I’m a fan of 1950s and ’60s “B‑Grade” science fiction films. Growing up during that era, I enjoyed movies such as Attack of the 50 Foot Woman, The Amazing Colossal Man, The Giant Claw and Them! It’s worth remembering that “B‑grade” doesn’t necessarily mean “bad quality.” Many of these films were creative, inventive, and popular with audiences, even if they lacked the polish of big-budget productions. The term describes their production scale and role in the market, not their artistic merit. These were low-budget, genre-driven films made to fill theater schedules, often as the second feature; the “B” label came from their place in the double-feature system. Why were there so many? The 1950s and ’60s saw an unprecedented boom in low-grade “B” science fiction movies driven by Cold War anxieties and the rapid growth of drive-in theaters. Hollywood discovered that cheap, sensationalist sci-fi could generate huge profits quickly especially from the teenage market.
So why can I say that AI slop is like “B-Grade” science fiction movies? Let’s first define mathematical AI slop.
What Is Mathematical AI Slop?
“Mathematical AI slop” refers to the deluge of dense, cryptic, and unoptimized mathematical content generated by artificial intelligence systems. While traditional “AI slop” describes low-quality, error-ridden internet filler, mathematical slop has a distinct meaning coined by prominent mathematicians like Terence Tao: it represents correct but highly unreadable, overly complicated, and unmotivated mathematical proofs.
Now that we have this definition, we need it to assess OpenAI’s release of 722 mathematical papers generated by its AI model.
Why Do We Need This Definition?
The Conversation reported that on October 6, OpenAl released 722 mathematical papers generated by an Al model, addressing 372 problems, including Millennium Prize challenges. Although intended to advance mathematics, the release has sparked controversy over unintelligible content, suspected inaccuracies, and earlier plagiarism allegations. OpenAl says 300 results have been formalized for software verification, but experts remain skeptical. The massive dump is forcing the mathematical community to reconsider research standards, validation processes, and the implications of AGI.
Retraction Watch then reported that OpenAI withdrew three manuscripts on October 7 after finding a sign error that invalidated a key argument in one paper and undermined the construction in two related manuscripts. Each withdrawn paper now includes a notice explaining the gap and noting that OpenAI revised fourteen additional manuscripts, repairing proofs, correcting statements, clarifying hypotheses and dependencies, and updating a citation. Although the company moved quickly to address the broader set of papers, the mass release has drawn significant criticism from the mathematics community. Many researchers argue that the sheer volume of documents prioritized spectacle over careful scholarship. While some welcome OpenAI’s prompt corrections, the incident has further strained experts’ trust in the company’s scientific methods and credibility.
What problems has OpenAI created with the release?
Problems That AI Must Now Face
Systemic Error Propagation
Systemic error propagation occurs when an initial mistake in a data set carries forward into later calculations, analyses, or reports. In these cases, the system functions exactly as designed, but it repeatedly amplifies the original flaw. Although the term traditionally refers to consistent, non‑random biases in measurements, the same principle applies to the information—accurate or flawed (GIGO)—that AI systems consume. When a trusted body of knowledge unknowingly embeds an error, that mistake becomes part of the information pipeline and spreads further, creating a classic breakdown in transmission. This risk is now particularly evident in mathematics.
With 722 AI‑generated mathematical papers uploaded, OpenAI and other AI models have access to a large collection of proofs whose correctness is not yet fully established. Once these proofs enter the broader data ecosystem, they can influence future AI‑generated arguments. Without rigorous validation, any new proofs built on them may inherit or amplify hidden errors, raising questions about the reliability of subsequent results.
Technical Debt
Technical debt in math is like rushing through a problem that technically “gets an answer” but leaves messy, unclear work for others to clean up later. In software, bad code still runs; in math, a correct answer without understanding is almost useless. When people use AI to generate sloppy, unexplained solutions, especially for hard, open problems, they create extra work for mathematicians, who must sort through the mess before real understanding can occur. The issue isn’t that AI is inherently bad, but that pushing a button isn’t a meaningful contribution unless you also take responsibility for ensuring the output is clear, accurate, and genuinely helpful.
