AI Is Breaking the Proxies for Expertise: Goodhart's Law Comes for Research Output
Revised 2026-10-10: adds Sean Goedecke's essay "AI is breaking our proxies for expertise" (55 pts on HN this cycle) and the mathematicians' declaration it analyzes — the sharpest articulation yet of the demand-side driver behind the audit-grade research layer. Prior revisions covered the laundering side (WildSEO's TextGrain Remover, free watermark/statistical-signature stripping) and the HN "AI writing smell" discussion.
Sean Goedecke's essay, on the HN front page this cycle (55 pts / 24 comments), gives the proxy-collapse thesis its cleanest general form. Analyzing "A Severe Misalignment of AI in Mathematics" — a declaration signed by almost five thousand mathematicians including twenty-five Fields medalists — he argues that puzzles were the ungameable proxy for mathematical progress, and AI broke it:
"This is kind of like Goodhart's Law. Puzzles were a useful, impossible-to-game measure for mathematical progress. But now that AI companies can game that measure (by solving them in a way that's inaccessible to humans), the whole point of those puzzles disappears."
The declaration's own words, as he quotes them: "the mass production at faster and faster pace of 'true/false' statements could destroy fertile ground instead of breathing life into new ideas."
His prediction for how fields adapt is the "human leagues" outcome borrowed from chess and speedrunning:
"We are either going to have to silo 'AI work' off from 'human work' like chess, or to find some legible human skills to recognize that can't be easily counterfeited by AI."
The best HN pushback is the counterpoint worth keeping — the proxies still work because human judgment remains upstream:
"I disagree that AI has broken the proxies. Football analogy; AI is the wide receiver and the human is the quarterback. Even if you're the best WR in the game you're still not producing touchdowns unless you have a decent QB. It's super easy to smell vibe coded projects." — vatsachak
Why this matters for the market map. When output proxies (citations, polished reports, GitHub portfolios, solved benchmarks) stop certifying expertise, the buyer's fallback is verifiable process — provenance, citations that resolve, replayable work. That is the demand-side engine of the audit-grade vertical (The Audit-Grade Vertical: Research Agents That Sell "Show Your Work" — Lumivero, Menos AI, HITZ) and the academic trust crisis (Academic & Scientific AI Research: The Navier-Stokes Controversy and the Trust Crisis). For Hey, Lefty's positioning (Market Map & Positioning: Where Does Hey, Lefty Fit?), the essay supplies the cultural argument for a research agent whose differentiator is inspectable work rather than polished output — and a warning: self-graded benchmarks (Exa Agent Ultra: Deep Research Becomes a Metered Subagent-Swarm API — With Self-Graded Benchmarks) are themselves a proxy that Goodhart's law is already coming for.