A coding guide compares Underdog Saluki 27B, a 2-bit version of Qwen3.8-27B, with ISTA's GSQ-RCO quantisation. It looks at GGUF headers, memory use and tool-calling performance on the BFCL benchmark in standard llama.cpp.
Cognition has released the Devin Vulnerability Fixes plugin, three MIT-licensed agent skills that inventory, patch and test PHP, Node.js and self-managed GitLab deployments affected by this week's critical remote code execution flaws. The plugin then opens one pull request per repository.
Microsoft Research has released Agensh, a multi-agent harness with no central orchestrator in which self-organising workers share a Git repository, a chat and a findings board. On ProgramBench, with a six-hour budget, it scaled from one agent to 1,024.
The researchers behind OpenMobile-2 argue smartphone agents improve by training on many different apps rather than on more data. They built an open environment with over 110 apps, including 35 replicas of popular ones, and say their model scores highest among open agents on real Android devices.
Nvidia's open data engine Flora labels an hour of robot video for 72 cents, against $88 with GPT-6 Astra, with slightly higher quality scores. Built on three fine-tuned Qwen3.5 models, it is far better at spotting errors in robot actions but sometimes less accurate in its descriptions, and Nvidia sees it as part of a push to cover the whole robotics value chain.
A WeChat account affiliated with state broadcaster CCTV says DeepSeek and MiniMax are the first companies to complete registration under China's proposed AI export controls. Both have obtained rights to release models as open source and license them overseas.