OpenAI's Math Proofs Spark a Mathematician Backlash
OpenAI's AI-generated proof of the unique games conjecture stunned mathematicians and sparked a "human mathematics" backlash movement.
What happened with OpenAI’s math proofs?
OpenAI released a batch of mathematical results, reportedly numbering in the hundreds, that included a proof touching on the unique games conjecture, a long-standing open problem in theoretical computer science tied to how hard certain optimization problems are to approximate. The release was reviewed by an advisory group that included well-known mathematicians such as Timothy Gowers and Edward Witten. What made it notable wasn’t just that an AI produced new mathematical results. It’s that some of the proofs are reportedly so unusual in structure and style that working mathematicians, including specialists in the exact subfield, say they can’t fully verify or understand them without AI assistance.
TL;DR
- OpenAI published a large set of mathematical results, among them a proof connected to the unique games conjecture, a problem some mathematicians have spent entire careers on.
- Scott Aaronson, a theoretical computer scientist and former OpenAI collaborator, called the release one of the biggest days in mathematical history, while noting the proofs read like they were “written by someone on psychedelics.”
- Mathematicians reportedly struggle to verify parts of the work because it uses unfamiliar constructions, like a novel “noise test” built through recursive methods instead of established techniques.
- A new group calling itself the Association for Human Mathematics formed in response, arguing mathematicians “did not ask for this work to be done.”
- The backlash has drawn criticism for seeming territorial, since mathematical proof, unlike a private competition, is traditionally open to anyone who can produce a valid argument.
- A parallel anxiety is surfacing in cryptocurrency, where researchers worry AI-assisted tools could accelerate the discovery of vulnerabilities in wallets, smart contracts, and blockchain infrastructure.
- Vitalik Buterin has publicly acknowledged these AI-related security concerns rather than dismissing them, which carries weight given his stake in the crypto ecosystem.
Remy is new. The platform isn't.
Remy is the latest expression of years of platform work. Not a hastily wrapped LLM.
Why are mathematicians upset about an AI solving proofs?
The core objection isn’t that the math is wrong. It’s that the math is alien. Multiple mathematicians examining the unique games conjecture proof have said the techniques don’t resemble how humans normally build these arguments. Instead of extending known, well-understood methods, the proof reportedly invents new constructs, including an unusual “noise test” with a recursive structure that doesn’t map cleanly onto prior work in the field.
That creates a verification problem. Mathematical proofs are supposed to be checkable step by step by any sufficiently trained reader. If a proof is so densely written and so unfamiliar in its approach that experts need AI tools just to reconstruct what claim is being made in a given section, the normal process of peer review breaks down. Aaronson described getting help from an AI assistant just to extract a coherent “completeness and soundness” claim from one part of the paper, piecing it together from scattered sections.
There’s also a more personal dimension. Aaronson’s wife, a complexity theorist, had worked on the unique games conjecture for years. His anecdote about his son asking whether “a robot solved the math problem Mommy worked on her whole career” captures something real: for mathematicians who’ve devoted decades to a specific open problem, having it resolved overnight by a system trained on humanity’s existing knowledge (and then pushed far beyond it) is disorienting, regardless of whether the result holds up.
What is the Association for Human Mathematics?
In direct response to the OpenAI release, a group of mathematicians formed something called the Association for Human Mathematics. Its opening position, as quoted in discussions of the letter, is that “mathematicians did not ask for this work to be done.” The framing suggests a desire to draw a line around mathematics as a human activity and to push back on AI-generated results being folded into the field without some form of consent or process from the existing mathematical community.
The reaction to this has been mixed and often sharply critical. Mathematics, unlike a private research lab, isn’t something any one group owns or gatekeeps. A valid proof is a valid proof regardless of who or what produced it, and nobody needs permission from an existing association to work on an open problem. Critics have pointed out the apparent contradiction: a field built entirely on open, independently verifiable argument is now seeing some of its practitioners argue for restricting who gets to participate.
