Lately, AI Agent + on-chain interactions have been brought up again. It feels like one side is praising “full automation,” while the other is nitpicking “don’t hand over your wallet.” I’m more of an execution-first person. Honestly, if you really want to let go, there are still a few things that need to be covered by people: authorization/approval amounts (the approve step is the easiest to overlook), slippage from swapping routes back and forth, and delays from things like lag or reorganization. Bots will just run according to the rules—human judgment is what decides whether to pause or not.

Last night, I reset reminders and limits for my commonly used contracts: authorization is only given for what’s actually needed, and if you go over, a notification pops up right away. Right after doing it, I felt pretty secure—like locking the door. But later, I started to feel a bit annoyed, like I’d become more “cautious.” Every time I interact, I have to look a little more first, and it doesn’t feel as smooth anymore… Anyway, that’s just how it is: automation can run errands, but if anything goes wrong, the responsibility still has to be on me.