Recently, everyone has been talking about AI agents being able to run on-chain independently, but I’m a bit skeptical. Frankly, most of what they can do automatically is “click buttons according to rules,” but when the rules change, they fail: that one signature, which contract permissions are granted, whether the authorization is unlimited, which cross-chain bridge/router to use—when something goes wrong, someone still has to take responsibility and cover the losses. And governance is even more absurd; agents can help you read forums and make summaries, but when it’s time to cast a decisive vote, someone has to check whether “this proposal is just for show,” or else everything is just procedural correctness, and the outcome could be disastrous.

Lately, there’s been talk about interest rate cut expectations and the US dollar index moving wildly along with risk assets, and the on-chain scene is quite similar: when sentiment shifts, agents chase signals, and slippage and MEV come to collect taxes. My roommate also mocked me: aren’t you against dramatization? Why are you still scrutinizing each authorization for so long… I can’t help it, automation is fine, but responsibility is automatically dumped on me too. That’s all for now.