AI digest: agents everywhere, and cyber warnings getting serious
From Cursor going vertical to Five Eyes sounding the alarm, this week's AI news has some real substance to it.
A busy few days. Agents are eating more industries, Cursor is making a big move, and governments are getting genuinely nervous about what frontier AI means for cyber conflict.
Cursor announces its own model, a Git platform, and a mobile app
Cursor is no longer just a wrapper around other people’s models. The coding tool has revealed its first fully in-house trained AI model, alongside a new Git platform and a mobile app. That’s a serious vertical integration play, and it puts Cursor in much more direct competition with the big labs. Read more at The Decoder.
Five Eyes warns frontier AI could reshape offensive cyber ops within months
The Five Eyes intelligence alliance has said publicly that AI models capable of taking down governments and businesses are only months away. That’s not a vague long-term risk, that’s a near-term operational warning. The framing matters: this is about offensive capability, not just defensive gaps. Full story at The Decoder.
Anthropic’s Claude Tag learns your company through Slack
Claude Tag is a new Anthropic feature that sits in Slack as an always-on AI teammate, picking up institutional knowledge as it goes. The productivity angle is real, but the deeper play is context capture. Anthropic wants Claude to know your company well enough that switching to a competitor becomes genuinely painful. Read more at TechCrunch.
Sakana AI’s Fugu orchestrates multiple LLMs to match top benchmarks
Tokyo-based Sakana AI has built Fugu, a system that routes tasks across multiple models at runtime to match the performance of Anthropic’s best. The interesting thing here is not just the benchmark result, it’s the idea that orchestration might be a valid alternative to training a single giant model. Reduces provider lock-in too. Full story at The Decoder.