AI Knapsack RL: Unlocking Exploration of LLMs via Optimizing Budget Allocation https://arxiv.org/abs/2509.25849

Periodic Labs

AI Julian Schrittwieser, co-first author on AlphaGo, AlphaZero, and MuZero, summarizes studies of recent progress and what we should expect in the next…

CA-1 Europa

New Temples for the Sand God

Google DeepMind discovers new solutions to century-old problems in fluid dynamics

Google and OpenAI at the International Collegiate Programming Contest World Finals

Gauss, an agent for autoformalization

AI system creates expert-level scientific software

AI New research explains why LLMs hallucinate, through a connection between supervised and self-supervised learning.

AI A vision and path forward for genetically encoding almost all chemistry, powered by new AI tools…

AI Jet-Nemotron: Efficient Language Model with Post Neural Architecture Search https://arxiv.org/abs/2508.15884v1

AI Genie 3: An infinite world model with Shlomi Fruchter and Jack Parker-Holder https://www.youtube.com/watch?v=n5x6yXDj0uo

AI SSRL: Self-Search Reinforcement Learning https://arxiv.org/abs/2508.10874

AI A conversation with Demis Hassabis on world models (genie 3), deep think, the need for better evals (game arena), and their progress towards AGI.

AI Neuroscience study provides yet more evidence that AI systems and human brains converge on similar ways of representing the world…

GPT-5 The wait after GPT-4 was about four months shorter than the wait between GPT-3 and GPT-4.

Open models by OpenAI

AI GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning https://arxiv.org/abs/2507.19457

New Temples for the Sandgod

Hierarchical Reasoning Model

ChatGPT agent

Meta Superintelligence

Kimi K2: Open Agentic Intelligence

AI “Reachy Mini: cute and low priced, hackable yet easy to use, powered by open-source and the infinite community.

AI Energy-Based Transformers are Scalable Learners and Thinkers — “We outscale (feed-forward) transformers while generalizing reasoning/system 2…

Self-play improvement

AI Chai-2, a major breakthrough in molecular design.

AI Steering Your Diffusion Policy with Latent Space Reinforcement Learning — “If you have a policy that uses diffusion/flow (e.g.

Gemini Robotics On-Device brings AI to local robotic devices

San Mateo-based Generalist showcased its end-to-end neural network in action

AI The upcoming GPT-3 moment for RL: “We suspect the next AI paradigm will emerge from leveraging existing software to efficiently build training…

Neural OS