The Diamond No Longer Waits

The ever accelerating progress of technology ... gives the appearance of approaching some essential singularity in the history of the race beyond which human affairs, as we know them, could not continue.

The production function has not become a constitution. That is the first fact to preserve after following computation from the power plant into the institutions where its output acquires consequence.

Electricity can now be organized into forms of cognitive work that differ profoundly from one another. Some vanish in use. Some persist as reproducible capability. Some satisfy protocol conditions that help establish an ordered settlement history. Computational systems can also participate in choosing experiments, designing components, and allocating resources toward successor systems. None of this makes energy a measure of value or capital identical to labor. It does alter the old dependence of capital upon human cognition at every point where a possible investment had to be found, compared, and directed.

That alteration is enough.

For much of industrial history, machinery amplified purposes formed elsewhere. An engine waited for a throttle. A ledger waited for an entry. A portfolio waited for judgment about what should be bought. The waiting did not make labor morally deserving by itself, but it gave human participants leverage inside production because conversion could not proceed without their situated work.

Computation has begun to remove some of that waiting. A model can search a design space, propose a course of action, evaluate alternatives, and help construct the next tool. Its activity remains framed by objectives, data, permissions, infrastructure, and institutions supplied by people. The point is not that capital has awakened. It is that a growing part of the judgment once supplied during production can be embodied in a process owned and directed as capital.

Cheaper cognitive work can enlarge what an undertaking attempts without giving it control over the means it needs. A model provider whose capability remains difficult to replace, an infrastructure owner controlling a needed site, or an institution authorized to let an act proceed can influence what another undertaking is able to attempt. That power may command a payment without yielding enough to repay the investment made to acquire it. The service can remain useful, and its customers dependent, while the owner's claims lose value. A change in ownership can preserve the service by changing which claims its proceeds must support.

Soddy distinguished the arithmetic by which a debt increases from the work by which wealth is maintained.¹ Productive capacity can grow, and finance can help build it; adding to what is owed does not perform the construction. Claims upon future output can be created in an instant; grids, fabs, skilled organizations, and lawful authority still take time. Even digital settlement depends upon physical machines and recognized rules. The apparent weightlessness of the claim rests upon a world that is stubbornly material.

The center of this volume has been a conversion, but not alchemy in the old fraudulent sense. Physical throughput is made productive through computation when a selected output performs an office that people and institutions recognize: answering a query, improving a capability, admitting a block, coordinating an act. Expenditure is necessary to these processes and never sufficient to establish their worth. The economic object emerges from the relation between material work and a rule, market, or institution able to use what the work produces.

Bitcoin made that relation unusually visible. Proof-of-work is not ornamental waste protecting value formed elsewhere, nor does each additional joule confer a corresponding increment of price. Successful work helps order a history and authorize issuance under public rules. A market then decides what claims within that history are worth. The achievement is constitutional: a protocol made costly computation one condition of producing and defending a permissionless settlement state.

Machine learning arranges the conversion differently. Training can leave behind a capability that is cheap to copy, but neither training expense nor benchmark performance guarantees a useful deployment. Value depends upon whether the capability enters an activity, survives evaluation, receives permission, and changes an outcome. A model held outside those relations is not stored economic value. It is an unrealized possibility with a maintenance bill.

The industrial novelty lies in joining these possibilities to allocation. Systems now help decide which configurations to test, which resources to commit, and which successor designs to pursue. The loop remains partial. It depends upon human principals, physical supply chains, capital markets, law, and acts of authorization. But partial recursion can be economically decisive long before it becomes autonomy. A factor need not reproduce its whole world to alter bargaining power within it.

This is why the volume ends at the gates. Ownership of energy and fabrication determines who can attempt the conversion. Control of data and distribution determines which capability can be made useful. Licenses, insurance, and institutional recognition determine which outputs may become acts. Those positions can protect the public, collect rents, or do both at once. Their design will shape not only the level of output but who possesses standing in the productive order.

The central distributional question is therefore not whether computation will produce a surplus. It is which claims will be recognized before that surplus arrives. Ownership can be broadened or concentrated. Authorization can remain an accountable office or decay into ceremonial toll-taking. Liability can follow control or be deposited upon the last visible human. States can preserve open competition, direct capacity toward public ends, or combine strategic secrecy with private monopoly.

None of these choices is written in the thermodynamics. Physics supplies limits and costs. Engineering supplies conversions. Markets register some forms of demand. Institutions decide what may count as property, authority, injury, and repair.

In the prologue, the imagined hunter lifted a diamond whose lattice could outlast his judgment. Computation reverses one part of the scene. The durable structure can now participate in evaluating where resources should go next. It still cannot establish the legitimacy of the purpose, confer authority upon itself, or remain as an answerer after every act performed in its name.

The diamond no longer waits. The institutions around it must decide whether that new capacity will widen human agency or merely perfect the machinery by which agency is owned.