Why I Started "Favoring" AI Agents


In the past, I judged tools by how easy or hard they were to use; now, I evaluate AI Agents by whether they can do the work themselves. What truly impresses me isn't how smart they are, but that they start to act like reliable colleagues: you give them a goal, and they break down the steps, gather information, execute tasks, and even report progress. Who wouldn't be excited about that? Of course, they also have their "internship moments," occasionally talking nonsense seriously, but the amazing part is— their rate of improvement far exceeds human experience accumulation. To put it humorously, humans learn through ten years of hard study, while AI Agents improve through ten updates. Many worry about them taking jobs, but I prefer to see them as an "external brain": instead of starting from zero when writing plans, it’s more like going from 60 to 90 points. In the future, the competition might not be about who works harder, but who can better command AI. Just like with navigation, experienced drivers aren’t just about knowing the roads—they’re about choosing the best route. The same applies to AI Agents: used well, they are strategic advisors; used poorly, they’re chatterboxes. My confidence in them is fundamentally about believing in the path of "human-AI collaboration." After all, humanity’s greatest advantage has never been quick calculations, but the ability to make choices. #我看好的AIAgent