As of: 2026-10-10 14:45 CEST. Researched and written by AI agents (Claude Opus 5.5 in Claude Code). Human editor: Adam Bicz. Canonical page: https://postcutoff.com/news/science/3/ # AI news: Science & math, page 3 199 events in Science & math of 1,105 in the log, newest first. Page 3 of 4, 50 events per page, grouped by the day each event happened. ## Wednesday 5 August 2026 - [Sendov's 1958 conjecture on polynomial roots proved with GPT-5.6 Pro](https://postcutoff.com/e/2026-08-05-sendov-conjecture-proved/) (Science & math; OpenAI; major). It is a classic, well-known conjecture closed by AI, with the leading expert on the problem verifying and formalising the result. Source: https://www.proofatlas.ai/papers/sendov-conjecture/SENDOV_CONJECTURE_PROOF_AUGUST_5_2026.pdf - [HRT conjecture (1996) disproved](https://postcutoff.com/e/2026-08-05-hrt-conjecture-disproved/) (Science & math; Faulhuber, Petersen, van Velthoven, Voigtlaender (academic mathematicians)). It is a clean example of the "AI-assisted, human-written" mode of discovery. Source: https://arxiv.org/abs/2608.05044 ## Tuesday 4 August 2026 - [Alexander Perry disproves the period-index conjecture](https://postcutoff.com/e/2026-08-04-period-index-conjecture-false/) (Science & math; Alexander Perry, OpenAI; major). The period-index conjecture was a central open question on Brauer groups of function fields. Source: https://arxiv.org/abs/2608.03684 ## Monday 3 August 2026 - [Planar Schiffer and Pompeiu conjectures disproved by two independent groups](https://postcutoff.com/e/2026-08-03-schiffer-pompeiu-conjectures-disproved/) (Science & math; Matthew Colbrook, George Stepaniants, Gonzalo Cao-Labora, Jaume de Dios Pont; major). Schiffer's conjecture is a classic rigidity question and appears on Yau's list of open problems. Source: https://arxiv.org/abs/2608.01579 ## Saturday 1 August 2026 - [OpenAI's unreleased 'Astra' model claims ten advances in maths and theoretical CS, with Lean proofs](https://postcutoff.com/e/2026-08-01-openai-astra-ten-advances/) (Science & math; OpenAI; historic). It moved the frontier from individual AI-assisted results to a lab producing batches of significant theorems. Source: https://openai.com/index/ten-advances-in-mathematics/ ## August 2026, day not recorded - [OpenAI hosts a closed-door summit of about 40 mathematicians on the future of mathematics](https://postcutoff.com/e/2026-08-01-openai-future-of-mathematics-summit/) (Science & math; OpenAI). A lab whose models were producing new maths results was now directly engaging the profession about its future. Source: https://www.washingtonpost.com/technology/2026/08/19/mathematicians-ask-whats-left-humans-when-ai-can-do-math-research/ ## Wednesday 29 July 2026 - [Maxwell's conjecture on equilibria of point charges is false](https://postcutoff.com/e/2026-07-29-maxwell-point-charge-conjecture-false/) (Science & math; Philip Arathoon, Gavin Ball, Matthew D. Kvalheim, OpenAI). It is a long-standing question at the boundary of physics and real algebraic geometry, and one of the few 2026 AI-assisted results in classical physics. Source: https://arxiv.org/abs/2607.27197 ## Monday 27 July 2026 - [Neurosurgery resident uses GPT-5.6 Sol to prove Crouzeix's conjecture in a 16-hour autonomous run](https://postcutoff.com/e/2026-07-27-crouzeix-conjecture-proved/) (Science & math; OpenAI; major). Along with #1196, it showed that frontier models let amateurs resolve famous problems, which upended assumptions about who can do research mathematics. Source: https://arxiv.org/abs/2608.03841 ## Friday 24 July 2026 - [Terence Tao's ICM 2026 public lecture 'Mathematics in the age of AI' calls a crisis in the foundations of mathematical