Recently, watching AI Agents run on-chain interactions has been quite lively, especially with the incentive and point expectations on the testnet. The group has started guessing whether the mainnet will issue tokens... But honestly, when it comes to spending money, it still depends on human oversight. I don't dare to fully delegate the authorization/signature step to it—unlimited limits, authorization to unfamiliar contracts, it's uncomfortable if I don't keep an eye on it in the middle of the night; then there's cross-chain/bridges—one failure can freeze assets, and I could blow up on the spot. Also, when encountering slippage, sudden gas spikes, or route hopping, the Agent might "calculate" quite accurately, but I trust myself more to hit pause. Anyway, I always reduce my position before sleep. Recently, I lowered my expectations, stopped fantasizing about "points = overnight riches," and as a result, my mindset has become much more relaxed—being able to fall asleep is already a win.