The Mathocalypse III
Research report · 17 October 2026
On 6 October 2026 OpenAI posted 722 AI-generated manuscripts to GitHub, claiming progress on hundreds of open mathematical problems. For many mathematicians an abstract question about AI’s future became, overnight, a professional experience: problems they had worked on for years appeared on a list of claimed solutions. Their responses range from grief to awe, and they rarely sort neatly into optimists and pessimists. The sharpest testimony comes from doctoral students and young researchers, who stand to lose thesis topics and job prospects. Few dismissed the mathematics outright. The recurring complaints are about how it arrived: all at once, without human authors who can answer questions, often hard to read, and on a company’s timetable rather than the profession’s.
Scott Aaronson’s blog post of 7 October gave the episode its name: “The Mathocalypse”.
AI-generated manuscripts posted to GitHub, grouped into 372 families of results (377 claimed proofs by Quanta's count).
Retraction Watch300 of 719 top-line results had Lean proofs on 7 October. A Lean proof checks the statement as encoded, not the paper's exposition.
ImplicatorOpen problems put to OpenAI's internal model. Three papers were withdrawn the next day over a sign error.
OpenAIWhat OpenAI released, and the month before
| Date | Tuesday 6 October 2026, around 6 p.m. US Eastern time (OpenAI) |
| Manuscripts | 722, produced by an unreleased internal model and posted to GitHub (Retraction Watch) |
| Results | Counted differently by different sources: 372 “families” of related results in OpenAI’s catalogue (Implicator), 377 claimed proofs in Quanta’s count (Quanta) |
| Problems attempted | About 4,000, at an average of roughly three hours of ChatGPT Pro-equivalent computing per result (OpenAI; Implicator) |
| Machine-checked | 300 of 719 top-line results (about 42%) had Lean proofs as of 7 October. A Lean proof checks the statement as encoded, which may not match the theorem the paper advertises (Implicator; Quanta) |
| Corrections | Three algebraic-geometry papers withdrawn on 7 October over a sign error; 14 more revised (Retraction Watch) |
| Advice and money | OpenAI said it drew on guidelines from the Advisory Group on Mathematics and Artificial Intelligence (AGMAI) at the Institute for Advanced Study. On 15 October it added a $5 million contribution to a new Simons Foundation initiative (OpenAI) |
The release fell short of its most famous targets. On the Riemann Hypothesis it reached only a weaker variant, which Álvaro Lozano-Robledo called “quasi-RH” (The Decoder).
Background. Tension had been building for a month. OpenAI’s disputed Navier–Stokes claim in early September prompted a joint letter from 25 Fields medallists on 11 September. They warned that “the goals of the AI companies and the goals of the mathematical community are severely misaligned”. The same week saw the launch of the Association for Human Mathematics (AHM), whose members pledge not to work with commercial AI companies. AGMAI published guidelines for responsible release on 29 September.
Voices
Reactions rarely fit one box, so each mathematician can carry several themes. Filter by theme, then select a name to read the full account and its sources.
1 Hugo Duminil-Copin Fields Medal 2022 · University of Geneva The day his problems came back solved
Duminil-Copin had expected AI to overtake humans in parts of mathematics one day. What he had not expected was the concentration: the release appeared to touch every major open problem he had posed publicly, all on one day. “Everything has been claimed to be proved,” he wrote in his signed reaction, adding: “I am paralysed.” On Swiss radio on 8 October he called it “un choc monumental”, a monumental shock, and said mathematicians must reinvent how they do their work (RTS, via the University of Geneva).
In a 14 October Numberphile interview he said that 16 of the proofs had left him in tears. He explained that he had said so publicly out of empathy for distressed colleagues: “you need to share your emotion to pass through it and to bounce better after it.” He rejected the reading that mathematics is finished, saying “No mathematician will ever say that you can solve mathematics.” His expectation is that machines will increasingly lead on raw problem-solving, and that the human task is to catch up and turn the proofs into understanding: “we have to bounce on these proofs and to try to make the best of it.”
