📄 中文摘要
在运行 AI 代理团队的过程中，最初只需管理一个代理，但随着代理数量的增加，协调问题逐渐显现。用户发现自己成为了瓶颈，因为所有的状态信息都掌握在自己手中。用户需要跟踪每个代理的工作进展，记住哪些代理因其他代理的输出而被阻塞，并识别出可能的冲突。这种协调问题在代理数量较少时并不明显，但随着代理数量的增加，用户的管理负担加重，影响整体系统的效率。
启动 AI 代理团队时没人警告你的协调问题
📄 English Summary
The coordination problem nobody warns you about when you start running AI agent teams
As the number of AI agents increases, a coordination problem emerges that often goes unnoticed until several agents are in operation. Initially, managing a single agent is straightforward, but as more agents are added, the user becomes the bottleneck. The user holds all the state information, tracking which agent is working on what, remembering dependencies, and identifying potential conflicts. This issue is subtle; while individual agents may perform well, the overall system's efficiency suffers due to the user's increased management burden.