复刻 Lauren Tan 的 Grok Bot 方案：Grok 干活、Jev 编排、ChatGPT Dots 验证
作者把单 agent 拆成三层团队：Grok Bot 执行、Jev 负责任务路由与下一步决策、Dots 核查关键结论，加上人工审批边界，最终压缩成一份晨间简报，实现「睡觉时 agent 团队继续跑」。
LAUREN TAN’S GROK BOT SETUP RUNS AN AI TEAM WHILE YOU SLEEP I rebuilt the idea with Grok Bot + ChatGPT Dots + JEV instead of asking one agent to do everything, the system splits the work across different layers and keeps the workflow moving without constant human input -> Grok Bot does the actual work: research, execution and task completion -> JEV routes each task, coordinates the agents and decides what happens next -> ChatGPT Dots verifies the important claims and catches weak evidence -> approval boundaries stop anything that actually needs human judgment -> everything gets compressed into one morning brief the agents don’t all think about the same thing each one owns a narrow job, passes evidence to the next layer, and only pulls you in when something actually requires your decision while you sleep, the system keeps moving: intake -> route -> execute -> verify -> retry -> approve -> ship and when you wake up, you don’t have to read 50 agent messages you get one brief: what finished what changed what failed what needs approval what happens next the interesting part isn’t making one AI agent smarter it’s giving Grok Bot + JEV + ChatGPT Dots different responsibilities and making them operate like an actual team Grok Bot works JEV coordinates Dots verifies you review only what matters you sleep, the agent team works you wake up and review