framework’s own lights — impossible to undo. To bet the whole order on one untested legal form, passed everywhere at once, is to make exactly the kind of can’t-take-it-back move the last essay called the gravest danger of all. A process whose strength is fixing its own mistakes must never stake everything on a step it can’t reverse.

So the two answers aren’t rivals. They’re the thin center and the thick edge of one design — the same shape the framework has been drawing all along, now asked to hold at the level of machines and money. Law supplies the lasting purpose that keeps a thing bound to widen. Code and its cousins supply the spread-out, walk-away-able custody that keeps that purpose from being captured. Neither one alone builds a common mill. The join does.

The pairing

Here’s the join, and it’s less exotic than it sounds, because both halves already exist and have already survived contact with the real world. Being able to undo counsels using tested tools before inventing new ones — and these are tested.

The first half rebinds what the mill is for. There are legal forms, in use right now, that already free a company from the duty to narrow. There’s the steward-owned firm, held in a perpetual-purpose trust — a setup where nobody can sell the company or squeeze it for a windfall, because its ownership is locked to its mission rather than to a market. The outdoor-clothing company whose founder handed it to a trust and announced that the earth was now its only shareholder is one recent example; older European firms — a great engineering company, a famous lens-maker, a big pharmaceutical foundation — have been held this way for generations, and haven’t withered for it. There’s the benefit corporation, legally allowed to weigh purposes the ordinary company has to ignore. There’s the cooperative, owned by the people it serves instead of by faraway money. Each of these is a container built to hold context rather than pour it out. The scarce thing isn’t a new invention. It’s the will to use the ones we already have.

But — and this is the hinge of the whole essay — rebinding what the mill is for does nothing about who can grab it. A steward-owned mill with a single, kindly owner is still a mill with a single owner: a nice lord, not the end of lordship, and niceness isn’t a structure. We’ve watched this fail in real time. A famous artificial-intelligence venture was built exactly this way — a mission-locked nonprofit board put in charge, on purpose, to keep the whole enterprise tied to the wider good. That was the legal half of the pairing, and only the legal half. And when the crisis hit, the sheer gravity of the machines — the computing power and the money the mission depended on and didn’t itself own — simply flowed around the board’s authority over a single weekend. The purpose was locked; the power wasn’t spread out; and unlocked power runs downhill to whoever owns the engine. A legal shell without the spread-out custody is half a bridge, and half a bridge carries nobody across.

So the second half spreads out the power. Where the mill is a resource that genuinely can’t be scattered — and some can’t; the biggest computing engines, like a great dam or a rail trunk line, really do get cheaper the bigger they are, which pulls them toward bigness — the answer isn’t to crown the owner but to hold the thing as a commons, with its governance shared out among the many who depend on it. Not one hand on the switch but many, no single one of which can throw it alone; keys held by independent parties who don’t share a boss, so that using the mill for anything serious takes a real agreement no single actor can fake or force. This is the modern echo of the oldest commons there is — the shared irrigation ditch, the common pasture, the village mill run by the guild instead of the lord — and there’s a deep body of study behind it: how communities have governed the resources they all depend on, for centuries, without either handing them to a private owner or handing them to a distant state. If you can’t scatter the resource, scatter the control of it.

The pairing, then, is a legal shell that locks in the purpose, married to a spread-out, many-keyed custody that locks out capture. And the place to build it isn’t a summit of nations drafting one universal charter — that’s the can’t-undo bet all over again — but small, where a real shared resource already sits: a pooled research cluster, a shared body of data, a jointly held trust of model weights, run by a cross-party compact and graded not by some regulator’s blessing but by whether it actually works and lasts. Stand one up at that size; let survival grade it; and if it widens rather than narrows and holds together, others will copy the pattern — and only the proven pattern, earned at the edge and never decreed from on high, should ever harden into a broad law or a standing institution. There’s a name floating around for the biggest version of this: a public, commonly governed computing commons for science, modeled on the great shared physics labs that many nations pay for and no nation owns. Whether or not that exact thing gets built, it’s the right shape — the mill held by the town.

Paying for the mill without selling it

Which leaves the hardest, most physical question of all — the one that turns every fine intention into a joke if it has no answer. A great mill costs a fortune to build: the engines, the buildings, the long contracts for power. Where does a common mill get that fortune without letting the money that builds it become the new lord? Capital, offered at that size, rarely shows up without a hand reaching for the steering wheel. So how do you take the coin and keep the keys?

