Good morning. Here is what mattered in AI on Sunday, November 8.
The top story is a safety audit. The nonprofit METR reports that Anthropic’s models have improved about twice as fast since October 2025, and credits most of the speed-up to AI used inside Anthropic’s own research, though Anthropic redacted much of the evidence. In policy, a New York Times investigation raised questions about how Nvidia H200 buyers in China are vetted, and Meta launched Spaces for its Muse agent. We also updated our report on OpenAI’s funding round.
Below that, our explainers: how AI translation works, a checklist for vetting AI study tools, and a myths-and-facts piece on AI detectors and watermarks.