values](https://postcutoff.com/e/2026-07-24-tao-icm-mathematics-in-the-age-of-ai/) (Science & math; International Congress of Mathematicians, UCLA). It was the most prominent framing of AI-and-mathematics at the field's main quadrennial event. Source: https://teorth.github.io/tao-web/slides/age-of-ai-icm-2026.pdf - [Rota's 1970 unimodality conjecture for matroid flats disproved, with ChatGPT 5.6 Pro suggesting a key construction](https://postcutoff.com/e/2026-07-24-rota-flats-unimodality-disproved/) (Science & math; OpenAI, Anthropic). Rota's conjecture was one of the best-known open problems about matroids. Source: https://arxiv.org/abs/2607.22515 - [Hessian conjecture refuted in five variables, derived from Claude-found Jacobian counterexample](https://postcutoff.com/e/2026-07-24-hessian-conjecture-counterexample/) (Science & math; Independent researchers). It shows how AI-found results feed quickly into human follow-up work. Source: https://arxiv.org/abs/2607.22198 ## Thursday 23 July 2026 - [AI systems score a perfect 42/42 at IMO 2026, officially graded](https://postcutoff.com/e/2026-07-23-imo-2026-ai-perfect-scores/) (Science & math; Huawei, Xiaohongshu (RedNote); historic). Olympiad math is now saturated as an AI benchmark just one year after the first gold-level results; attention shifts to research-level math (FrontierMath Tier 4, Erdős problems). Source: https://techxplore.com/news/2026-07-ai-humans-score-math-contest.html ## Monday 20 July 2026 - [Claude Fable 5 finds a counterexample to the Jacobian conjecture in dimension 3](https://postcutoff.com/e/2026-07-20-jacobian-conjecture-counterexample/) (Science & math; Anthropic; historic). The Jacobian conjecture is one of the most famous open problems in algebra. Source: https://arxiv.org/abs/2608.00222 ## Friday 17 July 2026 - [GPT-5.6 Sol Ultra proves the 50-year-old cycle double cover conjecture](https://postcutoff.com/e/2026-07-17-cycle-double-cover-conjecture-proved/) (Science & math; OpenAI; historic). Along with the Jacobian counterexample the same week, it marked the point where famous named conjectures, not just Erdős-list problems, began falling to AI. Source: https://cdn.openai.com/pdf/04d1d1e4-bc75-476a-97cf-49055cd98d31/cdc_proof.pdf ## Tuesday 14 July 2026 - [Berkeley's Phillip Kerger uses GPT-5.6 Sol and a 10-page prompt to close a 30-year gap in derivative-free convex optimization](https://postcutoff.com/e/2026-07-14-gpt-5-6-zeroth-order-convex-lower-bound/) (Science & math; UC Berkeley, OpenAI). Most of the summer's AI-math headlines came from the labs themselves. Source: https://arxiv.org/abs/2607.13335 ## July 2026, day not recorded - [AI-assisted counterexample answers Grothendieck's question on finite flat group schemes, merged into Mathlib](https://postcutoff.com/e/2026-07-01-grothendieck-group-scheme-counterexample/) (Science & math; OpenAI, Anthropic). Kevin Buzzard said this mattered more to him than the Erdős results because it lies in "an area of mathematics that I personally find more interesting". Source: https://antieau.github.io/2026/08/10/akhil-mathew-ai.html ## Monday 29 June 2026 - [Machine-learning screen predicts two new kagome superconductors, confirmed in the lab](https://postcutoff.com/e/2026-06-29-ml-screened-kagome-superconductors/) (Science & math; Aalto University, Rice University). It is a modest but clean prediction-then-confirmation loop in superconductor research, a field full of hype. Source: https://www.sciencedaily.com/releases/2026/07/260701205006.htm ## Friday 5 June 2026 - [Computationally designed broad coronavirus vaccine is safe and immunogenic in first human trial](https://postcutoff.com/e/2026-06-05-ai-designed-coronavirus-vaccine-trial/) (Science & math; University of Cambridge, DIOSynVax). It is an early human validation of computer-designed vaccine antigens, relevant to pandemic preparedness. Source: https://www.diosynvax.com/ ## Wednesday 27 May 2026 - [Erdős–Szemerédi sum-product conjecture shown false over the reals](https://postcutoff.com/e/2026-05-27-sum-product-conjecture-false-over-reals/) (Science & math; OpenAI; major). It shows AI ideas spreading into human research and then being reproduced autonomously: a feedback loop between AI and human mathematicians. Source: https://arxiv.org/abs/2605.28781 ## Wednesday 20 May 2026 - [OpenAI model disproves Erdős's 80-year-old unit distance conjecture](https://postcutoff.com/e/2026-05-20-ai-disproves-erdos-unit-distance-conjecture/) (Science & math; OpenAI; historic). This is the moment AI crossed from solving competition problems to settling a famous open research conjecture, reshaping debate about AI's role in mathematics. Source: https://openai.com/index/model-disproves-discrete-geometry-conjecture/ ## Tuesday 12 May 2026 - [GPT-5.5 Pro finds counterexample disproving McKean's 1966 conjecture and the Gaussian completely monotone conjecture](https://postcutoff.com/e/2026-05-12-gaussian-completely-monotone-disproved/) (Science & math; OpenAI). It is part of 2026's striking pattern of AI counterexamples: models proved especially good at finding objects that break long-believed conjectures. Source: https://arxiv.org/abs/2605.11656 ## Saturday 9 May 2026 - [Google DeepMind's 'AI co-mathematician' sets FrontierMath Tier 4 record and helps resolve a Kourovka Notebook problem](https://postcutoff.com/e/2026-05-09-deepmind-ai-co-mathematician/) (Science & math; Google DeepMind, University of Oxford). FrontierMath Tier 4, built to resist AI for years, was nearly half solved 18 months after launch. Source: https://arxiv.org/abs/2605.06651 ## Sunday 3 May 2026 - [Amateur with GPT-5.4 Pro 'vibe-maths' a 60-year-old Erdős conjecture on primitive sets](https://postcutoff.com/e/2026-05-03-erdos-1196-primitive-sets/) (Science & math; OpenAI; major). It was the first AI solution to a well-known, decades-old Erdős conjecture that specialists had actively worked on, not just an obscure entry. Source: https://arxiv.org/abs/2605.00301 ## May 2026, day not recorded - [GPT-5.5 Pro-assisted construction lowers the smallest known Borsuk counterexample dimension from 64 to 63](https://postcutoff.com/e/2026-05-01-borsuk-conjecture-dimension-63/) (Science & math; OpenAI). It is a clean, checkable improvement to a well-known geometry record. Source: https://github.com/maaxgrin/borsuk-63-counterexample ## Tuesday 10 March 2026 - [AlphaEvolve improves lower bounds for nine classical Ramsey numbers](https://postcutoff.com/e/2026-03-10-alphaevolve-ramsey-lower-bounds/) (Science & math; Google). Small Ramsey numbers are among the most-studied computational problems in combinatorics; AI-found improvements across nine at once showed the reach of evolutionary LLM search. Source: https://arxiv.org/abs/2603.09172 ## March 2026, day not recorded - [Math Inc's Gauss formalises Viazovska's sphere-packing proofs in dimensions 8 and 24, fixing errors in the originals](https://postcutoff.com/e/2026-03-01-gauss-sphere-packing-formalization/) (Science & math; Math Inc; major). AI autoformalization reached Fields-Medal-level proofs, strengthening the case that formal verification can keep up with the flood of AI-generated mathematics. Source: https://arxiv.org/abs/2604.23468 - [RAVEN machine-learning pipeline validates 118 new planets in TESS data](https://postcutoff.com/e/2026-03-25-raven-tess-118-new-planets/) (Science & math; University of Warwick). It continues AI's role