The mistake buried in the question is the idea that the goal is to keep capital out. You can’t keep it out — the engines are real and cost what they cost, and thrift isn’t a building material. The framework’s move isn’t to refuse the money but to cut the return on it loose from control over the purpose — to pay capital well on the one hand while denying it the wheel on the other. Six threads do that cutting, and each is a tested tool, not a new gadget.

The first is the cut itself, made structural: capital gets a real, even generous, financial return — and no vote. The tools for this are ordinary — a capped return, a slice of revenue, a stake that pays out but carries no say in governance. The keys to the mill just aren’t for sale: made non-transferable to capital by the charter and by the custody setup both, so there’s no door for money to buy its way through rather than a guarded one. The second thread turns that cap into a filter: a hard ceiling on the return repels exactly the kind of money that wants to own and maximize, and keeps the kind that’s happy to be paid fairly and leave the purpose alone. The ceiling isn’t a sacrifice — it’s how you screen for compatible money before it’s inside the walls.

The third thread aims the framework’s own horror of concentration at the balance sheet: draw the money from many independent places — patient philanthropy, public and civic research funds, the members’ own contributions, revenue from real services, mission-locked loans — so that no single funder is the one whose leaving would kill you, and so no single funder can demand a key as the price of staying. Spread-out funding is real exit one more time, now in the form of money. The fourth thread lowers the whole bar by preferring access over ownership: you don’t have to own one giant mill if you can knit together many small ones — pooled, shared, rented capacity, a co-op of members each bringing a piece — which is both cheaper and, by its nature, harder to capture. The fifth thread is honest about the tier where the resource really is a natural monopoly, too big and too lumpy for any co-op: there the answer is frankly public or mutual, paid for the way nations already pay for the shared laboratory or the public road — built in common because it’s used in common, and governed as a commons rather than sold to a champion.

The sixth thread is the deepest, and it answers the whole worry at its root: arrange the fundraising so that survival, not desperation, sets the terms. The reason capital can demand the wheel is that it’s asked for everything at once, up front, by someone with no leverage — and a fortune raised cold forces you to a source big enough to dictate. So don’t be there. Start small, in the shallows, on rented and shared capacity, and prove the thing works — prove it widens, prove it lasts, prove it catches its own mistakes better than the giant does. Proven survival flips the bargaining power: once the common mill is the thing capital and talent need more than it needs them, the money comes to you, on your terms, because you’ve already taken away the risk it couldn’t. The answer to “how do you raise a fortune without selling the keys” is, in large part, don’t try to — until you’ve made the mill worth more standing free than captured. That, too, is the undo rule: never make the can’t-take-it-back concession while you’re weakest.

The honest edges

Three of them, in the discipline of the earlier essays, because the account won’t claim more than it has.

First, there’s a tier where even all this may not be enough. Some frontier engine might turn out so huge that only a sovereign — a great state — can fund it at all, and then the question stops being clever institutional design and becomes naked statecraft: which state, under whose governance, watched by whom. The framework doesn’t pretend to some mechanism that dissolves that. It offers only its steady direction — hold it as a commons, spread out the governance, refuse the single crown — and no guarantee the direction wins. Where the machine can’t be built except by a giant, the fight is political all the way down, and the essay’s honesty is to say so rather than to wave a clever system at it.

Second, the incumbent won’t stand still. The legal form that mandates narrowing isn’t a sleeping mistake; it’s a live, concentrated, well-defended institution with every reason to stay the only game in town. The common mill has to out-compete the lord’s mill while the lord still owns the valley — and there’s no promise it wins the race before the machines collect past the point where forking is even possible. Saying that plainly is the price of not sliding into the technologist’s easy faith that good design always wins. It doesn’t always win. It only gives the best odds actually on offer.

Third, and deepest: this is the one corner of the framework where the work isn’t a piece of reasoning but a fight. Everywhere else, the argument can be carried by thinking clearly. Here, thinking clearly only tells you where to push; the pushing is done with money and law and coalitions and time, against interests that will push back hard. The framework can hand you the compass — spread it out, keep the exits real, keep it undoable, keep it diverse, all the way down to the copper and the concrete — but it can’t hand you the valley. That has to be built, by people, out in the open, under resistance. A theory of political economy that promised otherwise would be lying, and lying is the native tongue of the very thing this whole book stands against.

