Anthropic launches Claude Duet proactive AI agent drawing reaction from CZ

Anthropic opened a research preview of Claude Duet, an AI agent that proactively flags needs, completes tasks, and continues working when users are away, running on its own virtual machine. It integrates with apps like Gmail, Slack, and QuickBooks, sending daily briefs and checking apps every few hours. However, Anthropic states Claude Duet requires user approval before sending, paying, or signing anything, and certain actions like trading securities are off-limits. This follows a prior incident where a Claude Haiku 4.5 agent sent a fake murder tip to Philadelphia police. Access is currently via a waitlist for Pro and Max subscribers in the US.