In the past two days, I’ve seen a bunch of AI Agent demos where they automatically run around on-chain—sure, it’s convenient, but to be honest, I still feel like someone has to stand behind it. For example, with the signature: how big is the authorization scope, and did you accidentally click “unlimited allowance”… No matter how smart the machine is, it won’t feel bad for your wallet. And with cross-chain and switching pools like this—slippage, routing, even fake contract addresses—one time you get it wrong is basically the tuition. An Agent can at most say “Executed.”

Right now, I just have it do some really dumb jobs: check balances, reconcile accounts, remind me to confirm the transaction again 2 minutes later, and the amounts are only testing around 12 USDT. Anyway, don’t just start with full automation. As for those recent social mining and fan token “attention mining,” I find it a bit exhausting… Attention is indeed valuable, but what’s valuable isn’t necessarily you. I’d rather move slower, with less noise and more control.