General
- Google announces 40B$ investment in Texas (2025-2027).
- Anthropic is partnering with Nvidia and Microsoft; Claude will be available on Azure.
Research Insights
- OpenAI: Understanding neural networks through sparse circuits. We trained models to think in simpler, more traceable steps—so we can better understand how they work.
- SciAgent: A Unified Multi-Agent System for Generalistic Scientific Reasoning.
- Solving a Million-Step LLM Task with Zero Errors. They break problems into the smallest possible units, and use multi-agent voting for each step. This decreases the per-step error rate low enough that long task sequences can be handled.
LLM
- Google announces availability of Gemini 3 Pro. Record-setting scores across many measures (including ARC-AGI).
- Antigravity is their new agentic development platform, that uses Gemini 3 Pro. It plays the role of an IDE, but also allows one to manage agents.
Agents
Vision
- Meta releases Segment Anything 3 (SAM3), and introduces SAM3D for 3D inference on objects and people.
Image Synthesis
Video
- ElevenLabs is converging third-party image and video models with their own audio capabilities (voices, music, sound effects).
World Synthesis
- Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) introduces PAN World Model. A foundation model built for simulation, prediction, and agent reasoning.
Science
- A foundation model for atomistic materials chemistry.
- Google announces Google Scholar Labs.
- OpenAI: Early science acceleration experiments with GPT-5.
Robots
- Shenzhen MindOn Robotics is testing their robot brain in the Unitree G1 body. If the claims are true that this motion is not teleoperated, it is indeed remarkably fluid and capable.
- Agile Robotics (Germany) announces plans for Agile One humanoid for industrial context.