What Would Refute the Thesis

A position that names its failure conditions

A structural argument that cannot say what would prove it wrong is rhetoric. This appendix states, premise by premise, what the Discontinuity Thesis would have to lose to be wrong. It is the standard at which the argument can be attacked, and the standard against which entries for the Discontinuity Prize are judged.

The four premises

Premise One. Unit Cost Dominance is being crossed. AI plus verification produces a substantial and growing fraction of professional cognitive tasks at lower unit cost and equal or better quality than human-only production. The benchmarked domain overlaps heavily with the cognitive work that sustained middle-class absorption.

Premise Two. Task dominance propagates through interface collapse. AI is not confined to isolated task outputs. Once models operate across software interfaces, task-level dominance becomes workflow recomposition, which suppresses hiring, breaks training ladders and reduces mass absorption without needing whole occupations to disappear at once.

Premise Three. The propagation cannot be restrained as a competitive equilibrium. No actor or coalition operating within current institutional structures can prevent the propagation at scale. Restraint is dominated at the worker, firm, sector and state levels under the competitive conditions that obtain.

Premise Four. No available preservation route restores the wage-demand circuit. Category-based regulation fails as circuit defence, because the assistance-replacement boundary dissolves inside workflows and AI as capital in motion migrates across legal categories. Structural alternatives preserve consumption rather than productive necessity. Friction slows propagation without restoring general-purpose cognitive labour as mass scarcity.

Conclusion. The wage-demand circuit is no longer self-reproducing under the new technological condition. Postwar wage capitalism has lost its reproduction mechanism. What replaces it is a successor system, and the open question is which successor arrives, who designs it, and on whose terms.

What would refute Premise One

A refutation of Premise One would show that AI plus verification is not crossing unit cost dominance for a substantial fraction of professional cognitive tasks. It would have to engage the quality condition, the cost condition or the workflow condition.

A quality refutation would show that benchmark performance does not generalise to deployed quality, that the benchmarked domain does not overlap meaningfully with mass cognitive labour, or that capability has stalled in a way that prevents further crossover. Pointing to residual work AI cannot do, such as care, embodied work or roles that rest on trust, does not by itself refute the premise. The thesis already grants those exceptions and treats them as residual scarcity rather than mass absorption.

A cost refutation would show that verification, integration and oversight costs systematically consume the whole raw cost advantage for the relevant class of tasks. That requires engaging the verifier-cost arithmetic in Essay 1. Showing that one particular deployment costs more than the human baseline does not refute the premise, which concerns the trajectory and the class of tasks rather than every individual deployment.

A workflow refutation would show that the deployed cost of AI plus verification systematically fails to undercut human-only production in the benchmarked domain. The strongest version would draw on enterprise deployment data showing that AI deployments routinely reach total costs equivalent to the human-only baseline.

What would refute Premise Two

A refutation of Premise Two would show that task-level unit cost dominance does not propagate into workflow recomposition. It would need to show that software fragmentation remains a durable labour moat, that computer-use and tool-use capabilities do not generalise into enterprise workflows, or that the cost of integrating AI across interfaces routinely consumes the task-level advantage.

Showing that jobs contain residual human tasks does not refute the premise. The premise does not require whole-job replacement. It requires enough workflow recomposition to suppress hiring, break training ladders and reduce mass absorption.

The strongest refutation would be evidence that interface-level capability has stalled in a way that prevents propagation: frontier models that cannot reliably operate desktop software, navigate browsers, handle files or sustain multi-step work across applications, despite continued investment and progress at the task layer. OSWorld-Verified, BrowseComp and Tau2-bench Telecom track the relevant capability. A reversal of their trajectory across independent benchmarks would be a Premise Two refutation.

What would refute Premise Three

A refutation of Premise Three would identify an actor or coalition that can credibly restrain the propagation, and specify how that restraint becomes a stable equilibrium rather than a temporary delay.

It would have to engage the Multiplayer Prisoner’s Dilemma at all four levels. Worker-level adoption happens below the level coordination can see. Firm-level deployment is competitively dominant. Sector-level coordination is undermined by adjacent entrants and platform substitution. State-level coordination is undermined by competition between states.

Citing historical precedents for coordination, such as Basel, Paris, the WTO, nuclear non-proliferation, the chemical weapons convention or the Montreal Protocol, does not refute the premise unless the refutation explains why the conditions that made those precedents work hold for AI. Essay 3 makes the comparison directly.