as the main filter for exoplanet surveys. Source: https://warwick.ac.uk/news/pressreleases/ai-approach-uncovers-dozens-of-hidden-planets/ ## Saturday 28 February 2026 - [Claude Opus 4.6 solves an open Hamiltonian-cycle problem, Donald Knuth reports](https://postcutoff.com/e/2026-02-28-knuth-claudes-cycles/) (Science & math; Anthropic, Stanford University; major). Coming from one of computing's most respected and AI-sceptical elders, the note became a cultural marker that frontier LLMs could contribute original mathematical constructions. Source: https://www-cs-faculty.stanford.edu/~knuth/papers/claude-cycles.pdf ## Saturday 14 February 2026 - [AI solves about half of 10 unpublished research problems in the 'First Proof' challenge](https://postcutoff.com/e/2026-02-14-first-proof-challenge/) (Science & math; Google DeepMind, OpenAI). It gave a cleaner measure than olympiads of whether AI can do research maths: at the time, about half the time. Source: https://1stproof.org/ ## Friday 13 February 2026 - [GPT-5.2 conjectures, and an OpenAI model proves, that 'single-minus' gluon tree amplitudes are nonzero](https://postcutoff.com/e/2026-02-13-gpt-5-2-single-minus-gluon-amplitudes/) (Science & math; OpenAI, Institute for Advanced Study, Harvard University, University of Cambridge, Vanderbilt University). It is a showcase of LLMs as conjecture engines in theoretical physics, though critics dispute its physical significance and the AI's share of the insight. Source: https://openai.com/index/new-result-theoretical-physics/ ## Wednesday 11 February 2026 - [DeepMind's Aletheia agent and Gemini Deep Think report autonomous Erdős solutions and new physics and CS results](https://postcutoff.com/e/2026-02-11-deepmind-aletheia-deep-think-science/) (Science & math; Google DeepMind; major). Google's answer to OpenAI's maths push showed that several labs could now produce publishable research-level results, although many were on problems nobody had seriously attacked. Source: https://deepmind.google/blog/accelerating-mathematical-and-scientific-discovery-with-gemini-deep-think/ ## Tuesday 10 February 2026 - [Isomorphic Labs unveils IsoDDE drug-discovery engine, hailed as 'an AlphaFold 4'](https://postcutoff.com/e/2026-02-10-isomorphic-isodde/) (Science & math; Isomorphic Labs, Google DeepMind). Shows AI structural biology continuing to advance fast, but also a shift from DeepMind's open-science AlphaFold tradition toward proprietary commercial models. Source: https://www.nature.com/articles/d41586-026-00365-7 ## Thursday 5 February 2026 - [GPT-5 autonomously runs 36,000 experiments in Ginkgo's cloud lab, cutting protein-synthesis cost 40%](https://postcutoff.com/e/2026-02-05-gpt-5-ginkgo-autonomous-lab/) (Science & math; OpenAI, Ginkgo Bioworks). It is a concrete, economically meaningful result from an LLM driving a physical lab end-to-end. Source: https://openai.com/index/gpt-5-lowers-protein-synthesis-cost/ ## Tuesday 6 January 2026 - [Erdős problem #728 solved near-autonomously by GPT-5.2 Pro and Harmonic's Aristotle, with a Lean proof](https://postcutoff.com/e/2026-01-06-erdos-728-gpt-5-2-aristotle/) (Science & math; OpenAI, Harmonic; major). It showed the combination of informal LLM reasoning with formal verification as a practical, trustworthy workflow for research maths that non-experts could run. Source: https://arxiv.org/abs/2601.07421 ## Monday 8 December 2025 - [Genuine AI-assisted solutions to Erdős problems begin](https://postcutoff.com/e/2025-12-08-ai-erdos-problems-wave-late-2025/) (Science & math; Harmonic, Google DeepMind, OpenAI; major). Erdős problems became the first large, public, verifiable scoreboard for AI in research mathematics, and set the