This is quite interesting.

Today, I came across an article titled The Limits of Vibe Coding: 30 Million Developers Can’t Capture the Spark of 8 Billion People, and honestly, it struck a chord.

Lately, hasn’t everyone been hyping up how “AI programming tools will enable anyone to code”? GitHub Copilot, ChatGPT writing code, low-code platforms… It’s as if tomorrow, the guy selling pancakes downstairs could casually whip up an app. But this article poured cold water on that idea—“lowering the barrier” doesn’t mean “mass programming.”

To put it bluntly, current Vibe Coding (or AI-assisted programming) is still for those 30 million developers. Ordinary people? Forget it.

Why? Because “knowing how to ask the right questions” is itself a barrier.

If you want AI to write code, you still have to tell it what you need, right? But just “describing requirements” already requires a basic understanding of technical concepts. For example, you say, “Help me build a website,” and the AI shoots back: “Should the frontend use React or Vue? Do you want RESTful or GraphQL for the backend?”—instantly deterring 99% of people.

And that’s not even mentioning the grunt work of debugging, optimizing, and deploying. What if the AI-generated code doesn’t run? How do you fix errors? These “tacit knowledge” hurdles are still there, unchanged.

So, current AI programming tools are more like “a power-up for programmers” than “a programming teacher for the masses.” They make professionals faster but don’t truly bring outsiders into the fold.

This reminds me of the age-old topic of “democratizing technology.” Every time a new tool emerges, someone declares, “Now everyone can do XXX!” But how did that turn out? Photoshop has been around for 30 years, yet not everyone became a designer. Video editing tools are everywhere, but most people still only film 15-second clips of their cats.

Technology can reduce operational costs, but the barrier to creation was never just about operations. Programming is especially telling—it requires abstract thinking, logical breakdowns, and system design. These skills don’t magically appear just because AI can write code.

Of course, I’m not saying AI programming is worthless. It has undeniably revolutionized development efficiency and may even reshape the software industry. But “mass programming”? We might have to wait a bit longer.

Or, think of it another way—do all 8 billion people on Earth really need to code? Maybe the best future is one where technology is transparent enough for ordinary people to “use” without having to “understand.” Just like how you use a smartphone today without knowing how the baseband chip works.

That said, if the day comes when the pancake vendor can actually build an app with AI… I’ll be the first to invest.