AI Resonance · 阅读最新日报 · AI 入门推荐

2026-11-12 · America/Los_Angeles · 社区动态 · #13

What are you still not letting your agents do?

I’m building an open-source (MIT) workflow engine where the LLM proposes actions, but deterministic policy decides what actually runs. I’m looking for a real use case to test it against, not another demo where someone clicks Approve all day. What action have you kept manual or read-only? What kind of system(s) does it touch, and what’s missing from your current tooling? For example: replaying failed pipeline events for one customer without touching another customer’s data or running the same effect twice. If you’ve already found something that handles this well, I’d like to hear what’s…

What are you still not letting your agents do?

热度 44.9 / 100;排名与评分保留该期记录。