Recently, everyone has been hyping up AI Agents being fully automated on the blockchain. I’m both excited and a little scared… To put it plainly, for this to actually work in the real world, someone still has to be there as the backstop: if you don’t carefully review the authorization step, the Agent can go on to perform actions that open up your wallet permissions to “spend freely,” and when you realize it later, there’s nowhere to go cry about it. And then there are cross-chain/routing failures, sudden increases in slippage, and things like that—machines will only run according to the rules, while people need to be there to hit pause. Not to mention dealing with privacy coins and coin-mixing stuff—the communities end up tearing themselves apart arguing about it. Either way, I don’t want the Agent to “smoothly” step over compliance boundaries on its own.

Right now, my biggest skill is: waiting for confirmation, waiting for the callback/rollback, waiting until I’ve thought it through… When I get impulsive and want to add to my position, I first go make a strong cup of coffee, then wait two minutes before clicking confirm—it really works.