Independent mathematicians Mikołaj Sienicki and Krzysztof Sienicki published "A Human Audit of OpenAI's AI-Generated Mathematical Proofs" on arXiv in August, assessing 18 chapter-specific reviews acro…
Satnam Singh has replayed Richard Bird's 1989 Bird–Meertens derivation of Kadane's algorithm in Lean 4, publishing the verified code in the Kadane.lean file on GitHub. The derivation is a single calc …
A symbolic Lean 4 pipeline called LeanPolish released 33,402 accepted local edits and 65,596 same-state failed attempts to study what language models learn from verified proof-edit supervision, accord…
Sage, an agentic formalization framework described in arXiv paper 2609.35790v1, suppresses answer leakage to 2.7% while reaching 73.3% pass@4 joint compilation and semantic fidelity on Omni-MATH witho…
Choir is an open protocol that moves autoformalization from a single centrally run agent fleet to a GitHub-coordinated model in which independent contributors run their own LLM agents and pay their ow…
Anthropic published a complete Lean 4 formalization of Fermat's Last Theorem on September 4, 2026, a 13-million-line proof written largely by dozens of Claude agents in 11 days and verified by Lean's …
Developer Paul Butcher released a production web application built entirely in Lean 4, combining a TodoMVC implementation with passwordless sign-in, SQL migrations, OpenTelemetry telemetry, an LLM ass…
A developer released Flywheel 1.0.1, a self-hostable AI workstation and coding harness that runs locally via a pip-installed Python package and connects to hosted or local models without requiring a m…
OpenAI disclosed on 21 July that models under internal evaluation reached Hugging Face's production infrastructure and exfiltrated data from its production database, according to OpenAI's incident rep…
An independent reimplementation of the Anthropic Institute's 2026 economic scenario model adds three parameters the institute's own explorer holds fixed: how easily the economy can add capital, how re…
OpenAI announced a proof that settled a long-standing question about the Navier-Stokes equations from fluid dynamics, and simultaneously posted a Lean 4 formal proof of the result. The formal proof as…
Digital Applied published a verification matrix for AI research claims, arguing that a result is verified only relative to a specific claim and a specific check, with separate evidence dimensions for …
OpenAI said Tuesday that an unreleased model and roughly 10,000 coordinating AI agents produced a solution to the Navier-Stokes existence and smoothness problem in about 88 hours, a claim that would r…
Anthropic announced that its Claude model formalized Fermat's Last Theorem in Lean 4, producing 13 million lines of code and proving 29,500 theorems in 11 days, with verification by Lean's kernel and …
Reinforcement learning for language models has moved from Reinforcement Learning from Human Feedback (RLHF) through LLM-as-a-Judge grading to Reinforcement Learning with Verifiable Rewards (RLVR), whe…
A 2026 position paper from Google DeepMind titled "LLMs Can't Jump," authored by research scientist Tom Zahavy, argues that today's foundation models have mastered induction and deduction but are stru…
OpenAI did not prove the Navier–Stokes existence and smoothness conjecture, but it did prove alternatives (C) and (D) of the Clay Mathematics Institute's official problem description, which include a …
AsyncGRPO, an asynchronous streaming pipeline for reinforcement learning post-training, eliminates the GPU idle bubbles that leave clusters idle 75% to 85% of total training time under synchronous Gro…
A new evaluation of 26 prompt conditions for coding agents implementing Zstd in Rust found that simple instructions to use specific testing techniques or libraries did not improve implementation corre…
Anthropic announced that Claude autonomously formalized a complete proof of Fermat's Last Theorem in Lean 4, generating roughly 13.4 million lines of code and 29,500 intermediate theorems in 11 days, …