Dots 执行 + Jev 决策：搭一个准自主 AI 员工

Zero @zer0point_eth原推发布于 2026.10.02

在 X 上打开原推

OpenAI Dots + Jev might be the closest thing to an ai employee i've seen Dots gives you the persistent environment Jev decides what should happen next And together you get something much closer to an autonomous ai worker The loop is simple: your goal → dots → jev decides → dots executes → jev validates → repeat Here’s how i’d use it: step 1 → give dots a persistent workspace with browser, terminal, files and memory step 2 → connect jev as the decision layer before expensive or risky actions step 3 → let jev choose between tools, skills, models or recovery paths instead of hardcoding every workflow step 4 → give dots the boring execution work: research, browsing, coding, testing, docs and repetitive tasks step 5 → let jev handle routing. if a task is simple, use the cheap path. if it needs deeper reasoning, escalate it step 6 → add checkpoints before anything irreversible. sending messages, moving money or changing production should still require approval step 7 → run the whole thing on a schedule. dots keeps working while you’re offline step 8 → keep logs of every decision. over time, you can see which routes actually save time and money step 9 → give the system shared memory so it stops rebuilding context from scratch every time step 10 → close the loop. jev decides what matters, dots does the work, and the result becomes new context for the next decision The interesting part isn't that dots can do more things You give it an objective and let the system figure out how to get there That's a completely different way of thinking about agents

这条是引用着另一条写的：x.com/zer0point_eth/status/2105324196838457425