4
0Today, I'd like to discuss AI-assisted coding. Recently, I've built two projects using AI tools for coding, but before that, I experimented with what I'd call "vibe coding"—essentially, an intuitive, off-the-cuff approach. In my view, vibe coding doesn't work well for actual project development because the results are too unpredictable.
That's why I see AI coding as a fantastic executor, or perhaps more accurately, a reliable senior programmer you can count on. Based on my experience, the most effective strategy is to start by thoroughly architecting the project: outlining detailed requirements and documenting everything in Markdown files. I even leverage AI to help draft these architecture documents, guidelines, and design tokens in Markdown format. These files then serve as a clear roadmap, directing the AI throughout the implementation phase.
So far, this method has been incredibly smooth for me. While the AI handles tasks much like a human developer would, we still need to monitor its progress and fix any errors along the way. Ultimately, even with AI coding, a substantial amount of upfront preparation is essential before diving into a project
