how do you all decide which of your AI agents actually get access to real stuff?
ok so I have gone back and forth on this like three times now and still don't think I've actually landed on an answer. I have got a couple agents running for real, one drafts replies to stuff, one pokes through logs and flags weird things. thos…