In the past, when learning to code, many people would start by learning syntax, frameworks, and APIs before beginning to build products.

Now, it’s increasingly reversed: first comes the idea, then AI helps you bring it to life.
My biggest recent realization is that vibecoding isn’t about people who can’t code winning, but about those who can clearly express their needs starting to get stronger.
In the past, an ordinary person with a product idea might get stuck because they didn’t know front-end, back-end, deployment, or debugging, and many ideas never even started because the learning curve was too steep and discouraged them.
But now, many things become simpler—if you can clearly explain the logic, AI can help you complete most of the work.
The development process is increasingly like chatting with a computer to make a product—think of something, then build it first, and improve it as you go.
The most extreme example is that some things I used to never dare to touch, but now I can actually launch them.
Many ideas used to die because of “not knowing how to do it,” but now they often die because “no one uses it.”
However, I think vibecoding will also bring problems.
In the future, there will definitely be many products that run well but have poor structure and are hard to maintain.
Because AI is good at generating code, but it doesn’t necessarily understand long-term system design.
So I don’t think software engineers will disappear; I believe the threshold is simply changing.
What will be more important in the future might not be how much syntax you know, but product intuition, system design skills, aesthetics, and whether you truly understand what you want to create.
Because what might be the most valuable in the future isn’t code itself, but knowing what’s worth building.