Goodhart's Law Hits the Professions: "AI is breaking our proxies for expertise"
Sean Goedecke's essay starts from the mathematicians' declaration A Severe Misalignment of AI in Mathematics — "Almost five thousand mathematicians (including twenty-five Fields medalists)" signed it — and argues their complaint is not the usual automation grumble but a precise Goodhart's Law problem. The declaration's core line: "the mass production at faster and faster pace of 'true/false' statements could destroy fertile ground instead of breathing life into new ideas." Goedecke's frame: math has two modes, legible "puzzle-solving" (prestigious, prize-able) and illegible "idea-generating" (the actual work). AI now solves the puzzles "the hard way, without generating intuitive new ideas," so the puzzle stops being evidence of anything — "AI companies claim the prestige while not meaningfully advancing mathematical progress."
His pivot to software is what made the HN thread (55 pts) resonate: "It used to be that you could put a meaty project on your GitHub — say, an emulator, or a toy OS — and people would know you were a skilled engineer. But now projects like that are worthless, because everyone just assumes they're vibe-coded." The disagreement in the comments was over whether the proxy is broken or merely reweighted. Optimist case: "AI is the wide receiver and the human is the quarterback" (vatsachak) — immediately countered with the college-football rebuttal that when the receiver is too elite, "QBs who have these clearly elite WRs are discounted and considered more-or-less replaceable all the time" (BoxFour). The artisan counter-signal: "I am currently working on a game. If and when I ever finish it, I want to be able to say that I wrote every single line by hand" (hirvi74) — which got the demographic verdict: "In a few years we'll look at this the same way as people who unicycle or blacksmith" (CuriouslyC). The deepest cut came from exprez135 citing Berg & Baskin's "On Radical Preservation": unchecked LLM use doesn't just cheat the merit system, it risks making shared concepts like merit "not just decadent or debatable but unintelligible."
Why it matters: this is the generalization of the vibe-coding backlash beyond code — the verification tax now applies to credentials themselves (GitHub portfolios, proofs, published work). Every field whose legible outputs AI can counterfeit has to either segregate "human work" (chess-style leagues) or invent new legible signals. That's a multi-year reorganization of how expertise is priced. This extends What Happened to HackerOne? AI Slop and the Crisis of Bug Bounty Automation (AI slop destroying bug-bounty signal) and The 'Vibe-Coding' Hangover: AI Codebases, 'Meat Proxies,' and the Escalating Verification Tax to the prestige economy itself.