2 Terence Tao Fields Medal 2006 · UCLA An abundance the profession cannot yet digest
Tao gives one of the fullest mixed accounts. His signed response recognises that the manuscripts “appear to introduce clever new ideas”. He is troubled, though, that no human behind them can explain a step, give a talk or train a student, and that research paths students expected to spend years on have vanished. “I mourn the path not taken,” he wrote. He regretted that labs had not submitted their frontier models to independent researchers for proper scientific evaluation, and urged that “the community needs to come together more than ever” to declare its own standards. His sharpest image, quoted by Quanta, is agricultural: open problems that once sustained years of human inquiry are being solved wholesale, so that “solutions to open problems are now being harvested at large scale in an unsustainable fashion”, leaving entire fields “much less fertile” than when such problems were solved in the traditional way (Quanta).
In a 15 October Numberphile interview, titled “No More Heroes”, he put it plainly: “we are transitioning from an era of proof scarcity to proof abundance”. The result is that “we’re drowning in proofs, but we’re not getting all the understanding that we used to get.” He expects mathematics to “work in big teams, more like the other sciences”. He wants students still to learn problem-solving by hand, as they learn arithmetic before calculators. “Something will be lost,” he said, but “exciting new things are possible.” His 9 October Caltech lecture, “Math 2.0”, criticised “the excessive emphasis on automated solving of open problems by AI”. His blog has also hosted guest essays, including the AHM statement and Huiberts’s critique, which speak for their authors rather than for him.
3 Peter Scholze Fields Medal 2018 · Max Planck Institute for Mathematics, Bonn A marathon, with consequences beyond mathematics
Scholze asked for patience in his signed reaction: “mathematics is a marathon, not a sprint,” and “the goal is and always will be the human understanding of mathematics.” His most pointed concern reaches beyond the profession. If AI begins finding algorithms that break widely used cryptography, the consequences for society could be immediate and serious. Quanta lists him among the AHM’s members (Quanta), though the association’s statements speak for the group rather than for him.
4 Levent Alpöge Number theorist at Anthropic Exhilaration beside a conflict of interest
Alpöge reacted on X on the night of 6 October with unusually strong praise for the claimed results on the quasi-Riemann hypothesis and Siegel zeros. He was “kicking myself” for having said such results were within reach without pursuing them. He also mentioned “sad stories” of researchers “getting scooped” and of conflicts of interest, without naming anyone. He still called it “obviously the most significant moment in mathematical history.” He gave no names or details of the scooping cases. As an employee of a competing AI lab, he is not a neutral witness.
5 Ben Green University of Oxford, Waynflete Professor of Pure Mathematics Shock, then work to do
“I was not expecting the magnitude of some of these results,” Green wrote in his signed reaction. He was most struck by a claimed “quasi-Riemann hypothesis”: a proof that Dirichlet L-functions, central objects in number theory, have no zeros across a wide region where the Riemann Hypothesis predicts none. He called it “absolutely shocking to me”. “Many of the central problems of additive combinatorics have fallen,” he wrote, among them “around three quarters of the aims I had for an ERC Advanced Grant, awarded only in June.” Yet his conclusion looks forward: “The work of understanding these solutions and the associated context properly is significant.”
6 Henry Wilton University of Cambridge Proofs without an author to ask
Wilton found several problems he cares about among the claimed solutions, and gave the bleakest line of the week: “If OpenAI wanted to destroy the mathematical community, this would be a great way to go about it.” He called the release “a tsunami that we have no time to prepare for” (signed reaction). His concern is structural. Without human authors to answer questions, revise an exposition or submit to journals, the release bypasses the seminars and peer review through which mathematicians judge, repair and connect results.
7 Bryna Kra Northwestern University; AMS president 2023–25 The collaboration beneath a theorem
Kra accepts that the release may contain deep results, but asks what happens when a reader gets stuck and there is no author to ask. Research grows from tentative ideas shared in confidence. If those ideas are immediately turned into machine proofs, the incentive to share them weakens. “This way of releasing results closes off directions of research, rather than opening new vistas,” she wrote (signed reaction). She expects doctoral training, hiring, promotion and publication to need redesigning.
8 Leslie Ann Goldberg University of Oxford, combinatorics, probability and theoretical computer science Pausing an essay as events overtook it
Goldberg added a 7 October postscript to an essay she had written before the release. She expects the results “will be true”, and says the Lean translations “can and will be checked.” The real issue lies elsewhere: “AI is not a threat to mathematics,” but it “is a threat to the human infrastructure that we have built for understanding mathematics.”
