Recently, everyone’s been hyping AI Agents that can run around on-chain by themselves, and honestly, I’m pretty uneasy about it… Automation is great, but if something really goes wrong, people still have to step in as backup. For example, during the authorization step (approve—granting the contract permission), I’d rather click it myself, slowly, than wake up the next day to find the permissions have been opened for “ages”; and with cross-chain and exchange-rate swaps, if slippage drifts or the routing changes, the Agent will just keep doing what it was told—humans are the ones responsible for calling a stop.

Also, risk control: on the perpetuals side, when funding rates and positions suddenly act up, the machine might still be “executing according to plan,” but people need to watch the thresholds. When emotions flare up, go cool off, come back, and cut positions—don’t stubbornly hold on. Lately, RWA and U.S. Treasury yields have been compared to on-chain yield products; as for me, what I care about more is this: who is actually guaranteeing the underlying, and whether they can exit if extreme conditions hit. The Agent can calculate, but whether to trust it—and whether to withdraw—still comes down to your own call.