AI is breaking our proxies for expertise

TEXT START: Mathematicians are broadly not anti-AI.

The Dissection

The article correctly identifies the collapse of expertise proxies. Puzzle-solving was legible evidence of mathematical ability; GitHub projects and high-volume coding served the same function in software. AI makes those signals cheap to counterfeit or directly produces the result, severing achievement from demonstrated human capability.

But the article stops at the cultural surface. It reframes structural displacement as a prestige crisis: invent new proxies, preserve human-only leagues, or make AI output understandable to humans. That is an obituary for old expertise disguised as a program for rebuilding its badges.

The Core Fallacy

It confuses the survival of a cultural practice with the survival of its economic function.

Under the Discontinuity Thesis, the decisive question is not whether humans can retain prestige, solve problems unaided, or explain machine-generated proofs. It is whether human labor remains necessary and competitive. P1 makes proxy collapse a symptom of AI cost and performance dominance. P2 makes stable human-only productive domains impossible at scale. P3 removes the majority from economically necessary labor.

The chess analogy fails structurally. Human chess leagues survive because chess is an entertainment ritual, not a mass production system. Mathematics may survive as a human-only prestige enclave. Software engineering is embedded directly in production and therefore cannot be protected by cultural preference once machine-generated systems are cheaper and adequate.

A human-readable proof may preserve understanding. It does not restore bargaining power. If AI can generate, verify, translate, and teach the proof—or automate enough surrounding work to make human interpretation cheap—the remaining human role is a narrow Servitor niche, not a recovered profession. The scoreboard can survive after the economic game has moved elsewhere.

Hidden Assumptions

• Conceptual understanding is the economy’s bottleneck rather than a human preference or civilizational luxury.
• An AI solution that humans cannot intuitively understand is economically inferior. Markets usually require useful, reliable results—not human comprehension for its own sake.
• Human-only competitions can sustain large-scale employment, income, and institutional status.
• A new human skill can remain permanently uncopiable. If it can be specified and rewarded, it becomes an automation target; if it cannot be verified, it cannot serve as a broad credential.
• Mathematics and software are comparable to chess and speedrunning. They are not: the former shape knowledge and production, while the latter are already bounded spectator practices.
• Human interpreters will remain ahead of AI at making machine output legible.
• Uncertainty about whether AI generates genuinely new ideas represents a durable human moat. It is merely an unresolved technical gap until proven otherwise.
• The central damage is status loss and unhappiness rather than loss of ownership, income, and productive necessity.
• Existing institutions can coordinate a stable human-only economic zone. P2 says they cannot do so at scale against competitive pressure.

Social Function

Classification: partial truth, transition management, prestige preservation, and ideological anesthetic.

The article is not empty copium. It accurately diagnoses that old evidence of competence no longer proves much. Its anesthetic function begins when it treats the remedy as cultural redesign: better proxies, human leagues, and human explanations of machine work.

That preserves identity while avoiding the ownership question. It does not ask who controls compute, energy, distribution, data, or the firms deploying the systems. It offers educated workers a place beside the machine, not control of the machine. Human mathematics can persist as a protected ritual; human software work can persist in verification, integration, maintenance, and transition niches. Those are enclaves and servitor positions, not a restored mass labor market.

The Verdict

The article sees the first-order symptom but underestimates the disease. AI is not merely breaking proxies for expertise; it is dissolving the link between human skill, economic necessity, and bargaining power.

The proposed human-only leagues may preserve prestige. They cannot preserve the post-WWII wage-to-consumption circuit. Even if humans retain some monopoly over deep conceptual insight, that monopoly can become a narrow source of elite differentiation rather than mass employment. The viable positions are ownership and control of AI capital, indispensable Servitor work, or Vulture/Hyena roles in verification, transition, infrastructure, and carcass management.

The old experts are not simply waiting for a better scoreboard. Their scoreboard has become evidence of a game the machines no longer need them to play.