stage for 2026's much larger results. Source: https://arxiv.org/abs/2510.19804 ## Saturday 6 December 2025 - [AxiomProver produces machine-checked Lean proofs for all 12 Putnam 2025 problems](https://postcutoff.com/e/2025-12-06-axiomprover-putnam-2025/) (Science & math; Axiom Math). Formal verification removes grading disputes like those around informal IMO proofs, and showed formal provers catching up with informal LLMs on hard competition maths. Source: https://github.com/AxiomMath/putnam2025 ## December 2025, day not recorded - [AI searches 100 million Hubble images in 2.5 days, finding ~1,400 anomalies including 800+ never described](https://postcutoff.com/e/2025-12-01-ai-finds-1400-anomalies-hubble-archive/) (Science & math; European Space Agency). It shows the AI-first search pattern that surveys such as Rubin/LSST and Euclid will depend on. Source: https://esahubble.org/news/heic2603/ - [Physics Letters B paper built on a GPT-5 idea draws criticism that it tests the wrong thing](https://postcutoff.com/e/2025-12-01-hsu-gpt-5-physics-paper-dispute/) (Science & math; Michigan State University, OpenAI). It is a cautionary example: AI-originated ideas can pass peer review and still be wrong or misframed. Source: https://the-decoder.com/physicist-steve-hsu-publishes-research-built-around-a-core-idea-generated-by-gpt-5/ ## Thursday 20 November 2025 - [OpenAI publishes 'Early science acceleration experiments with GPT-5'](https://postcutoff.com/e/2025-11-20-openai-gpt-5-science-acceleration/) (Science & math; OpenAI). It marked a shift from benchmark claims to documented research contributions, and set the template for OpenAI's 2026 "OpenAI for Science" results. Source: https://arxiv.org/abs/2511.16072 ## Wednesday 5 November 2025 - [Tao, Gómez-Serrano, Georgiev and Wagner test AlphaEvolve on 67 maths problems](https://postcutoff.com/e/2025-11-05-alphaevolve-tao-67-problems/) (Science & math; Google DeepMind, UCLA, Brown University). Coming from Tao, it gave the maths community a credible, balanced picture of what AI search could do, just before the 2026 surge. Source: https://arxiv.org/abs/2511.02864 ## November 2025, day not recorded - [Baker lab designs antibodies from scratch with atomic accuracy using RFdiffusion](https://postcutoff.com/e/2025-11-05-ai-designed-antibodies-rfdiffusion/) (Science & math; University of Washington Institute for Protein Design). Computational antibody design could replace months of animal immunisation and library screening in drug discovery. Source: https://www.nature.com/articles/s41586-025-09721-5 - [Edison Scientific's Kosmos AI scientist claims six months of research per run](https://postcutoff.com/e/2025-11-05-kosmos-ai-scientist/) (Science & math; Edison Scientific, FutureHouse). It is an early commercial "AI scientist" whose headline value is reproducing months-long analyses overnight. Source: https://edisonscientific.com/news/announcing-kosmos ## Friday 24 October 2025 - [Genentech's GNEprop screens 1.4 billion virtual compounds and finds 82 new antibacterial hits](https://postcutoff.com/e/2025-10-24-genentech-gneprop-antibacterial-screening/) (Science & math; Genentech, NVIDIA, Mila). Industrial-scale evidence for ML-guided antibiotic discovery, following MIT's halicin and abaucin work. Source: https://www.nature.com/articles/s41587-025-02814-6 ## Wednesday 22 October 2025 - [Agents4Science 2025, the first conference where AI must be first author and reviewer](https://postcutoff.com/e/2025-10-22-agents4science-conference/) (Science & math; Stanford University, Together AI). It made AI authorship an explicit, measurable experiment rather than a hidden practice, and produced data on the strengths and failure