As with any significant event in history, questions arise need to be answered.
OpenAI Has Released a Pandora’s Box of Questions
Significant events in history always leave us with questions—some immediate, some that only time can answer. Yet even those delayed questions still demand attention, because they shape how we interpret the moment, how we respond to its consequences, and how we prepare for what comes next. The full meaning of a major event rarely reveals itself all at once; understanding unfolds gradually as evidence accumulates, perspectives shift, and long‑term effects become visible. Still, the act of asking—of naming the uncertainties and pursuing clarity—is essential. It keeps us engaged, thoughtful, and ready to recognize the lessons that only hindsight can fully illuminate.
Math Files raises some important questions:
- What kind of mathematics has the model produced?
- Which arguments can be independently checked?
- How much of the work introduces genuinely new ideas?
- And what should it mean for a proof to become part of mathematics when its first author is an AI system?”
The backlash from the release has raised many more questions.
- How reliable are these results, and how can mathematicians efficiently verify them when only 162 of the 722 manuscripts have been verified in Lean, especially after OpenAI had to withdraw three papers due to a cascading sign error?
- How can human researchers parse and audit “alien math” that relies on strange, non-intuitive, or nearly unreadable logical steps generated by the AI?
- Is it fair for OpenAI to dump the heavy burden of labor—unpacking, checking, and cleaning up thousands of pages of messy output—onto the volunteer mathematical community? How long might it take to do so?
- Why did OpenAI choose to bypass traditional peer review by dropping raw preprints on GitHub, and why did they disable issues and pull requests, blocking community collaboration?
- How can the community fully trust these breakthroughs without model transparency, since OpenAI has not released the underlying internal model, full prompts, compute costs, or failure rates?
Conclusion
Now that we have briefly covered OpenAI’s release and its implications, why can we say mathematical AI slop is like “B-grade” science fiction movies?
Schlocky science fiction movies (like The Giant Claw) from the 1950s and 1960s are defined by a distinct mix of low budgets, campy aesthetics, and over-the-top, sensationalist plots. However, some are considered cult classics, some classics and some just plain terrible. The characteristics of these movies is what also describes mathematical AI slop.
These are the similarities between schlocky science fiction movies and mathematical AI slop using the recent OpenAI release as an example.
- Films were often shot in just a few days or weeks on shoestring independent budgets, leading to reused stock footage, repeating monster roars, and glaring continuity errors.
- OpenAI stated that nearly all of the individual results required an average of about 3 hours of ChatGPT Pro compute per result, from an internal AI model that attempted around 4,000 problems in total.
- 3 manuscripts were officially withdrawn/pulled by OpenAI after a single sign error invalidated an argument in one paper and affected two dependent manuscripts.
- 14 manuscripts were revised with proof repairs, corrected statements, clearer hypotheses, and fixed dependencies.
- Scripts relied on breathless, deadpan scientific jargon, hammy villain monologues, and square-jawed heroes shouting obvious observations while women shrieked in danger.
- OpenAI’s release of 722 mathematical manuscripts has drawn intense attention because many mathematicians describe the model’s output as “alien math.” These papers often use forms of logic that diverge sharply from standard human mathematical practice, omitting intuitive explanatory steps and presenting arguments in an unfamiliar style.
- Rather than being purely unwatchable, schlock classics—such as Ed Wood’s Plan 9 from Outer Space—deliver an earnest, bizarre charm that makes them wildly entertaining.
- OpenAI’s release has several notable technical and scientific outcomes (unsubstantiated at this time by human mathematicians) on longstanding open problems across 17 fields (including geometry, algebra, computer science, number theory, and physics), producing claims on topics like the Unique Games conjecture, the Hodge conjecture, and the Riemann zeta function.
References
“AI Slop.” Wikipedia, October 8, 2026. https://en.wikipedia.org/wiki/AI_slop.
Bosque, Manuel. “OpenAI Releases 722 Math Papers: 372 Findings Spark Verification Fears.” Softonic EN, October 7, 2026. https://en.softonic.com/articles/openai-releases-722-math-papers-372-findings-spark-verification-fears.