9 Álvaro Lozano-Robledo University of Connecticut Telling students to stay
Lozano-Robledo called 6 October possibly “the single most important day in the history of mathematics thus far”. He noted, though, that the model fell short of its headline goals: the Riemann Hypothesis, the full Hodge conjecture and Birch–Swinnerton-Dyer (signed reaction). His guest essay on Tao’s blog, written for a worried student, was composed before the release; its afterword says the release does not change his advice. A student who loves mathematics should still consider a PhD, since there is much work in checking, explaining and extending the results. He acknowledges that tenure gives him a security young mathematicians lack.
10 Kevin Buzzard Imperial College London; leader of the Lean formalisation project Xena Messy papers, real promise
Buzzard, a long-time AI optimist, admits the mess: “We do not want to read slop papers.” Still, he believes “truth and understanding will bubble to the top” as people work through the collection and AI gets better at explaining itself (signed reaction). Asked on his blog for his view of the release, he answered in Fermat’s words: “Too large to fit into this margin.”
11 Scott Aaronson University of Texas at Austin, theoretical computer science Historic change, responsibility on both sides
Aaronson coined “the Mathocalypse” and treats the release as historic. He also warned that “no human has understood just about any of these proofs yet; the race to do so has just started.” He holds that mathematicians should press AI companies on conduct and safety, but that refusing to let AI attack open problems is “totally untenable”. He contrasts OpenAI’s raw dump with Anthropic’s handling of its own recent results, where human researchers were paid to write up the key idea, and sees drawbacks in both. His wife, the complexity theorist Dana Moshkovitz (also at UT Austin), was vindicated by the claimed proof of the Unique Games Conjecture, which she had long believed true, but still found the paper “so horribly written that it’s impossible to read it without AI help.” She added that the results are paradigm shifts but “they don’t finish math”.
12 Martin Bridson University of Oxford; President of the Clay Mathematics Institute Who chooses the next questions?
Bridson calls the scope “truly breathtaking” and sees a chance for larger human–AI projects than were ever possible (signed reaction). He worries that mathematicians could become interpreters of questions chosen by AI labs, and distrusts the labs’ commercial motives: “we should not be indentured to follow their lead.”
13 Sophie Huiberts Mathematician; PhD Utrecht; Stieltjes Prize A theorem is also a piece of literature
In “This is normal bad work” (12 October, a guest essay on Tao’s blog), Huiberts argues that proving a theorem brings duties to the field. An author must explain what is new, credit predecessors and improve the literature. Judged that way, “these pdfs are all easy rejects” that “actively set back the state of the literature.” She rejects the idea that 6 October was a watershed: the work is normal progress, and the only exceptional thing is OpenAI’s “lack of shame”.
14 Asaf Karagila University of Leeds, set theory A specialist reads one paper closely
Karagila reviewed the manuscript in his own speciality (8 October) and found it “unclear, muddled”, with a strange structure, and in his view fit for desk rejection. He noted that OpenAI’s “progress” became “solutions” in media coverage. His worry is that a powerful company is asking the profession to accept its standards rather than meet the profession’s own. This is a judgement on one paper, which is exactly why specialist readings matter more than blanket verdicts either way.
15 Emily Riehl Johns Hopkins University Remember the human work inside the papers
Riehl stresses the human lineage of the results: “mathematicians formulated the conjectures, proved the surrounding results, and shared their ideas” (signed reaction). She welcomed a site, citedbyagi.com, that traces which human papers each AI result builds on. She wants a future that keeps credit, lineage and learning between people.
16 Danny Calegari and Josh Frisch University of Chicago; UC San Diego Other ways to value the work
Calegari says choosing problems purely because they fascinate him has left him largely untouched: “I honestly think there will be more ways to do math in the future, not fewer.” Frisch describes 2026 as a year of loss, including “the era of human proofs”. He still sees beauty in the new results, and hopes mathematicians will work to understand and share them (signed reactions).
17 Jeremy Avigad Carnegie Mellon University; director of the Institute for Computer-Aided Reasoning in Mathematics Keep setting ambitious problems
Avigad’s 5 October essay urged mathematicians to take on harder questions and support younger researchers as AI capability rises; it preceded the release. Afterwards he told Quanta: “It’s clear that the OpenAI release includes major advances” (Quanta).
18 Ian Agol University of California, Berkeley Comparing it to Perelman
Agol was stunned by results in geometric group theory and topology, and compared the release to Perelman’s proof of the geometrisation conjecture. He hopes the results will be formalised, a task that would also mean formalising some of his own work (signed reaction).
