Recently, I've been seeing everyone hype up AI Agents running automated on-chain tasks. I'm quite tempted, after all, being able to click in bulk and save on gas (mainly on L2). But honestly, the parts that truly require human oversight are not few: the authorization/signature step must be carefully checked by oneself, or else you might give unlimited permissions and not know how you'll end up dead later; also, issues like cross-chain transfers, slippage during token swaps, and wrong routing—when Agents get excited, they can easily cause a failure, and in the end, someone still has to cancel orders, revoke permissions, and fix things.

Airdrop season is also quite surreal now. Task platforms are becoming increasingly strict against anti-witching tactics, and the points system has turned the grab-and-earn crowd into a bunch clocking in like at work... Agents can run tasks on your behalf, but whether they look "like a real person" is something they truly don't understand, which actually makes it easier to violate rules.

The information environment is full of noise. My noise reduction strategy is simple: only focus on the chains and contracts I need, treat other tweets as ads first, and revisit the next day if I still want to click.