Recently, everyone has been talking about AI Agents automatically on-chain, which indeed saves effort, but I think some steps still need someone to take responsibility: especially authorization, particularly in cases of unlimited limits; also when cross-chain/routing, the "which route to take"—the Agent chooses quickly, but someone has to be responsible for the worst-case scenario; and then there are contract upgrades and parameter changes—problems often aren't due to insufficient computing power, but because no one looked at it more carefully.

In the past couple of days, there's been more discussion about RWA and comparing U.S. Treasury yields to on-chain yield products. I'm actually more concerned about whether the source of the yield is clearly explained, whether liquidation/redemption gets stuck, Agents can automatically click buttons, but can't explain "why it suddenly stopped working today."

If I could only keep one habit, it would be: pause for three seconds before each signature.