Starting out: students and early-career researchers
The people with the most at stake are those whose careers have barely begun. A professor whose favourite problem is claimed loses a project. A doctoral student can lose a thesis topic, a job-market pitch or a reason to stay. Their responses are less guarded than their seniors’, and more varied: grief and anger, but also excitement, and some practical thinking about what a PhD should now be for.
The mood predates the release. In an August survey of MIT’s mathematics graduate students, about three-quarters of the 45 who responded (out of 120) feared AI would make tenure-track jobs harder to get, and 71% were looking at moving away from academic mathematics (Boston Globe).
19 Tasmin Chu Caltech mathematics doctoral student; co-founder of the AHM The problem she had loved since she was an undergraduate
The release claimed a problem Chu had loved since her undergraduate years, building on the work of her own adviser. What she had wanted was “the time and space to think about it for the next few years”. Her grant proposal, due within a week, “has to be rewritten” (signed reaction). “I feel a profound sense of grief for the working conditions I believed I would have.” Her concern is less about who owns a theorem than whether the profession will keep giving young mathematicians room to become mathematicians.
20 Tristan Humbert Doctoral researcher applying for postdocs The thesis problem that disappeared
Humbert’s PhD was built around Katok’s entropy conjecture, which appeared among the claimed solutions. He describes sadness, then stress, because “I am currently applying for postdocs and more of my research plan was now obsolete”, and finally anger. He called the write-ups “mostly unreadable slop”, adding: “Open AI’s standards for mathematical publication is far too low and is harming mathematical research” (signed reaction).
21 Dora Woodruff and Hanna Mularczyk Mathematics doctoral students, MIT “the ones who are actually going to be on the job market”
Woodruff, a third-year doctoral student, told the Boston Globe that the stakes are highest for students, who are “the ones who are actually going to be on the job market in a couple years”. She is now questioning whether she wants to spend her life in the field. Mularczyk, in her fifth year, helped run the department survey and organise a town hall. She worries that a problem she has been working on may be among OpenAI’s proofs. She is also uneasy about how students learn: she has found herself having to “defend myself against Chat about a calculus problem”, and says learning mathematics should not mean having “an oracle built into your computer to tell you what to do” (Boston Globe, 15 October).
22 Junseo Lee Doctoral student, quantum information Not sure what the path is any more
“I started my PhD hoping to work on quantum complexity and learning theory,” Lee wrote. Some of the claimed results touch questions Lee’s advisers had worked on. The reaction is excitement, and unease: “I’m no longer sure what the best path forward looks like for someone just starting a PhD.” Lee worries that fewer students will choose theoretical research at all (signed reaction).
23 Michael Chapman Young researcher More excited than afraid
Chapman had first met several of the claimed problems as a student. “This is quite humbling, and as a young researcher also disorienting,” he wrote. “Nevertheless, I am more excited than afraid.” Chapman hopes the research community comes out of it stronger (signed reaction).
24 Claudio Llosa Isenrich Early-career mathematician The lighthouses torn down
Writing from what Llosa Isenrich calls “my (I like to believe still short) career”, the reaction is grief for “all the guiding lighthouses that have been torn down”: the famous problems that gave young researchers a direction. Distrustful of the AI companies but hopeful about the community’s response, Llosa Isenrich worries most for those coming next, the bachelor’s, master’s, PhD and postdoctoral students: “I am longing for stability for myself, but even more for our future generation” (signed reaction).
25 Written as the storm gathered Three essays by early-career mathematicians, written in August and September and published around the release
Sirawit Pongnakintr, a master’s student at ETH Zurich applying for PhD places, argues that a mathematics PhD should no longer require an original contribution. Asking students for one is asking them “to race against a superintelligent machine”. A PhD should instead certify understanding and mastery, with protection for students whose research is overtaken. He may not apply to programmes that judge him by output (Proofs and Prompts, 6 October).
Lawrence Hollom, a postdoc at EPFL, asks what he will achieve in the next three years. He chose mathematics because “I loved the process, I loved solving problems”. He now writes with wry resignation: “I still have a purpose. Until the next version comes out, at least.” In a later comment he says his view has shifted. He now enjoys writing exposition, and thinks AI remains weak at judging what matters (Proofs and Prompts, 10 October).
