Today's Top 30 AI Stories

Sundar Pichai introduced Gemini Omni, which reasons about physics, history, and context to generate realistic scenes. It rolls out today to Google AI subscribers via the Gemini app and YouTube Shorts. The model combines intuitive understanding with knowledge.

NVIDIA unveiled SANA-WM, a 2.6B-parameter open-source world model for controllable 60-second video generation. It runs on a single consumer GPU with precise camera control from one image and text. The model targets embodied AI, robotics, and simulations.

Demis Hassabis was shown reacting live to the decision to release all known protein structures via AlphaFold. The moment highlights the impact of AI on biology. It underscores DeepMind's Nobel-winning contributions to science.

Zen Technologies launched India's first modular AI counter-drone system capable of detecting drones over 15 km. It handles swarms, jamming, and intercepts with full indigenous design. The system strengthens national defense capabilities.

Microsoft released an open-source AI Agent Governance Toolkit for runtime policy enforcement and zero-trust security. It covers all OWASP Agentic risks with thousands of tests. The toolkit supports safe deployment of autonomous agents.

Google launched Gemini Spark, a 24/7 personal AI agent that works autonomously in the background. It checks with users before major actions and runs even when devices are off. The agent is rolling out via the Gemini app.

JPMorgan CEO Jamie Dimon stated AI will cure cancer, stop car crashes, and create new materials. He predicted kids will work 3.5 days a week and live to 100. The comments come from a major financial leader.

Greg Brockman emphasized OpenAI's autonomous solution to the Erdős planar unit distance problem. The model outperformed decades of human assumptions with new constructions. This represents a landmark in AI-driven scientific discovery.

Jensen Huang explained how AI handles routine tasks to free humans for meaningful work. The comments focus on productivity gains across industries. NVIDIA continues to position itself at the center of the AI infrastructure boom.

Posts highlight China's strong open-source AI ecosystem with models like DeepSeek outperforming expectations. Multiple high-signal accounts note the gap with Western closed models. This fuels global competition in accessible AI.

Anthropic shared a 2-hour official workshop on building reliable Claude agents with memory and tool use. The session covers terminal access, hooks, and large codebases. It is presented by the engineers behind Claude Code.

Mistral AI acquired Austrian startup Emmi AI for physics-aware simulation models. The deal strengthens Mistral's industrial AI offerings in Europe. It positions the company as a key player in sovereign European AI.

GitHub introduced the first official Agentic AI Developer certification (GH-600). It focuses on building, supervising, and integrating AI agents in real workflows. The credential formalizes a new engineering discipline.

Ukraine's AI-powered Delta system provides real-time tracking of every Russian soldier. Drones receive coordinates instantly for targeted strikes. The platform demonstrates battlefield AI superiority.

Tencent released Hy-MT2, open-source multilingual models in three sizes up to 30B parameters. The models outperform larger competitors across 33 languages. They run efficiently on mobile hardware.

Greg Isenberg shared a viral list of 36 major startup opportunities driven by AI. Categories include agent employees, elder tech, and physical AI. The post sparked widespread discussion on the next wave of AI businesses.

Brian Roemmele detailed how NVIDIA's new open-source SANA-WM enables controllable physics-rich worlds on consumer GPUs. It supports embodied AI and zero-human simulation loops. The model was trained in just 15 days.

Capcom confirmed generative AI is used solely for repetitive work like research and notes. All creative game content remains fully human-driven. The policy aims to boost efficiency without replacing artistry.

Cleo Abram released the full Demis Hassabis conversation covering AlphaFold, drug discovery, and military AI use. The interview explores AI's hardest problems and human-AI boundaries. It has garnered massive engagement.

Japan's ruling party approved a proposal for AI and blockchain-powered 24/7 automated commerce. AI agents will handle purchases with verifiable blockchain payments. The plan aims to modernize financial infrastructure.

Hitachi announced a strategic partnership with Anthropic to combine AI with industrial expertise. The deal includes training 100,000 AI professionals and deploying Claude. It accelerates enterprise AI adoption in Japan.

Kawasaki Heavy Industries partnered with NVIDIA to establish a robotics hub on the US West Coast. The collaboration focuses on physical AI for advanced robotics. It strengthens Japan-US tech ties in embodied intelligence.

Dograh, a fully open-source voice agent platform, was launched as a free self-hosted alternative. It supports any LLM, STT, and TTS with no per-minute fees. The tool runs entirely on user infrastructure.

Lisa Su presented AMD's compact AI PC capable of running 200B parameter models locally. The device targets developers needing powerful on-device inference. It highlights AMD's push into the AI hardware market.

Microsoft AI CEO Mustafa Suleyman forecasted human-level AI performance on most professional tasks within 18 months. Accounting, legal, and project management are highlighted as vulnerable. The prediction comes from a top industry executive.

Google published 13 official skills for AI agents compatible with Claude, Cursor, and Copilot. The free open-source skills enable advanced task automation and workflow integration. They lower the barrier for building capable agents.

Cohere launched Command A+, its most powerful open-source LLM optimized for minimal hardware. The model is now the fastest frontier-class open model available. It strengthens Canada's position in global AI.