Today I was looking at another pretty interesting log of an AI Agent running arbitrage. It automatically scans the price spreads across a few Dexes, and its speed is definitely much better than what I can do by hand, but when it hits new pools with very shallow liquidity, its slippage-range calculations are often mistaken by around the slippage amount—then it ends up eating up all the GAS fees at once. In situations like this, you still need a person to watch and set a protection line, or manually adjust the parameters; otherwise it just barrels in on its own and loses money. That said, the whole narrative around modular blockchain DA may excite developers, but from my layer’s users, it just makes on-chain data flows feel even more abstract. Even the Agent itself might not be able to judge which DA layer is truly reliable—ultimately, it still comes down to humans to step in and cover. Forget it, let’s leave it at that; at least today I didn’t let it run wild.