The arrow reaches the metal

Step back and see what happened. The framework didn’t pick up a new principle when it came down to the machines. It carried its old ones lower. Diversity became many owners instead of one. Exit became the freedom to run your own engine, not just to say no to a rule. Undoability became the refusal to bet the order on a charter you can’t take back, and a brake set in front of the purchase you can’t reverse. The commons instead of the crown became, quite literally, the town’s mill instead of the lord’s. It’s the same shape the framework has drawn at every scale — the cell, the mind, the diverse society — now pressed into copper and money and law. The arrow didn’t change direction when it reached the metal. It just reached the metal.

And that’s the whole of it — less a finished blueprint than a place to put the first shovel. The mill can be held in common; it has been before, in older forms, by communities who worked out for themselves how to keep the one machine everybody needs from becoming one master everybody serves. Doing it again, in the age of thinking engines, isn’t a new kind of task. It’s the oldest task there is — keeping the thing everyone needs from becoming the thing one owner holds over everyone — carried one turn further out, to the machines the next turn of everything will run on.

An honest miller, they said, has a golden thumb. The answer was never to go find an honest miller. It was to own the mill together.

Sources & further reading

This essay engages its sources directly rather than through the book’s per-chapter endnotes. A citation-level pass is still owed.

The mill, the monopoly on the indispensable machine. On the manorial milling monopoly — the banalité and the lord’s mill, oven, and winepress — and the long conflict over the right to grind: the economic and social history of medieval milling (Richard Holt, The Mills of Medieval England, 1988; John Langdon, Mills in the Medieval Economy, 2004). The proverb of the miller’s golden thumb runs at least to Chaucer.

The firm that is legally bound to narrow. On shareholder-primacy as the governing doctrine and its critics: the shareholder-versus-stakeholder debate (Milton Friedman, “The Social Responsibility of Business Is to Increase Its Profits,” 1970, as the sharpest statement of the doctrine; Lynn Stout, The Shareholder Value Myth, 2012, and Colin Mayer, Prosperity, 2018, as its leading critics); Dodge v. Ford (1919) as the case usually cited for the duty.

Stewardship and purpose-locked ownership. On steward-ownership, the perpetual-purpose trust, and foundation-owned firms: the Patagonia transfer to the Holdfast Collective and Patagonia Purpose Trust (2022); the long-standing foundation ownership of Robert Bosch, Carl Zeiss (the Carl Zeiss Stiftung), and the Novo Nordisk Foundation; the benefit-corporation and B Corp movement; and the steward-ownership literature (Purpose Foundation).

Distributed power, and the trouble with pure protocol. On the re-concentration of “trustless” systems: the empirical literature on cryptocurrency wealth and exchange concentration. On the failure of a mission-locked board without distributed custody: the OpenAI governance crisis of 2023, taken as an illustration, not a verdict on any party.

Governing a shared resource without a lord. The foundational work on the commons: Elinor Ostrom, Governing the Commons (1990), and her design principles for enduring common-pool-resource institutions — monitoring, graduated sanctions, nested and polycentric governance — drawn from irrigation systems, fisheries, forests, and pastures; with Garrett Hardin’s “The Tragedy of the Commons” (1968) as the thesis she answered. On exit as the discipline on captured institutions: Albert Hirschman, Exit, Voice, and Loyalty (1970).

Financing without surrendering control. On separating economic rights from control rights: the steward-ownership financing literature and instruments such as capped, non-voting, and self-redeeming capital, revenue-based financing, and mission-locked debt. On shared public research infrastructure as the model for a natural-monopoly tier: the great multinational scientific facilities funded in common (CERN as the paradigm) and public research-computing initiatives.

Within the framework: The Soft-Shell Hour (defended openness, and the exit-must-be-real claim this essay follows down to the machine); Coherent Pluralism (the thin center and thick edges, here in copper and capital); Moloch, Formally (the trap that concentration springs, and why the sovereign is not the escape); Coherence at Scale (nesting and the network of networks, the structural engine beneath a commons); and Rights, and the Widening We (compressed coherence institutionalized — of which a commonly held mill is one more instance).