Recently, everyone has been hyping up AI Agents that can go on-chain and do work by themselves.

I find it pretty cool, but also a bit funny and frustrating: they can indeed help me monitor pools, compare quotes, automatically sign a bunch of steps, but when it comes to “should I sign this one” or “this contract looks like a skin swap scam,” I still need a human to cover the back... otherwise, if it gets too excited and treats my wallet as a resource library, I might as well turn into performance art on the spot.

And these days, with the main public chain undergoing upgrades/maintenance, everyone in the group is guessing whether projects will migrate.
I’m even less willing to fully rely on the Agent: if the chain shakes, RPC calls fail, cross-chain bridges show a bunch of red warnings, it might just fail very seriously. In the end, I still have to manually revoke permissions, stop the bot, change parameters, and maybe curse the floor price a couple of times.
Automation is fine, but I can’t let it take all the blame—I can’t afford to carry it all myself.