Simon Willison on Muse Glimmer: "optimized for exactly the kind of things I'm looking for in a local model."

His same-day hands-on leads with the clean Apache 2.0 license and reports positive results on code exploration and image description, flagging the GGUF build immediately with llama.cpp, MLX, and ExecuTorch integrations following. His read carries weight because he's spent a year publicly hunting for a local model good enough to run agent loops on a laptop, and this is the closest he's said any open-weights release has come. Separately, Zuckerberg blamed US training-data rules for ceding open models to China, which drew 535 points and 488 comments on HN, one of the most contested threads of the day.