Citing friction, whether regulation, liability, integration cost or professional gatekeeping, does not refute the premise either. Friction modulates timing. Essay 7 addresses friction as rescue and shows that drag is not equilibrium.

The strongest version would identify an enforcement mechanism that survives the competitive structure. That means showing how defection can be observed and punished when it is internal, diffuse, continuous and highly rewarded, and when the enforcers are also the actors most exposed to the cost of restraint.

What would refute Premise Four

A refutation of Premise Four would specify a mechanism that preserves the wage-demand circuit through the propagation. Preservation means maintaining mass productive necessity for the majority of working-age adults at socially sustaining wages, where subsidy, protection, artificial scarcity, make-work or political mandate are not the main source of the wage relation.

Postwar capitalism always had stabilisers. The welfare state, public employment, industrial policy, collective bargaining, public procurement and unemployment insurance all moved purchasing power outside the immediate wage relation. The question is whether such stabilisers supplement a labour market that remains primary, or replace it. Preservation requires the first. The thesis claims the technological condition has shifted so that the second now holds.

The refutation must engage all three routes together: regulation (the within-face Sorites problem and across-face Categorical Recursion), structural alternatives (the synthetic wage problem), and friction (the argument that friction modulates propagation).

A regulatory refutation would show that some category-based instrument can defend the assistance-replacement boundary as a defence of the wage-demand circuit, despite the Sorites gradient inside every use and despite the migration across categories that the deployed object shows in response to the category chosen. Regulations that succeed at other purposes, such as fraud, safety, liability or documentation, do not refute the premise. The thesis is scoped to circuit defence. See Essay 4 and Essay 5.

A structural refutation would show that direct compute ownership, mandatory wage-share redistribution, public deployment or sovereign AI funds preserve the wage-demand circuit, rather than producing a successor that delivers consumption without productive necessity. It must engage the synthetic wage distinction in Essay 6. A wage maintained by political obligation after productive necessity has ended is a transfer routed through payroll, not the wage-demand circuit.

A friction refutation would show that some specified friction is both permanent and sufficient to restore mass productive necessity, rather than merely changing the timing of its erosion. It would need a friction that does not decay, is not routed around, and does not simply spread deployment across time.

The strongest positive refutation: durable mass complementarity

The strongest refutation of the thesis as a whole would be evidence of durable mass complementarity. That means AI adoption raising the marginal value, wages, bargaining power and career mobility of ordinary workers across the affected cognitive labour market, rather than concentrating gains among capital owners, senior verifiers, platform firms or protected incumbents.

If AI deployment produced broad wage growth, more entry-level hiring, restored training ladders and a rising labour share in exposed sectors, the thesis would be wrong. The mechanism would have failed to operate as described, and the wage-demand circuit would be visibly reproducing itself under the new technological condition.

The refutation must show mass complementarity, not elite complementarity. Evidence that some workers use AI well does not refute the thesis. Evidence that the median cognitive worker in an exposed sector is gaining bargaining power, pay or mobility does.

This is the route most useful to empirical observers, because it is still open. The labour market will answer it over the next several years, and the evidence currently points in two directions.

The most direct test is Stanford’s Digital Economy Lab analysis of early-career employment. In an August 2026 revision on a larger dataset, it puts employment of workers aged 22 to 25 in AI-exposed occupations 19 percent below where it would be had it kept pace with less-exposed peers, and finds the gap widening since it was first reported in August 2025. That trajectory runs with the thesis, not against it.

The strongest counter-evidence comes from postings rather than payroll. In May 2026, researchers at the Federal Reserve Bank of New York found no divergence between junior and senior postings within highly exposed occupations, found that the relative decline in exposed vacancies began before ChatGPT, and put fewer than 10 percent of workers and vacancies in highly exposed occupations. The two instruments measure different things. Payroll records who was hired. Postings record whom firms say they want. PwC’s 2026 Global AI Jobs Barometer finds AI-exposed junior roles seven times more likely than the least-exposed to ask for traditionally senior skills, which would let postings stay flat while real entry-level hiring falls. That reconciliation is a hypothesis and is not yet tested. The disagreement is a live test of Premise Two, and the evidence register files the New York Fed finding as disconfirming until it is resolved.

A reversal of the Stanford trajectory across independent measures would be the empirical signal of mass complementarity.

What does not count as refutation

So that this appendix is not read selectively, the following moves do not by themselves refute the thesis. Each is addressed in the essays, and each offers comfort without engaging the mechanism. Appendix II treats them as patterns.

