The Counter-Thesis

The first principle is that you must not fool yourself—and you are the easiest person to fool.

A customer of a computational service may obtain a useful saving while the economy records little improvement. An employer may acquire more control without learning how to produce much more. A transformation in the organization of work can succeed on one party's terms and fail the larger promise made for it.

The difficulty reaches settlement as well. Businesses need not choose the same rail to make use of computational services, and an open base protocol can support an intermediary that its users cannot afford to leave.

Automation Without the Dividend

Acemoglu and Restrepo show why displacement cannot be used as evidence of productivity.¹ Automation can substitute for labor in tasks without producing an improvement large enough to justify the social cost. Firms may automate because it strengthens control, follows managerial fashion, or transfers income, even when aggregate output scarcely moves.

Learned inference does not escape this objection by being more general than an industrial robot. Generality expands the range of tasks on which mediocre substitution can occur. If error, integration, review, and rework consume the apparent gain, cheap cognition may chiefly become a cheaper way to supervise or displace people.

V/C cannot rescue the thesis by relabeling every failure as expensive verification. Task choice, capability, demand, organizational design, market power, and regulation may explain more. The ratio survives only if it adds explanatory power after those variables enter.

The failure condition is demanding and ordinary. Widespread deployment accompanied by weak gains in measured output, quality, or resource use would reduce Factor Prime from a production transformation to an important change in managerial technique. Measurement problems cannot be invoked forever. A claimed dividend has to become visible somewhere resistant to the claimant's own accounting.

Productivity May Remain Slow

Even useful systems may fail to alter the aggregate growth regime. Reorganization takes time. Complementary capital can be scarce. Improvements may concentrate in services whose quality is difficult to measure, or be offset by congestion, energy cost, and rising complexity. The history of new tasks also supplies no guarantee that reinstatement will keep pace with displacement.²

Part III established partial recursion, not an accelerating law. Models can assist code, evaluation, and design while harder discoveries, fabrication, experiments, power, and institutional approval keep the overall loop slow. If the gains remain local, capital may still reorganize around computation without producing the civilizational expansion implied by the strongest thesis.

This objection changes the interpretation of investment. Heavy spending can be evidence of expected transformation, strategic rivalry, or a competitive race that destroys returns. Construction does not certify productivity. Market capitalization does not certify construction. The book must allow a future in which computational capability improves spectacularly while growth disappoints.

Settlement Is Plural

Bitcoin demonstrates one constitutional conversion, not the only viable settlement system. Nakamoto's protocol makes qualifying work part of block admission and cumulative-work history.³ Stablecoins, bank deposits, central-bank money, proof-of-stake networks, commodity claims, and platform credit solve different problems and expose users to different authorities.

For many transactions, price stability, legal recognition, reversible error, and integration with existing accounts matter more than resistance to an issuer. A business operating inside one jurisdiction may rationally prefer a regulated bank or stablecoin. A claimant who needs a court to reverse fraud may value the very institution that a permissionless design removes from the validation path.

Proof-of-work retains a distinctive office. It joins open verification and chain selection to a cost incurred outside an issuer's discretion. Yet practical use can reintroduce exchanges, custodians, pools, software providers, and legal dependencies. Budish's analysis further shows that security expenditure and the value exposed to attack are mutually conditioning.⁴ Energy does not purchase independence once and for all.

The Joule Standard therefore becomes local. Mining is a relevant outside option only for an operator able to switch suitable power and capital into it. Bitcoin may be indispensable for some threats and needlessly costly for other uses. If regulated or alternative rails satisfy most machine-mediated commerce without intolerable capture, the book's settlement claim contracts to the margin where discretionary denial is itself the central risk.

Open Protocol, Closed Exit

A protocol can be open while ordinary access consolidates around intermediaries. Scale economies, convenience, regulation, liquidity, and technical difficulty can make a formal right to exit practically remote. In that world, the base rules remain permissionless and the lived system does not.

The answer cannot be that anyone could run the entire stack. Exit has constitutional force only when a person or institution can exercise it without forfeiting assets, identity, reputation, or the ability to continue operating. Portability, interoperable data, multiple providers, and fork rights may strengthen that option. None guarantees it.

Nor is intermediation itself the failure. A custodian, insurer, exchange, or specialized interface may reduce risk and make participation possible. The failure begins when dependency becomes noncontestable: the intermediary can change terms, deny access, or retain the user's accumulated standing while no independent forum or practical alternative can answer the act.

This objection changes the political economy of the final movement. Control does not reside only in model weights, power plants, or base protocols. It can collect in the convenient layer everyone is nominally free to leave. Openness at one level can coexist with sovereignty at another.

What Survives

A firm that can buy work it once had to organize through employment acquires a new choice. Its gain depends on what it can obtain for its expenditure; the gain to the economy depends on what improves in production, rather than merely changing hands. If the larger dividend fails to appear, the transfer of power within the workplace can still be extensive. People whose work has been displaced do not recover their positions merely because the wider promise of prosperity fails.

A useful intermediary presents the same difficulty from the customer's side. Its convenience may be worth paying for, and its protection worth retaining. As more of the customer's activity comes to depend on it, however, the initial decision to enter tells us less about the practical freedom to leave. A choice that made participation easier can help establish the terms on which participation will later be allowed. The objection concerns that change in the relation, not the customer's failure to appreciate an open protocol.

The attraction of a scarce asset begins with someone's need for it. How that need becomes an owner's income depends on the terms under which access is supplied, and those terms can be contested. Capital can now own capability that performs part of management and design. The resulting choice of what to build next enters a conflict among people whose assets, work and authority need not serve the same end.