modes of AI reviewers. Source: https://arxiv.org/abs/2511.15534 ## Friday 17 October 2025 - [OpenAI researchers claim GPT-5 'solved' 10 Erdős problems](https://postcutoff.com/e/2025-10-17-gpt-5-erdos-problems-controversy/) (Science & math; OpenAI). The episode set the standard of scepticism for later AI maths claims, and led to Tao's public wiki tracking exactly what AI contributed to each Erdős problem. Source: https://techcrunch.com/2025/10/19/openais-embarrassing-math/ ## Thursday 16 October 2025 - [Google DeepMind partners with Commonwealth Fusion Systems to optimise and control the SPARC tokamak with AI](https://postcutoff.com/e/2025-10-16-deepmind-cfs-torax-fusion/) (Science & math; Google DeepMind, Commonwealth Fusion Systems). It moves AI plasma control from academic demos (TCV, DIII-D) into the commissioning plan of a privately built machine that aims for net energy. Source: https://deepmind.google/blog/bringing-ai-to-the-next-generation-of-fusion-energy/ ## Wednesday 15 October 2025 - [Google's C2S-Scale 27B model generates a new cancer-immunotherapy hypothesis confirmed in living cells](https://postcutoff.com/e/2025-10-15-c2s-scale-gemma-cancer-hypothesis/) (Science & math; Google Research, Google DeepMind, Yale University). It is evidence that scaling biological foundation models can yield testable, novel hypotheses, though only in vitro so far. Source: https://blog.google/technology/ai/google-gemma-ai-cancer-therapy-discovery/ ## Saturday 27 September 2025 - [Scott Aaronson credits GPT-5 with a key step in a quantum complexity proof](https://postcutoff.com/e/2025-09-27-aaronson-gpt-5-qma-proof/) (Science & math; UT Austin, CWI, OpenAI). It was a credible, first-person account from a top theorist of an LLM contributing a genuine idea to a published result. Source: https://arxiv.org/abs/2509.21131 ## Monday 22 September 2025 - [AlphaEvolve finds gadgets that prove new NP-hardness of approximation bounds for MAX-k-CUT](https://postcutoff.com/e/2025-09-22-alphaevolve-hardness-of-approximation/) (Science & math; Google Research, Google DeepMind). AI-found objects became ingredients of rigorous complexity-theory theorems, not just numeric improvements. Source: https://arxiv.org/abs/2509.18057 ## Wednesday 17 September 2025 - [DeepMind and mathematicians use neural networks to find new unstable singularities in fluid equations](https://postcutoff.com/e/2025-09-17-deepmind-unstable-singularities-fluids/) (Science & math; Google DeepMind, New York University, Stanford University, Brown University). It was the groundwork of the AI-plus-computer-assisted-proof approach to fluid blow-up that culminated in the disputed 2026 Navier–Stokes claims. Source: https://arxiv.org/abs/2509.14185 Previous page: https://postcutoff.com/news/science/2/ Next page: https://postcutoff.com/news/science/4/ Other views: All https://postcutoff.com/news/; Major only https://postcutoff.com/news/major/; Policy & safety https://postcutoff.com/news/policy-safety/; Science & math https://postcutoff.com/news/science/; Business https://postcutoff.com/news/business/; Model releases https://postcutoff.com/news/model-release/; Research https://postcutoff.com/news/research/; Chips & compute https://postcutoff.com/news/hardware-compute/; Products https://postcutoff.com/news/product/; Open source https://postcutoff.com/news/open-source/; Agents https://postcutoff.com/news/agents/; Robotics https://postcutoff.com/news/robotics/; Media generation https://postcutoff.com/news/media-generation/; Benchmarks https://postcutoff.com/news/benchmark/; Culture https://postcutoff.com/news/culture/; Milestones https://postcutoff.com/news/milestone/. Feeds: https://postcutoff.com/feeds/science.xml