An unreleased OpenAI model, informally called 'Astra', reportedly solved ten open problems in mathematics, quantum complexity theory, and theoretical computer science, including one unsolved for about…
Tom Zahavy, a Google DeepMind researcher and co-author of AlphaProof, argues in an ICML 2026 position paper that large language models have mechanized induction and deduction but lack a mechanism for …
Astra, an AI model from OpenAI, claims to have solved 10 open math problems, but the lack of formal proof artifacts, benchmarks, or peer review raises skepticism, according to an analysis. The verific…
OpenAI's internal model Astra has solved ten open problems in mathematics and theoretical computer science, with proofs formalized as machine-checkable Lean certificates. The total compute cost was ro…
A mathematician argues that the volume of mathematical research has outpaced human ability to verify it, creating a crisis of trust, and proposes 'AIcademia'—AI-assisted proof checking and conjecture …
OpenAI previewed Astra, an unreleased model built for long-running research tasks, by publishing 10 claimed advances in mathematics and theoretical computer science on August 1, with an internal versi…
OpenAI published ten advances in mathematics and theoretical computer science, including results on sphere packing, non-sofic groups, and a counterexample to Connes' rigidity conjecture, with Lean cer…
OpenAI disclosed on August 1 that an internal version of Astra, described as its next major model, produced ten new results in mathematics and theoretical computer science, with problems that had seen…
OpenAI announced the solution to ten open problems in mathematics, all discovered by a yet-unreleased model, marking a significant advance in AI's capability to do frontier mathematical research. The …
OpenAI announced on August 1st that an internal version of its next major model, Astra, produced new results for 10 longstanding problems in mathematics and theoretical computer science, a claim suppo…
Lean's kernel had a soundness bug that allowed a proof of False, exploited by an AI-assisted disproof of the Collatz conjecture; the bug was fixed within an hour of the report. The Lean development te…
An AI-generated formal proof in Lean claiming to solve the Collatz problem was found to exploit a bug in the Lean kernel, allowing any statement to be proven. The proof's author was aware of the sound…
A new analysis argues that large language models (LLMs) function as a 'mutator' or fuzzer, producing rubbish by design and requiring an oracle program to verify correctness, with the key benchmark bei…
Rocq is better than Lean for program verification, according to a LangSec keynote slide by an unnamed author, because Rocq natively supports executable coinductive types and cofixpoints in Type, while…
Terence Tao's International Congress of Mathematicians 2026 lecture on Goodhart's law and verification bottlenecks in mathematics maps directly onto AI code generation, arguing that AI-scale proof gen…
Amazon is making the largest single donation in the history of the Lean Focused Research Organization (FRO) to support the development of Lean, a programming language that enables mathematical proofs …
Anthropic mathematician Levent Alpöge used the Claude Fable 5 AI model to generate a 216-character polynomial map in three variables that disproves the 87-year-old Jacobian conjecture, marking the mos…
On 2026-07-22, ChatGPT and Claude generated counterexamples to open conjectures associated with Erdős and Grothendieck, with some verified in Lean, the formal proof language. Terence Tao published a p…
Harvard mathematician Levent Alpöge used Anthropic's Fable model to produce a 216-character counterexample to the Jacobian Conjecture, an 87-year-old open problem proposed by Eduard Ott-Heinrich Kelle…
Harmonic, the AI startup co-founded by Robinhood CEO Vlad Tenev and Tudor Achim, announced on July 28 that its reasoning model Aristotle achieved gold medal-level performance on the 2025 International…