[Updated 2025-Aug]
My journey with Vibe coding
With the rise of the term AI Code, or “vibe coding” recently, I’ve been reminded of my high school days (the 2000s), when I built a forum using phpBB without really knowing how to code (don’t get me wrong, vibe coding isn’t about coding without any knowledge, nor is that something I prefer to do)
Back then, I had no experience in coding. I just read what I could find, copy/pasted, mixed and matched, failed, retried, and eventually ended up with a functional forum. I even added some of the trending plugins of that time like announcements, a chat box, a leaderboard,.. (yeah the era of Yahoo chat and blog 360, and forums)
I customized themes/plugin without truly understanding HTML, JavaScript, or PHP. Still, that experience gave me my very first impression into what programming is, how a web development work, and the fundamentals of hosting a website.
Of course, when bugs appeared, I often didn’t know why. I could identify issues at a high level but lacked the skills to fully understand and resolve them without relying heavily on search engines and technical forums.
Looking back, this feels a lot like vibe coding today at but in a hard way.
Back then, I used to search for solutions and plug them into my system without worrying about how things worked behind the scenes. Today, I have an AI assistant that can understand my prompts and return a solution instantly, without me having to jump around searching for answers. Even more impressively, it can apply those solutions directly to my codebase, without me lifting a finger.
The nature of evolution
Given the story, I believe that Vibe coding (or maybe something else in the future) is great for getting things done quickly, and it’s an amazing way to try something new.
Whether we name it “vibe” or something else, I think this is simply the natural direction of software development where always want to build applications at higher levels of abstraction to reduce the development cost. Just as we no longer hand-code everything in assembly or C++, and there are also many systems is working very well with the low-code approach.
Vibe coding represents another level of abstraction, fueled by decades of community knowledge and experimentation and delivered to us via LLM model and Coding assistant.
The Good and Bad
Right now, I see vibe coding as especially powerful in greenfield projects and especially useful for building the starter structure, prototyping, proof of concepts, or learning a new framework, etc,.. where we won’t have to worry much about the legacy or existing stuff.
In brownfield projects, though, the challenge will be providing enough instruction and context for assistants to truly understand an existing codebase and the business logic behind it. And the cost (financial) is also a challenge.
Overall, I think vibe coding is a positive signal where it pushes the industry to improve its development pipelines, requirements engineering, testing, and collaboration workflows through natural language, …
But there’s also a danger. It can be like having a helpful baby sister who does almost everything for us and just work (but not sure It’s the most suitable): convenient, but it risks making you dependent, less creative, and less thoughtful about the “why” behind the code. And when something goes wrong and the AI fails you, you may not have the skills to handle it yourself.
Final though
At this stage, where both the community and AI-powered search are growing at an incredible pace, I believe coding assistants and their acceptance in the developer community are still far from reaching their final form.
For now, Coding with AI should be seen as a productivity tool – something to automate tasks you already proficiency with, or as an advisor that helps you refine/discuss your work & exploring more ideas.
If you’re just starting out with the coding journey, don’t let it write all your code. You won’t actually learn anything that way. Treat the AI bot as a companion learning alongside you, not as a delegated assistant or mentor. You should write the code yourself, then use the AI to review your work, challenge your thinking, challenge back the AI bot as its answers won’t always be right.
AI is just a tool, it follows the crowd and doesn’t truly create new knowledge or thinking. That’s why, whenever possible, it’s important to expand your network. Having a real human mentor or peer is always a huge advantage for both your learning journey and your professional development.
While today’s coding assistants still have limitations and don’t always provide the right help, they’ve already proven themselves to be valuable and effective in many cases. Ongoing improvements from both established tech giants and innovative startups are steadily pushing them forward.
Unlike the short-lived AI hype cycles of the past, I believe today’s coding assistants represent a practical and lasting advancement.
Hung Doan 
Leave a comment