General
- Microsoft announce: Inside the world’s most powerful AI datacenter. They plan to build “the largest” AI cluster, with “hundreds of thousands” of GB200 GPUs.
- OpenAI & NVIDIA Announce Strategic Partnership to Deploy 10GW of NVIDIA Systems.
- OpenAI, Oracle, and SoftBank expand Stargate with five new AI data center sites. New data centers put Stargate ahead of schedule to secure full $500 billion, 10-gigawatt commitment by end of 2025.
- Sam Altman: Abundant Intelligence.
Research Insights
- Psychologically Enhanced AI Agents.
- Teaching LLMs to Plan: Logical Chain-of-Thought Instruction Tuning for Symbolic Planning.
- Internalizing Self-Consistency in Language Models: Multi-Agent Consensus Alignment.
Multi-modal
- Qwen3-Omni is a model that unifies text, image, audio, and video.
Audio
- Suno announces v5 music generation model.
Video
- Luma announces Ray3 (example, example for stop motion).
- ByteDance and Stanford: Mixture of Contexts for Long Video Generation. They claim this long-range memory retrieval achieves minute-scale coherence in video generation.
- Kling AI 2.5 Turbo Video Model. Capable of many challenging topics (examples: cyberpunk, human motion, motion and emotion, painting, gymnastics, breakdancing, combat, cinematic, horror camerawork, extended sequence).
Science
- Generative design of novel bacteriophages with genome language models. They claim the first AI-generated genome. They use DNA language models (Evo 1, Evo 2) to generate novel bacteriophage genomes.
- Google Deepmind: Discovering new solutions to century-old problems in fluid dynamics (preprint: Discovery of Unstable Singularities).
Hardware
Robots
- Humanoid HMND 01 Alpha, a non-mobile humanoid torso/arms intended for industrial applications.