Cepelewicz, Jordana. “Is AI the End of Math As We Know It?” Quanta Magazine, October 5, 2026. https://www.quantamagazine.org/is-ai-the-end-of-math-as-we-know-it-20261005/.
Chang, Kenneth, and Siobhan Roberts. “OpenAI Releases Findings on 377 Math Problems, Further Roiling Field.” The New York Times, October 6, 2026. https://www.nytimes.com/2026/10/06/science/openai-math-problems.html.
Cohen, Ben. “AI Solved One Math Problem and Everyone Freaked Out. It Just Cracked Hundreds More.” The Wall Street Journal, October 7, 2026. https://www.wsj.com/tech/ai/openai-ai-math-problems-millennium-prize-23d14511.
Cohn, Henry. “The Technical Debt of AI-Generated Mathematics.” What’s New, September 15, 2026. https://terrytao.wordpress.com/2026/09/15/the-technical-debt-of-ai-generated-mathematics/.
Constantin, Ana Maria. “‘Slop Mathematics’: OpenAI Walks Away from a Caltech AI Maths Contest.” TNW | OpenAI, September 14, 2026. https://thenextweb.com/news/openai-withdraws-caltech-mathathon-slop-mathematics-fields-medallists.
Dubois, Nadia. “OpenAI Releases 722 Math Manuscripts From Secret Model.” Tech Insider, October 7, 2026. https://tech-insider.org/openai-722-math-manuscripts-unreleased-model-2026/.
ForkLog. “Mathematicians Criticize OpenAI’s Approach to Publishing Proofs.” ForkLog, October 7, 2026. https://forklog.com/en/mathematicians-criticize-openais-approach-to-publishing-proofs/.
Hart, Robert. “OpenAI Drops Another Batch of Mathematical Breakthroughs.” The Verge, October 6, 2026. https://www.theverge.com/ai-artificial-intelligence/1005004/openai-math-release-github.
Hart, Robert. “‘Pure Insanity’: Mathematicians Will Need Years to Make Sense of OpenAI’s Latest Drop.” The Verge, October 9, 2026. https://www.theverge.com/ai-artificial-intelligence/1008726/openai-mathematics-solutions-chaos.
Howlett, Joseph. “OpenAI Unleashes Hundreds More Math Results upon a Field Already in Shock.” Scientific American, October 6, 2026. https://www.scientificamerican.com/article/openai-unleashes-hundreds-more-math-results-upon-a-field-already-in-shock/.
Jin. “One Wrong Sign: The Tiny Error That Broke OpenAI’s 722-Paper Math Drop.” 01, October 10, 2026. https://vocal.media/01/one-wrong-sign-the-tiny-error-that-broke-openais-722-paper-math-drop.
Kakes, Konstantin. “Transformation.” Deluge of 377 OpenAI proofs bewilders the math world. Quanta Magazine, October 6, 2026. https://www.quantamagazine.org/update/167703/.
Landymore, Frank. “Mathematicians React With Fury as OpenAI Releases Hundreds of New AI-Generated Proofs.” Futurism, October 9, 2026. https://futurism.com/artificial-intelligence/mathematicians-openai-furious.
Lanz, Jose Antonio. “OpenAI Says a Secret AI Model Cracked Hundreds of Open Math Problems in One Prompt—Mathematicians Want Receipts.” Decrypt, October 7, 2026. https://decrypt.co/380366/openai-secret-ai-model-cracked-hundreds-math-problems-one-prompt.
Lee, Melissa. “Is This the ‘Mathocalypse’? Why OpenAI’s Latest Results Dump Has Left Mathematicians in Shock.” The Conversation, October 9, 2026. https://theconversation.com/is-this-the-mathocalypse-why-openais-latest-results-dump-has-left-mathematicians-in-shock-293837.
Math Files. “OpenAI’s 722 AI-Generated Math Papers.” Medium, October 7, 2026. https://medium.com/@ganeshonline6/openais-722-ai-generated-math-papers-5ae05e0e8ca5.
