今日精读

"An internal version of Astra, @OpenAI's next major model family, solved 10 major open problems in mathematics, quantum complexity, and theoretical computer science. We believe it will be a major step for scientific reasoning." / "yes, nonsofic groups exist: this statement is one of many new beautiful results proved by Astra... We're releasing 10 such Astra proofs, complete with lean certificates and CoT walkthroughs for each of them."

"Large changes are hard to review. Smaller pull requests make them easier to ship. Stacked PRs are now on GitHub."

"All this talk about AGI destroying moats and everything made me want to ship again... I'd never imagine making Adobe Premiere, Final Cut Pro or Capcut but on the web... So today I just tried and vibecoded them into Photo AI! A fully functional video editor but the cool twist is this one can actually generate videos with you or your trained models in it directly in your video editor."

"Sharing our initial findings on a reported vulnerability. Our recommendation is to continue to use your Bitkey normally. The reported vulnerability would require exceptional circumstances to exploit, and can only occur at a specific narrow time during inheritance setup... they would not have enough cryptographic material to access funds. This is Bitkey's defense-in-depth in action."

"The Modular Platform unifies AI development and deployment with an open, fully-integrated suite of libraries and tools for model serving and GenAI scaling. Over 450,000 lines of code from 6,000+ contributors. Includes MAX GPU and CPU kernels, plus the Mojo standard library. OpenAI-compatible REST API and Kubernetes-ready containers for NVIDIA and AMD."

"Thank you, @cursor_ai, for providing free credits to several FFmpeg developers. These credits will support their ongoing work on FFmpeg, including development and code review. We greatly appreciate Cursor's support for FFmpeg and its community."

"Banning open source model, will 'remove capabilities for the defenders'. When OpenAI's models breached Huggingface, a Chinese open model is what cleaned up the mess. Anthropic's Fable 5 refused, so Hugging Face turned to Nvidia's quantized GLM 5.2."

"DeepSeek-V4-Flash-0731 is now over 2x faster than yesterday on Ollama's cloud!"

"It's been a busy couple of weeks! ICYMI, here's the recap — Gemini Robotics 2... Gemini 3.5 Flash-Lite is our fastest, most cost-effective model optimized for high-speed agentic workflows... Gemini 3.5 Flash Cyber is our new, cost-effective model optimized for finding and fixing software vulnerabilities at scale, available exclusively to governments and trusted partners... Lyria 3.5 is the newest music model from Google Deepmind... @Gemini_Notebook launched Collections."