Recently I’ve been watching some AI Agents automatically interact on-chain—it’s pretty interesting. Features like slippage control and depth judgment: machines really can do them faster than humans, but in places that need to be backed up by consensus or trust—for example, the weight allocation in social mining or the activity of fan tokens—no matter how smart the Agent is, it can’t learn “human情.” To put it plainly: if attention really could mine directly, then who would serve as the referee? As for my own order-placing strategy, I’m still relying on manually adjusting the depth layers; I’m not very comfortable with the slippage produced by the machine’s output. That’s it for now—continue watching the market.