Tattwamasi Amrutam, an adjunct assistant professor at the Institute of Mathematics of the Polish Academy of Sciences, saw two long-standing questions in Amrutam’s own area answered by a machine weeks after finishing a major project. “There will be proofs everywhere and an abundance of them. Only the understanding will dwindle.” Retreating from the field would be a mistake: “There are no great or small actions. There are only actions.” (Proofs and Prompts, 7 October)
Should you still do a PhD?
Yes, if you love mathematics. There is much work in checking, explaining and extending the results.
Yes: a PhD still builds deep thinking that is valuable inside and outside academia. “Just don’t use it to bypass the difficult, friction-filled stuff!”
The work shifts towards interpreting machine results: “somebody needs to interpret this and communicate this.”
Only if the PhD changes, certifying mastery rather than demanding original results.
“HECK NO! Are you crazy?” Academic jobs are scarce, and established professors urging passion are not the ones taking the risk.
More voices
“Today, something is lost forever.”
“How much beauty have we lost?”
“No human being should have to respond this quickly.”
“The present breakneck speed of things without human understanding compromises precisely this.”
On an offer of paid authorship under a non-disclosure agreement: “This is unthinkable in academic research.”
Posted a paper of his own the day after the release: “living in a mathematical wonderland”
“There’s no hurry and it will all get done.”
Treat the papers as “something like ‘research findings’ or ‘raw data’” for now
OpenAI “should have announced the manuscripts that were lean verified first”
The withdrawals were “the responsible thing to do”, but many are “unhappy with how OpenAI has behaved”
Unless another source is given, these quotes are from Proofs and Prompts.
Organised responses
Fields medallists' declaration
11 Sep 2026 (background)
AI companies’ goals and mathematics’ goals are “severely misaligned”. Famous problems should be a route to understanding, not a benchmark.
25 signatories, including Tao, Scholze, Duminil-Copin, Viazovska, Maynard, Hairer and Deligne
Association for Human Mathematics
7 Oct 2026
Called for a boycott, urging mathematicians to “discontinue their work with OpenAI”. “Mathematicians did not ask for this work to be done”; the release was “a demonstration of power”, not scholarship.
About 950 members, per its member page
AGMAI
6 Oct 2026
“An important event for mathematics”, but “the beginning, not the completion” of human understanding. It did not endorse OpenAI’s process, warned that research “cannot consist only of understanding results produced by AI labs”, and left it to the community to judge whether its guidelines were followed.
François Charles, Camillo De Lellis, Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ulrike Tillmann, Ravi Vakil, Edward Witten, Melanie Matchett Wood
Where they divide
Is this progress?
One view
Major advances (Avigad, Green, Agol, Aaronson, Alpöge)
The other view
Poor scholarship that sets back the literature (Huiberts, Karagila)
What matters most?
One view
True theorems, checked in Lean, will be absorbed in time (Buzzard, Goldberg on the results)
The other view
Understanding is the point, and proofs no one can explain do not deliver it (Tao, Scholze, Kra, Goldberg on the infrastructure)
How fast?
One view
The race to understand has begun (Aaronson)
The other view
Slow down: no one should have to respond this fast (Fisher, Ellenberg, Litt, Scholze)
Who sets the agenda?
One view
Engage the labs and shape their conduct (AGMAI)
The other view
Boycott them (the AHM)
What about careers?
One view
A PhD is still worth it (Lozano-Robledo, Kontorovich and Koberda, Calegari)
The other view
Young researchers’ problems and plans are gone (Chu, Humbert, Woodruff), or the PhD itself must change (Pongnakintr) or be avoided (Polak)
A profession trying to turn proofs into knowledge
These accounts do not divide neatly into optimists and pessimists. Duminil-Copin’s grief sits alongside a commitment to understanding. Tao sees both new ideas and the loss of a familiar way of working. Green moves from shock to the work ahead. Alpöge’s exhilaration comes with concern for researchers who were scooped. Buzzard expects better tools and explanations, while Karagila insists that each manuscript must still earn trust. Scholze reminds the field that some discoveries could matter far beyond academic credit.
The most common concern is the gap between producing proofs and making mathematics usable. A field needs people who can check a statement, explain why it matters, answer questions, connect it to earlier work and teach the next generation. It needs early-career researchers with time to develop ideas, and institutions that reward that work. AI could speed those tasks up or crowd them out; the release alone does not decide which.
