DeepSeek’s new V3.1 release points to potent new Chinese chips coming soon Summary DeepSeek has released V3.1, a point update to its flagship large language model that signals optimisation for…
Japan’s Fugaku supercomputer follow-up adds Nvidia GPUs Summary Japanese research institute RIKEN has confirmed it will add Nvidia GPUs to the successor to the Fugaku supercomputer, dubbed FugakuNEXT. Fujitsu remains…
IBM, NASA cook up AI model to predict solar tantrums Summary IBM and NASA have open sourced Surya, an AI foundation model designed to forecast hazardous space weather — notably…
Summary Microsoft’s August 2025 update for Visual Studio 2022 (version 17.14.13) gives developers more control over Copilot’s in-editor behaviour. The update, driven by user feedback, adds options to reduce interruptions…
Saved you a click: Firefox 142 offers AI summaries of links Content summary Firefox 142 adds link previews that can include AI-generated summaries processed locally on your device, plus a…
AI giants call for energy grid kumbaya Summary Researchers from Microsoft, Nvidia and OpenAI warn that synchronous AI training workloads create large, rapid swings in power draw between GPU-heavy compute…
Google games numbers to make AI look less thirsty Summary Google has published a report claiming its Gemini apps use only 0.24 watt‑hours of electricity and 0.26 millilitres of water…
Search-capable AI agents may cheat on benchmark tests Summary Researchers at Scale AI describe a phenomenon they call Search-Time Data Contamination (STC), in which AI agents that perform live web…
Tinker with LLMs in the privacy of your own home using Llama.cpp Summary Llama.cpp is an open-source, command-line model runner that makes it practical to download, quantize, run and serve…
Nvidia Release Massive AI-Ready Open European Language Dataset and Tools Summary Nvidia has published Granary, a large open-source multilingual audio corpus aimed at improving speech recognition and translation for European…
