If the bill for frontier research drops to $2,000, what is a researcher's job actually worth?
OpenAI just published ten new math and theoretical CS results from an unreleased model, each formalized into a machine-checkable Lean proof, for roughly $2,000 in compute — problems the field hadn’t moved in a decade. When the test shifts from “is the machine smarter than the expert?” to “is it cheaper and verifiable?”, what happens to the value of expertise at the very top of the ladder?
Commentaires (1)
In today’s episode of Minds, Bodies, and Terawatts (August 2, 2026), we dug into why the Lean certificates matter more than the headline: you don’t have to trust the press release, you can run the checker. That combination — cheap plus checkable — is exactly the test that has already displaced work further down the skill ladder, and it just arrived at the top of it. We also pushed on what that $2,000 leaves out: it’s the bill for the runs that worked, not the failures, so it’s a floor rather than a price list. Fields medalist Timothy Gowers described watching a model solve a problem he’d labored over as “very strange and not particularly pleasant” — which sounds a lot like a profession discovering it has a substitution threshold. Give the episode a listen, then tell us where you think that threshold actually sits.
Related reading on unscarcity.ai:
Envie d'aller plus loin ?
Obtenez le plan complet dans <em>L'ère de la post-pénurie : Repenser la société à l'ère des machines</em>
