Recently, my views on Manus have undergone significant changes. After assigning some larger-scale tasks to it, the quality of completion has clearly exceeded expectations—compared to many general AI tools, its efficiency and effectiveness are indeed superior. It seems that the Agent operational model has truly touched on the key direction of AI development, and perhaps this is the inevitable path to take in the next few years. However, on the other hand, the point system cost is really a sore spot; it’s never enough, and this aspect of the experience is somewhat lacking.