Lately, everyone has been hyping up AI agents as if they can run on the blockchain and do arbitrage on their own, but I actually prefer to slow down… When it really comes to moving my lending positions, I still need someone to cover the risk. For example, during authorization, whether I give someone unlimited access, whether the contract address is correct—no matter how many reasons the agent gives, I can't resist my shaky hands making a mistake; or when near liquidation thresholds, whether to add collateral or reduce positions—these are choices of “prefer to earn less than to blow up,” and I have to bear the consequences myself.

The same goes for testnet points expectations or guesses about whether the mainnet will issue tokens—agents might be very active in interacting, but in the end, I bear the risk if the thresholds are breached… So my current approach is pretty simple: let it handle monitoring and alerts, but when it comes to signing, leveraging, or changing collateral, I still confirm a bit slower—just to be safe.