Daily AI Briefing
The 6 top stories of the 120 AIBriefs curated that day.
Meta Superintelligence Labs shipped the Muse Wonder series, its second-generation family, roughly 14 months after the lab was founded. Muse Wonder is available via the Meta Model API and Muse Code; Muse Wonder Full is limited to Muse Code subscribers.
Gemini 4 Argon rolls out first to trusted cyber defenders via the Fairwind Program, priced at an introductory $2/M input and $10/M output tokens. It tops Text Arena at 1525 pts and the Vals Index at 68.9%, with a 15% hallucination rate on Artificial Analysis.
OpenAI published 722 manuscripts (719 after three withdrawals) organized into 372 families, covering 90 of the top 500 open problems per Proof of Atlas. The single model averaged three hours of compute per solution after being asked to attempt ~4,000 problems, with Lean proof formalizations on GitHub.
OpenAI says it began training a new internal model on August 28; within roughly 24 days it resolved more than 100 long-standing open problems across most areas of mathematics. Results are being held back pending discussions with the math community and an independent advisory group.
Anthropic's Frontier Red Team says a 9B model got inference running on 8 of 9 hardware types tested, and a 26B model split across 15 $55 Android TV boxes ran at 0.39 tokens/sec. Only GLM-5.2 completed end-to-end replication across 8 emulated devices, in 4 of 5 runs.
Anthropic reports its new methods reduced data harvested by distillers by roughly 90 to 99 percent. It names Zhipu, which distilled cyber capabilities into GLM-5.3, and says one network ran over 20,000 fake accounts at once.
Also that day
102 more stories that day — open the live feed.
Get tomorrow's AI brief in your inbox