Been trying out AI programming with Cursor these past few days. Took on two simple little tasks: one was a token data dashboard, and the other was monitoring price changes, trading volume, and OI anomalies for Binance contract trading pairs. Calling it Vibe Coding is pretty accurate—there’s definitely a vibe, but I barely wrote a single line of code myself. The logic for both tasks is relatively simple. The first one pulls data from Binance API and CoinMarketCap API, while the second one uses WebSocket to listen for data and sends a TG notification when conditions are met. But if I were to build everything from scratch—UI, concurrency, handling various data interfaces, etc.—it would still take quite a bit of time. With Cursor, all I had to do was describe my requirements in natural language, and it generated a basic, functional version of the code. From there, I just kept testing and refining it. Plus, you can have multiple agents working on different tasks simultaneously, which really boosts productivity. Of course, AI programming is far from perfect. If your requirements aren’t specific or clear enough, you often get results that are “vaguely correct.” Or sometimes, the AI confidently gives you what it claims is a perfect solution, but when you point out an issue, it’ll immediately respond with, “You’re right.” This made me realize that as AI tools become more powerful and widespread, the ability to “communicate accurately with AI and express your needs” is becoming more important than mastering some basic entry-level skills.