Pointing to residual cognitive work AI cannot do. The thesis grants it. Residual scarcity does not preserve mass absorption. The threshold is majority economic agency, not the last worker.

Pointing to premium human niches, such as high-end law, bespoke medicine, craft production or luxury services. They exist. They absorb thousands or tens of thousands per category. Mass absorption needs hundreds of millions worldwide, and the arithmetic does not close.

Citing slow deployment. Friction modulates timing. Slow destruction is still destruction, and the successor question arrives whenever the propagation completes.

Citing welfare states or redistribution. Postwar capitalism contained redistribution, but welfare states supplemented a wage market that remained primary. A post-AI redistribution regime replaces the wage market as the source of mass demand. The difference between supplement and replacement is the discontinuity.

Citing the political possibility of new policy. The thesis does not predict that no policy response will come. It claims that policy responses preserve consumption rather than the circuit. That is a structural claim about what redistribution and ownership do, not a forecast of what is politically achievable.

Citing imperfections in particular benchmarks. Every benchmark has limits, and no single one is decisive. The thesis rests on the trajectory across several benchmarks, including GDPval, OSWorld-Verified and Tau2-bench Telecom, together with the labour-market evidence. Refuting one benchmark anchor does not refute the thesis. Refuting the broader trajectory across independent measures of deliverable quality, interface operation, deployment cost and labour-market absorption would.

Pointing to the messiness of history or the unpredictability of the future. The thesis is structural. It claims a mechanism operates under specified conditions. That does not depend on the future being smooth. It depends on the conditions continuing to hold.

Recommending softer framing. That may be prudent, but it is not refutation. A refutation must show which premise fails or which preservation route succeeds. How the argument is phrased is a separate question from whether its structure is valid.

Evidence counts when it attacks a load-bearing premise. Isolated examples that do not engage the mechanism do not.

How engagement should be structured

A serious engagement names the premise or premises it challenges, proposes the mechanism it puts in their place, and addresses the essay that covers the relevant ground rather than a summary.

Refutations of Premise One should engage Essay 1. Refutations of Premise Two should engage Essay 2. Refutations of Premise Three should engage Essay 3. Refutations of Premise Four should engage Essays 4, 5, 6 and 7 together, because the regulatory, structural and friction routes interact, and a successful refutation has to close all three.

Engagements that improve particular claims, find failures of scope, correct empirical points or sharpen the formulation are valuable even when they are not refutations. The essays have been improved by such engagement over several rounds, and remain open to more.

The current state of the evidence

As of August 2026, the author’s assessment is as follows.

Premise One is substantially established by vendor-published evidence. OpenAI marketed GPT-5.2 as the first model to perform at or above human expert level on benchmarked professional knowledge work. It reported GPT-5.5 at 84.9 percent wins or ties on GDPval, 78.7 percent on OSWorld-Verified against a 72.4 percent human baseline, and 98.0 percent on Tau2-bench Telecom. In July 2026 it reported GPT-5.6 Sol at 92.2 percent on BrowseComp. Deployed verifier-cost arithmetic appears in case studies including Novo Nordisk and OpenAI’s internal use of Codex. All of these are vendor figures.

Premise Two is substantially established by deployment evidence. Workflow execution is operating at scale across regulated and high-stakes sectors, with named chief executives of deploying organisations, and senior figures at the model providers, stating the structural claim in their own words. Its labour-market leg is the contested part, as set out above.

Premises Three and Four rest on the structural arguments of the Multiplayer Prisoner’s Dilemma, the Successor System and Drag Is Not Rescue, and stand or fall with them.

The space for refutation is therefore narrower than the structure of this appendix suggests. It now requires either a sustained reversal in the capability trajectory, a reversal in the deployment trajectory at scale, or the labour-market evidence resolving towards mass complementarity. None of these is observed as of August 2026. The evidence register is where they would first appear.

What this appendix is not

This appendix is a published refutation standard. The Prize applies it, but the appendix is not a contract. Its purpose is to make explicit what the thesis would have to lose to be wrong. An argument that can name its own failure conditions is a position that can be debated.

The thesis is not the last word on AI and labour. It is a structural claim about why postwar wage capitalism no longer reproduces itself, and an attempt to move the policy debate onto what comes next. The refutation that would change the position would show, with argument and evidence, that mass productive necessity can be restored under the conditions that actually hold.

That refutation has not been produced. The evidence has moved the other way. The thesis has shifted from forecasting a discontinuity to documenting one in progress.