Google DeepMind releases its AI Control Roadmap, a defense-in-depth security framework that treats internal AI agents as potential insider threats, using trusted AI supervisors to monitor behavior in real-time after analyzing over one million coding tasks revealing most security flags stem from misinterpretation, not malicious intent.
AI startup General Intuition is raising $300M at a $2B valuation, backed by Jeff Bezos and Eric Schmidt, leveraging 2 billion gaming videos annually to train embodied AI agents with unique spatial-temporal reasoning capabilities.
Amazon is entering the AI chip market as a seller, with its AI chief confirming talks to supply custom Trainium3 chips to outside companies, directly challenging Nvidia's dominance in the booming AI infrastructure space.
A top-rated GitHub machine learning repository just surpassed 5,700 stars with a full 2025 update, offering a complete curriculum from NumPy basics to deep learning with TensorFlow, including three hands-on projects covering heart disease classification, bulldozer price prediction, and dog breed recognition.
Sina Weibo's tiny 3-billion-parameter AI model, VibeThinker-3B, is shaking up the industry by claiming to rival Google, OpenAI, and Anthropic on key reasoning benchmarks, but skeptics warn the results may reflect 'benchmaxxing' rather than real-world capability.
Qualcomm unveils the Snapdragon Reality Elite chip at Augmented World Expo, delivering 48 TOPS of on-device AI performance, 4.4K-per-eye visuals at 90fps, and 20% better battery life to power next-gen AR headsets, including the newly launched XREAL Project Aura glasses running Google's Android XR platform.
CoreWeave unveils 'Valvey' and 'Racky,' two groundbreaking infrastructure innovations designed to eliminate AI data center bottlenecks by revolutionizing coolant control and rack management for Nvidia's powerful 72-GPU Vera Rubin NVL72 system.
Factory 2.0 launches AI-powered 'software factories' that autonomously manage the entire software development lifecycle — from bug reports to deployment — with major enterprises like NVIDIA, Adobe, and Blackstone already running the systems in production, signaling a major shift where engineers govern AI rather than write code themselves.
IBM, NVIDIA, and Red Hat are spearheading DocLang, a groundbreaking open AI-native document format built to replace PDFs in enterprise AI pipelines, delivering up to 30x lower token costs while preserving semantic structure and slashing hallucination risks.
ErdosBench launches as a new benchmark challenging AI systems to tackle complex Erdős-style mathematical research problems, featuring a public smoke test with 14 problems and a full private corpus of 226 problems with blind evaluation tools.