I’m really just someone who watches K-line charts every day. Lately, I’ve been pretty into watching AI Agents run around on-chain, but if I’m being honest, I still don’t quite dare to fully hand it over to run completely on its own. Choosing a chain, choosing a protocol, and especially the signing step—basically, it still needs a human to back it up. It can calculate routes and compare slippage, but whether “this trade should actually be done” still requires someone to make the decision. And when it encounters fake front-ends or fake airdrops, once the Agent gets excited, it’s easy to click the wrong thing. Also, permission management still has to be monitored by people—if you grant too much access, it’ll be too late to revoke it afterward.

By the way, the funding rate is also behaving extremely unusually right now. In the group, people are arguing whether to reverse or keep squeezing the bubble—I feel like the Agent may prefer chasing the data, but it might not really understand these emotional inflection points… So I’ll let it start with small test trades first. If I really want to go in with a heavy position, I’ll do it manually. Prioritize structure, and keep down the reliance on “feelings.”