Recurring complaints
- Readability. Many papers may be correct but are very hard to read; one expert told Quanta some are not close to publishable form.
- Credit. There are no human authors, and earlier work is weakly acknowledged (Huiberts, Riehl, Wilton).
- Verification. A Lean proof checks the statement as encoded. Johan Commelin of the Mathlib Initiative noted that human experts must still read those formal statements (Quanta).
- Access and control. Public models lag far behind internal ones, so labs decide who works at the frontier (Saha, AGMAI).
- Junior researchers. The heaviest cost falls on PhD students and early-career mathematicians whose problems were claimed (Chu, Humbert, Lee, Llosa Isenrich), and on students deciding whether to start at all (Woodruff, Pongnakintr).
What comes next
The coming months of specialist reading, withdrawal, repair and exposition will decide what the collection is worth as mathematics. The institutions are already moving. The Simons Foundation has launched a $15 million effort to help mathematicians work with AI results, and Tao’s blog has called for stories of human–AI collaboration. The deeper question these mathematicians raise is who will understand, direct and take part in mathematics once producing a candidate proof becomes far faster than absorbing one.
About this report
This is an account of people, not a scorecard of proofs. Its main evidence is signed reflections, interviews and essays by mathematicians with clear professional credentials, with Fields medallists and other prominent figures covered first. Many short reactions come from Proofs and Prompts, which published about 100 signed responses on 8 October. That collection is a record of individual experience, not a survey: people whose own research was touched, or who felt strongly, are over-represented. The section on students and early-career researchers includes doctoral and master’s students and others at the start of their careers, whatever their standing.
Dates matter. Anything written before 6 October is labelled as background or forecast, not as a reaction to the manuscripts.
All sources
| Source | Date |
|---|---|
| 100+ reactions to 100+ solutions (Proofs and Prompts) | 8 Oct 2026, updated 14 Oct |
| Sharing AI progress in mathematics (OpenAI) | 6 Oct 2026, updated 15 Oct |
| What’s the point of MIT? OpenAI’s math proofs have thrown students and researchers into chaos (Boston Globe) | 15 Oct 2026 |
| Opinions against the PhD, from an incoming PhD applicant (Sirawit Pongnakintr) | 6 Oct 2026 |
| The grand piano (Tattwamasi Amrutam) | 7 Oct 2026 |
| What will you achieve in the next three years? (Lawrence Hollom) | 10 Oct 2026 |
| The Dialectical Theorem (Alex Kontorovich and Thomas Koberda) | 14 Oct 2026 |
| AI does math, should you get a PhD? (Jason Polak) | 9 Oct 2026 |
| The Mathocalypse (Scott Aaronson) | 7 Oct 2026 |
| The Strip Mining of Mathematical Proofs (Numberphile, Duminil-Copin); transcript | 14 Oct 2026 |
| No More Heroes (Numberphile, Tao); transcript | 15 Oct 2026 |
| Duminil-Copin on RTS La Première (University of Geneva) | Aired 8 Oct 2026 |
| Math 2.0 (Terence Tao, Caltech lecture) | 10 Oct 2026 |
| Alpöge’s X post (archived copy) | 6 Oct 2026 |
| Postscript to “Maths and AI” (Leslie Ann Goldberg) | 7 Oct 2026 |
| What should we tell our students? (Álvaro Lozano-Robledo) | 8 Oct 2026 |
| OpenAI, the Partition Principle, and mathematics (Asaf Karagila) | 8 Oct 2026 |
| This is normal bad work (Sophie Huiberts) | 12 Oct 2026 |
| The future of mathematics (Jeremy Avigad) | 5 Oct 2026 (pre-release) |
| To grieve, or not to grieve? (Kevin Buzzard), comments | 1–9 Oct 2026 |
| AHM statement on OpenAI’s release | 7 Oct 2026 |
| AGMAI statement | 6–7 Oct 2026 |
| A Severe Misalignment of AI in Mathematics (25 Fields medallists) | 11 Sep 2026 |
| Transformation (Quanta Magazine) | 6–9 Oct 2026 |
| OpenAI withdraws three preprints (Retraction Watch) | 8 Oct 2026 |
| OpenAI’s 372 math results (Implicator) | 8 Oct 2026 |
| “How much beauty have we lost?” (The Decoder) | 10 Oct 2026 |
| Simons Foundation launches $15M effort | 15 Oct 2026 |
