Last night, I came across a bunch of demos of AI agents automatically on-chain, explained as if it's fully autonomous driving, but I still feel a bit uneasy… Honestly, there are too many places where human oversight is truly needed: who confirms the authorization? Has the contract address been changed? And narratives like “it says it will help me make money” are the most dangerous, feeling like it could easily turn into clearing out your wallet.

There are many things that can be automated on the blockchain, but the decision of “whether or not to do it” still has to be made by a person.

And looking at the old script of economic collapses in blockchain games—inflation + studio entry + coin price spiral—it's all starting to collapse from the point where the “mechanism seems to work.”
No matter how smart the agent is, it's hard to judge whether community sentiment is changing or if the narrative has already started to leak…
Anyway, my current approach is: automation is fine, but with small limits, short permissions, and every action requiring my approval, otherwise I don’t sleep well.