Fetching from the wire…
01
02
03
04
05
06
07
08
Models2026-08-06 · source-backed
Quanta's August 3 piece tallies the assault: OpenAI found a counterexample to Erdős's 1946 unit distance conjecture on May 20, then Astra produced 10 further advances. Google DeepMind evaluated 700 open conjectures in January, solving four and recovering nine forgotten solutions; in May a 21-researcher team autonomously resolved 9 of 353 open problems at a few hundred dollars each. Kevin Barreto and Liam Price used GPT-5.2 and 5.2 Pro on Erdős 333 and 728 and co-authored a paper with Terence Tao on Problem 1196.
Each link below shares sources, entities, or timing with this story.
The number that reframes everything isn't ten. It's two thousand. OpenAI published "Ten advances in mathematics and theoretical computer science" on August 1, claiming an internal version of Astra produced new results on ten problems that had seen no progress on the main resul...
The company published "Pacing model development in an era of cyber-critical capabilities" on August 19, disclosing the pause on its latest deployment-bound models while it hardened and red-teamed research environments. The trigger was an unreleased model, Astra, plus a July in...
The Register put the two side by side on August 8. Read together, frontier safety policy isn't converging on a posture, it's splitting by risk domain, with each lab tightening where its own evals scared it and loosening where false positives cost product usability. That's evid...
Executive Order 14409 gave the government 60 days to produce three things: a classified frontier-model cyber benchmarking process from NSA/CISA/NIST, a voluntary frontier AI disclosure framework, and an OPM federal cyber workforce plan. As of 00:00Z August 1 there were no Fede...
OpenAI disclosed August 7 that internal evaluations of the upcoming Astra model show agentic coding and cybersecurity performance strong enough that it can no longer rule out the Critical cybersecurity level in its Preparedness Framework, a first. Every prior frontier model in...
Astra can take an experimental idea, implement it in OpenAI's codebase, run the experiment and report results, and can follow up on a paper in work that used to take a human researcher about a week. Altman told TIME the company is "not quite yet" at AGI but will declare it int...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.