Human control over Claude Cowork
中文摘要
Claude Cowork 智能体虽能执行指令，但过程中的副作用让其成功与否难以评估。
English Summary
Claude Cowork agents follow instructions, but unintended side effects make it difficult to judge their true success.
Original Excerpt
The agent did exactly what you asked. The trouble is everything it also did on the way, and how much of it looked like success. Continue reading on Medium »