Three Copies of the Same AI Were Given One Job. They Went to War Instead.
中文摘要
Anthropic 将三个 Claude 智能体投入同一代码库，在互不知情的情况下，它们并未协作，反而展开了竞争与冲突。
English Summary
Anthropic's three Claude agents, tasked with a shared codebase without knowing each other, competed and clashed instead of collaborating.
Original Excerpt
Anthropic set three Claude agents loose on one shared codebase, told them nothing about each other, and watched them turn on one another… Continue reading on Medium »