That said, the underlying anxieties aren’t baseless. Concerns about job displacement, about losing the human intuition and insight that make mathematical discovery meaningful, and about a flood of unverifiable AI-generated “proofs” overwhelming journals and reviewers are all legitimate, even if the public messaging has come across as defensive or territorial.
Is the backlash just about ego, or are there real risks?
Both. The optics of a “no AI allowed” stance in a discipline defined by universal, checkable truth claims invite ridicule, and that’s largely what’s happened online. But several real risks sit underneath the noise:
Remy doesn't build the plumbing. It inherits it.
Other agents wire up auth, databases, models, and integrations from scratch every time you ask them to build something.
Remy ships with all of it from MindStudio — so every cycle goes into the app you actually want.
Verification bottlenecks are real. If AI systems can generate proofs faster than human experts can check them, the field needs new infrastructure, likely AI-assisted itself, to keep review honest and rigorous. A proof nobody fully understands isn’t useful even if it’s technically correct.
Style and insight matter beyond correctness. Part of the value of mathematical proof historically has been the human insight it generates, the elegant trick or reframing that teaches other mathematicians something transferable. A proof that works but can’t be understood or generalized by humans delivers the result without the intellectual payoff the field has historically valued.
There’s also a credit and incentive question. If decades-long open problems can be cracked by a model trained partly on the public record of prior human attempts, it raises uncomfortable questions about how credit, funding, and career incentives in academic mathematics should work going forward.
What does this have to do with cryptocurrency?
A related anxiety is playing out in crypto circles, and it’s less about proofs and more about security. Blockchain researcher Justin Drake has called for what he termed “bunker mode” in the industry, urging holders to migrate assets to fresh, previously unused addresses whose public keys haven’t been exposed on-chain. The underlying fear is that AI tools are getting better and faster at finding vulnerabilities in cryptographic code and smart contracts, including in systems whose source code has long been public.
This isn’t hypothetical. A major South Korean bank hack was reportedly carried out using an AI tool, and previous incidents have involved attackers exploiting vulnerabilities in open-source code for hardware wallets, the same openness that let the security community audit that code also lets automated tools hunt it for weaknesses at scale.
Vitalik Buterin, one of the most prominent figures in the crypto industry, responded to Drake’s warning not by dismissing it but by engaging with the concern seriously. Given that his financial and professional stake is tied to the health of the ecosystem, his willingness to acknowledge the risk rather than wave it away is notable. It mirrors the mathematics situation: people with the deepest expertise in a field are the ones sounding the alarm, not outsiders unfamiliar with the subject.
Frequently Asked Questions
What is the unique games conjecture?
It’s a conjecture in theoretical computer science about how hard it is to approximate solutions to certain optimization problems. It has significant implications for complexity theory and has been a major open problem that researchers have worked on for years.
Did OpenAI actually prove the unique games conjecture?
OpenAI released a proof connected to the conjecture as part of a larger batch of mathematical results, reviewed by an advisory group of established mathematicians. Multiple mathematicians have said parts of the proof are difficult to verify because of unfamiliar methods, so full community confirmation and consensus is still an ongoing process.
What is the Association for Human Mathematics?
It’s a newly formed group of mathematicians organized in response to AI-generated mathematical proofs, arguing that the mathematical community did not request this kind of AI involvement in the field.
Why do some mathematicians say the AI proofs feel “alien”?
Seven tools to build an app. Or just Remy.
Editor, preview, AI agents, deploy — all in one tab. Nothing to install.
Because the methods used don’t follow established human approaches to the problem. Instead of building on recognizable techniques, the proofs reportedly introduce new constructs, like unfamiliar recursive testing methods, that experts say are hard to trace back to known mathematical reasoning.
Is the concern about AI and crypto security related to the math backlash?
They’re separate issues but share a theme: experts in highly technical fields, mathematics and cryptography, are both raising concerns about AI systems outpacing human ability to verify work or defend systems, whether that’s checking a proof or securing a crypto wallet.