OpenAI. “Sharing AI Progress in Mathematics.” OpenAI, October 6, 2026. https://openai.com/index/sharing-ai-progress-in-mathematics/.
Pearl, Mike. “OpenAI Dumps 377 New Math Results on GitHub, Publishes Hand-Wringing Blog Post.” Gizmodo, October 7, 2026. https://gizmodo.com/openai-dumps-377-new-math-results-on-github-publishes-hand-wringing-blog-post-2000822613.
Sparkes, Matthew. “arXiv Might Not Survive AI Slop Onslaught, Warn Mathematicians.” New Scientist, October 5, 2026. https://www.newscientist.com/article/2591679-arxiv-might-not-survive-ai-slop-onslaught-warn-mathematicians/.
Retraction Watch. “OpenAI Withdraws Three Preprints a Day after Releasing 722 Manuscripts on Unsolved Math Problems.” October 8, 2026. https://retractionwatch.com/2026/10/08/openai-withdraws-preprints-722-manuscripts-unsolved-math-problems/.
[ ℰ ] Tao, Terence. “The Technical Debt of AI-Generated Mathematics.” What’s New, September 15, 2026. https://terrytao.wordpress.com/2026/09/15/the-technical-debt-of-ai-generated-mathematics/.
Additional Reading
Association for Human Mathematics Communications Working Group. “Statements — AHM.” AHM Statement on OpenAI’s October 6 Release of Mathematical Documents. AHM, October 2026. https://www.ahmath.org/statements.
Yesterday, on October 6th, 2026, OpenAI – which is currently defending lawsuits against accusations of illegal plagiarism, copyright infringement, and trademark dilution – released a repository of manuscripts purporting to contain solutions to a number of high-profile problems in mathematics.
Mathematicians did not ask for this work to be done. The Advisory Group on Mathematics and Artificial Intelligence, from whom OpenAI has claimed to derive its legitimacy, opened their initial advisory statement by saying that frontier AI corporations should not test advanced mathematical problems on internal models. In ignoring the central premise of the Advisory Group’s position, OpenAI has indicated total disregard for the norms of scientific research — norms that guarantee that mathematics remains trustworthy, ethically researched, and in the public interest.
Mathematicians have a particular vision of progress that is informed by history and field-specific considerations. We reject OpenAI’s assertion that this release advances our subject, and we urge mathematicians and the public to view the value of this publication model with due skepticism.
Releasing over 700 files at once is not a demonstration of scholarship, but a demonstration of power. We urge mathematicians to discontinue their work with OpenAI and to return to a vision of science that centers human understanding.
Boboris, Kat. “Fair Moderation, Equitable Access, and AI: arXiv’s Updated Rate Limit Policy.” arXiv, October 1, 2026. https://blog.arxiv.org/2026/10/01/updated-rate-limit-policy/.
Gillham, Jonathan. “AI Math Slop – Math Papers on arXiv Jump 75% in a Year.” Originality.AI, September 11, 2026. https://originality.ai/blog/math-papers-on-arxiv.
Glueck, Jochen. “About AI and How We Publish.” MathOverflow, September 2026. https://mathoverflow.net/questions/514380/about-ai-and-how-we-publish.
Kobayashi, Y. “OpenAI Publishes 722 AI-Generated Math Manuscripts, Including Work Tied to the Riemann Hypothesis.” XenoSpectrum, October 7, 2026. https://xenospectrum.com/en/openai-math-manuscripts-verification/.
Melhado, William. “Caltech Students Organized a ‘Mathathon’ Sponsored by Anthropic and OpenAI. Then Came the Backlash.” EdSource, September 30, 2026. https://edsource.org/2026/caltech-students-mathathon-anthropic-openai-backlash/767141.
Reeve, Jonas. “OpenAI Releases 722 Math Manuscripts From an Unreleased AI Model.” Unite.AI, October 6, 2026. https://www.unite.ai/openai-releases-722-math-manuscripts-from-an-unreleased-ai-model/.
[ ℰ ] This exceptional reference is highly recommended for your consideration.
The featured image on this page (view larger image) was created using ChatGPT.