agents

Ollama switches to llama.cpp—faster local inference

SQL agents, Copilot billing changes, Node.js patches

Cut token use by 98%, Apple Silicon, and LLMs