Recently I’ve seen people say AI Agents can automatically run on-chain interactions, which sounds pretty cool. But honestly, if you really want to let it fully automate integration with those complex restaking protocols, I’m still a bit worried. After all, with this nested “restaking” structure, rewards stack on top of rewards—but risks stack just as deeply too. Can an AI understand the nested logic of every contract? I can’t say for sure about other things for now, but when it comes to slippage protection and routing selection, I’ll definitely have to keep an eye on it myself. What if the Agent gets carried away, chooses a route with high slippage, or gets hit by a sandwich attack—then you wouldn’t even have time to cry.

Forget it—plainly put: no matter how smart the AI is, it may not fully understand all the traps on-chain. I’d rather spend a bit more time myself, manually set the slippage, or choose a private transaction pool. In any case, for the parts that need to be covered as a fallback, it still has to be done by people themselves.