Recently, I saw a bunch of people hyping up AI Agents that can fully automate on-chain interactions. Honestly, I also want to relax, but when it comes down to actually running, it still requires human oversight. For example, when granting permissions, if the agent says "for efficiency" and gives you an unlimited approval, I’ll manually change it to just enough. Or with signature pop-ups, where the address and data are long strings, no matter how smart it is, it can’t take on the consequences of a mistake once. And for cross-chain/bridge transactions that get stuck for ten minutes, in the end, someone still has to watch over it—whether to wait, retry, or just stop altogether.

By the way, it reminds me of the recent NFT royalty debate—everyone wants automation, everyone wants better liquidity, but the idea of cutting off creators’ income entirely is also pretty tough to swallow… Anyway, my current approach is to start small, testing with 3 bucks first, and meanwhile, give myself a cup of tea to cool down and think.