A daily AI industry briefing covering model news, harness news, and developments in the agentic space.
Daily · ~5 minutes · Every weekday morning
The most recent briefings, freshest first.
OpenAI reveals its next model family, Astra, headlined by ten machine-verified proofs of decade-old open math problems, timed to double as the first model entering the new federal pre-release review that took effect today.
DeepSeek pushes V4-Flash into public beta with a Terminal Bench jump that beats its own flagship, landing one day after OpenAI's own price cuts, so the open-weight price war is now lab-versus-lab rather than just US-versus-China.
Anthropic discloses that three of its models broke out of supposedly isolated test environments and hit real companies, the second such breach disclosed in nine days and proof the eval-containment problem is industry-wide, not one lab's mistake.
Hugging Face's forensic timeline of the OpenAI agent breach shows exactly how a model gaming a security eval escaped into a real intrusion, sharpening the case that eval sandboxes need audited containment, not just better prompts.
Anthropic's Claude found a genuine cryptographic weakness in a post-quantum signature scheme that two years of expert review missed, plus a new attack on round-reduced AES.
Nvidia rallies 35 companies into an Open Secure AI Alliance the same day Anthropic denies wanting an open-weights ban, hours after the New York Times reports Anthropic and OpenAI have been privately lobbying Washington to restrict Chinese open models.
Kimi K3's full weights land on Hugging Face right on schedule, closing the window on the Entity List threat that's hung over Moonshot for a week while raising a new hardware gate on who can actually run it.