Unverified50% confidenceFactExact time
Boris Cherny提出了AI代理的循环工作模式，通过/loop命令让代理进入持续的watch-act循环，而非执行完单条指令后退出
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026
Sources
Codex循环模式：AI代理从单次任务到持续循环的范式转变
rssDevelopers Digest5/5/2026
Related Claims
UnverifiedBoris Cherny has proposed a working model where AI agents run in continuous loops using the /loop command, rather than executing one-shot tasks83% similarVerifiedAgent Loop本质上是一个While循环，持续进行思考、行动、再思考、再行动，直到目标达成70% similarPartially VerifiedAgent Loop的工作流程包括观察输入、思考决策、执行行动、检查终止条件、循环或终止五个步骤69% similarUnverifiedAgent Loop的底层运行逻辑是一个「感知-规划-行动-观察」的迭代循环68% similarUnverifiedAgent Loop 的工程实现经历了三个阶段：裸 while 循环、LangChain AgentExecutor 封装、LangGraph 和 AutoGen 的图编排与多智能体框架67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/49397API
curl https://kongchang.com/api/v1/knowledge/claims/49397MCP
get_claim(id=49397)