==Vibe Coding Zero Foundation Tutorial== Through Vibe Coding, I realized the development of information flow product, released Mac App, built AI Agent to meet my needs, and more new tasks are being unlocked. Next, I will write a series of tutorials to help you easily develop products that meet your own needs with minimal steps and just the right knowledge. one ️⃣ How to choose models and tools In terms of programming tasks, these three models can be considered: OpenAI's Codex, Claude series Gemini 3 Pro。 If you can solve network and payment problems, I recommend Claude (more balanced and mature) as your first choice, followed by Codex. Starting with an entry-level subscription is sufficient, and it is not recommended to use the free version. In terms of tools, if you are accustomed to using IDEs, you can use VS Code or Cursor. If you prefer an intuitive graphical interface, cursor would be more suitable for beginners. However, in the initial stage, directly using the official applications or web versions of these three teams is completely sufficient. two ️⃣ Organize requirements Building complex products is not a problem at all, but it requires gradual progress. Even the most complex tools are composed of basic modules. If you don't know where to start, then start by detailing the requirements. Tell AI as much detail as possible about all the features and logic you have in mind, preferably using more advanced models such as GPT-5-Think, Gemini 3 Pro, or Sonnet 4.5/Opus 4.5. You can tell AI like this: > Hi AI， I want to develop a product that has the following requirements. Could you please help me organize them into a logically clear and structurally rigorous product requirements document. [Then wrap the line and write down all the functions you want to achieve one by one] AI can effectively transform vague and uncertain requirements into complete and professional documents. If you feel that there are any areas that do not meet expectations, just inform it in a timely manner for repair. Save this document well, it will become a good starting point for your product development. If you are unsure, you can also ask AI to help you preliminarily evaluate the complexity and implementation difficulty of the entire product. If it's really too complicated, it's better to adjust the direction earlier. three ️⃣ Disassembly task Hot knowledge: Do not chat with AI for a long time in the same conversation window, as its "intelligence" will gradually decrease as the conversation accumulates. Restart a new dialogue window and directly provide the previously organized product requirements to AI, allowing it to analyze the first feature that can be quickly implemented and easily verified. You can request this: > Hi AI， Here is my product requirements document, but I have never programmed it before. Please help me analyze what the first task I can quickly accomplish in 10 minutes is and guide me on how to quickly validate this feature. [Then wrap the line and attach the previous product requirements document] Let AI teach you step by step how to set up an environment, run code, and see results. There are two additional suggestions: 1. Most programming models are better at Python, and Python has a lower entry barrier, so it can be a priority choice. If the task requires networking, priority should be given to local tasks that are not networked, as networked tasks have higher complexity. If you don't understand, just ask AI directly. four ️⃣ Continuously iterate and clarify validation methods Creating a 'positive feedback mechanism' is the key to delivering the final product, avoiding completing overly complex tasks at once. After completing the first small task, continue to tell the AI what functions should be implemented next based on your needs and existing code, and clearly tell you how to verify each task. Ensure that the effectiveness of each developed feature can be quickly verified. five ️⃣ Fix bugs During the development process, it is inevitable to encounter bugs, such as results that differ from expectations. When encountering problems, you can also rely on AI to help you debug. Provide a detailed description of your actual problem, including error messages or running results, and AI will guide you on how to fix and optimize the code based on the situation. It may take several iterations to solve the problem smoothly. six ️⃣ Learn Fundamentals of Software Engineering As the complexity of tasks increases, it is necessary to gradually understand some basic knowledge of software engineering, otherwise many problems will be difficult to fix. It's like building furniture: assembling a simple table only requires intuition and instructions, but more complex cabinets require learning to use professional tools and understanding the basic structure. How to learn? Of course, through AI, one can gradually understand "data structures and algorithms", the basic composition of engineering teams, and the overall workflow. seven ️⃣ Development tools and GitHub After the task becomes more complex, tools at hand are also essential. You can choose VC Code or cursor to edit the code, or open Codex or Claude directly through the terminal. I personally prefer the latter, but I suggest you try more and find your own comfortable way. Code management is also crucial. When you frequently modify code, Git becomes indispensable. The simplest graphical tool is the official GitHub app, which allows AI to help you learn step by step how to commit and manage code. eight ️⃣ Server or Cloudflare Most modern products require networking to obtain data or continuously run certain tasks, such as capturing data, computing, storing data, etc. You can rent servers, such as Amazon Web Services or Digital Ocean, starting at a few dollars per month. But you also need to learn Linux system from scratch, and the environment setup is quite complex. I recommend Cloudflare Worker, the free version is a great way to get started, and the paid version only costs $5 per month. My own products are mostly built with Workers, which helps you reduce a lot of infrastructure complexity and focus on business logic. Cloudflare also has many useful tools, such as a 5GB free database with strong scalability. You can also have AI evaluate whether your product is suitable for you. nine ️⃣ Accumulate Vibe Coding experience I believe that building products through this Vibe Coding method will become increasingly popular in the future, and the threshold will also be lower. From now on, as you accumulate experience, you will find that you have grown together with LLM technology and will be able to easily handle more product development scenarios in the future. URL: https://(chatgpt.com) https://gemini. (google.com) https://claude.ai/