Recently, I've seen everyone talking about AI Agents running automatically on the blockchain. I'm quite optimistic about it, but if I really let it run completely autonomously, I still feel a bit hesitant. To be honest, most of what can be automated is "execution," but whether it "should be executed" often requires human oversight: for example, not giving too much permission, reviewing the transaction details more carefully before signing, pausing if slippage suddenly spikes, and handling small pitfalls like failed contract interactions or duplicate submissions. Agents are diligent, but humans need to be responsible for braking.

When cross-chain bridges had issues back then, I became even more convinced of the "wait for confirmation" approach, especially when oracles go haywire reporting outrageous prices—it's better to wait and do less, it’s really not embarrassing. My roommate even teased me for acting like an old person, checking a transfer three times... Anyway, I prefer to be a bit slower and sleep more peacefully.