Daily AI Briefing

Open-weight AI is having a moment: Anthropic staked out its position on open-weights models, a $500 RL fine-tune of a 9B model reportedly beat frontier systems, and a new 8B model (Neutrino-1) plus firsthand reports say running open models "feels surprisingly good." Alongside that, devs are digging into how Claude Opus 5 benchmarks on coding tasks, watching Go's garbage collector move through the heap, and debating whether microservices and PyTorch's compiler still make sense as "authoring tools."