Recently, many exchanges have started accepting AI agents, allowing users to place orders, check positions, and set stop losses using natural language. Sounds cool, but upon careful consideration, where is the difference between this and Vibe coding? Vibe coding is valuable because it replaces programming language with natural language. Programming itself has a threshold, which most people do not know. So when AI can help you turn ideas directly into code, the elimination of this threshold is real. But what does the AI agent of the exchange replace? It's an app operation. How much is the difference between asking AI to help you check your position and opening the app yourself to take a look? How much is the difference between asking AI to place a stop loss order for you and setting it at 3 on your own? The app was originally designed for ordinary people, and the threshold is already very low. Replacing an already easy operation with natural language saves very limited resources. It's like someone made a TV remote control that can be operated with their mouth. It sounds convenient, but that's just how it works when pressed by hand. And even if the operation becomes more convenient, it won't make you earn an extra dollar. Someone used VBIE to create a backend that looks like Bloomberg Terminal, making it look very stylish. But Bloomberg's strength lies not in its interface, but in the exclusive data and analytical models accumulated over decades behind it. No matter how well the interface is designed, without that data, it's just a pretty empty shell. Transactions need to have alpha, relying on information that others cannot obtain, or seeing things that others cannot see from the same information. K-line and trading volume are publicly available information, and technical analysis has been playing for decades. What should have been squeezed is already being squeezed out. AI can help you see the results of K-line analysis, which are not much different from those of mass-produced technical analysis teachers. The only truly meaningful scenario is when you already have your own trading logic and used to have to write your own programming string API to automate execution. Now, through AI agents, you can describe strategies in natural language and have them translate them into order instructions for you. This indeed lowers the technical threshold for quantitative trading. The problem is that even if the deployment strategy becomes easier, the quality of the strategy itself still depends on what data and risk control model you use. No matter how convenient the tool is, without an edge, there is no edge. If an exchange only achieves "letting users place orders with their mouths" on AI, it is simply doing something that is already not difficult in a different way.