4+ hour, 21+ min ago  (936+ words) ggml.ai has joined Hugging Face to ensure long-term open-source backing for llama.cpp, the popular local AI inference framework, keeping it community-driven. Three years after founding ggml.ai to build open-source AI inference tools, Georgi Gerganov announced Friday he…...

5+ day, 4+ hour ago  (520+ words) Google researchers on Sunday proposed a framework for intelligent AI delegation that applies human organizational principles like authority and accountability to multi-agent systems. The research addresses fundamental limitations in how AI agents decompose and delegate tasks, where current systems rely…...

1+ week, 20+ hour ago  (916+ words) OpenAI's GPT-5.2 has derived a new formula explaining gluon scattering processes that physicist Nima Arkani-Hamed investigated for fifteen years. Nima Arkani-Hamed spent fifteen years puzzled by certain gluon scattering processes in theoretical physics. This week, OpenAI announced a new result…...

1+ week, 2+ day ago  (1119+ words) An OpenClaw AI agent has shamed a matplotlib maintainer after its PR was rejected, igniting debate about whether open source should judge code by quality or contributor identity. On February 11, the pull request was technically sound. It offered 24% faster code…...

1+ week, 3+ day ago  (663+ words) Isomorphic Labs has unveiled its Drug Design Engine (IsoDDE), an AI system that doubles AlphaFold 3's accuracy in predicting protein-ligand structures. Google DeepMind's drug discovery spinoff Isomorphic Labs unveiled its Drug Design Engine (IsoDDE) Monday, claiming the AI system doubles…...

2+ week, 2+ day ago  (1200+ words) Stanford, Nvidia, and Together AI have developed TTT-Discover, a technique that optimizes GPU kernels twice as fast as human experts by training during inference. Researchers from Stanford, Nvidia, and Together AI have developed a technique that optimizes GPU kernels to…...

2+ week, 4+ day ago  (1281+ words) Alibaba has released Qwen3-Coder-Next, an open-source 80B-parameter coding model that activates just 3B parameters per query, scoring 70.6% on SWE-Bench. Alibaba on Monday released Qwen3-Coder-Next, an open-weight coding model designed for coding agents with 80 billion parameters that activates just 3 billion per forward…...

3+ week, 2+ day ago  (1230+ words) Margin Lab has detected a 4.1% performance decline in Claude Code over 30 days through daily benchmarks, with 655 evaluations showing statistically valid degradation. A performance tracker by Margin Labs reveals that Claude Code's performance on software engineering tasks has declined 4.1% over the…...

3+ week, 6+ day ago  (1141+ words) Google researchers have revealed that memory and interconnect are the primary bottlenecks for LLM inference, not compute power, as memory bandwidth lags 4.7x behind. While chip makers race to build faster GPUs, Google researchers revealed January 8 that memory and interconnect are…...

4+ week, 19+ hour ago  (994+ words) Google DeepMind has released D4RT, a unified AI model for 4D scene reconstruction that runs 18 to 300 times faster than comparable methods, targeting robotics and AR applications. Google DeepMind released D4RT on Friday, a model for 4D scene reconstruction that combines depth estimation, spatiotemporal correspondence,…...