Practical Sharing - Zero Base Vibe Coding: I Hand Rubbed a US Stock Intelligence Station with Google Antigravity for 10 Minutes

Today, I would like to share a practical case that I just completed on a business trip: how to use Google Antigravity (as well as Gemini/ChatGPT) to develop a dedicated "US Stock Daily Intelligence Station" for myself without any knowledge of code. First, let's take a look at the finished product: see the picture below. Step 1: Clarify the requirements The first principle of Vibe Coding: Don't think about how to write code, think about how to use the product. As a follower of the US stock market, my pain points are very direct: Personalization: I only care about the stocks I hold (NVDA, TSLA, etc.) and don't want to see the noise of the entire market. Automation: I often travel around the world with chaotic time differences. I hope the webpage can automatically convert all times to 10am Beijing time (GMT+8) and push me the latest financial reports and news at that time. Minimalist maintenance: I don't have a technical background, it would be best if I could run it with just one file and support CSV import of stock lists. Step 2: Write a Prompt Similar to learning ancient methods of coding in college and starting with C language 101, the most important thing in programming 101 now is to learn how to write good prompts, so that AI can work better, understand and implement your needs. We recommend using Gemini or ChatGPT to assist in writing this Prompt, following these three golden principles: -What is my goal -How do I want to achieve (Tech Stack&Constraints) -What rules do you want AI to follow This is the core prompt I ultimately fed to Google Antigravity (recommended to bookmark): **Role: * * You are a senior Python development expert Agent. **Goal: * * Build a locally running US stock tracking web application. **Core Features:** 1. Watchlist: Supports manual input of stock codes (Ticker) and uploading CSV files for import. The data needs to be saved to a local file (for data persistence after restart). 2. News Aggregation: Use the 'yfinance', 'gnews', and' finviz 'libraries (free sources only) to capture the latest news. 3. Timezone processing: Automatically converts the timestamps of all news and data to GMT+8. 4. * * User Interface (UI): * * Provide a concise and clear dashboard. The core display is the "Daily Briefing" section, which includes percentage changes in stock prices and foldable news detail cards. **Constraints:** *Code robustness: The code must have high robustness. If a data source (such as FinViz) fails to capture or times out, the program must not crash. Simply record a warning in the background and skip the source. *Free principle: It is strictly prohibited to use any API that requires payment or complex key applications. *Zero foundation adaptation: I have no programming background at all, please ensure that the generated deployment plan strictly follows the standard of "one click run", the simpler the better. Step 3: Practical Programming in Google Antigravity Got the Prompt, open Google Antigravity（ https://antigravity.google/ ）. The reason why I use Google Antigravity here is that one fish eats two. Buying Google AI allows me to use Gemini3+Antigravity. Enter Manager View (Command Mode), create a new Task, and paste the Prompt above. Then all you need to do is go buy a cup of Bluebotel coffee. During your coffee break, the AI is continuously working for you, and the Planner Agent begins to plan the steps. Coder Agent starts writing Python code. Terminal Agent automatically installs dependency libraries such as streamlit and yfinance. The next step is to wait for Google Antigravity to prompt that the task is completed, and you need to verify whether your goal has been achieved: AI will automatically start the local server. Open a browser and enter localhost: 8501. I tried entering nvda, and the system immediately captured the latest FinViz exclusive news, and the time was converted to Beijing time according to the requirements. Additional requirement: At first, I only used yfinance as the source for the financial API. I wanted to add more news sources, so I told AI, 'I think there is too little news, help me add more similar free news APIs.'. AI automatically added Google News and FinViz as well. ” summary Through this practice, I have found that as long as the logic is clear, code ability is no longer the threshold for creating tools. Is this the end? Of course not. The current version, although usable, is not yet perfect. For example, the interface is not cool enough and must run on a computer. The most crucial principle when proposing requirements is the minimum feasible principle. When the first step is to tell AI the requirements, only mention the most critical core requirement and let it run the logic. After it runs, you can then raise additional requirements in the future. Otherwise, there will be too many requirements, and once errors occur, debugging will be very troublesome. In the next article, I will talk about how to increase further requirements and deploy them to the cloud so that everyone can access them openly.