In the Era of AI-Generated Code, Why Is Product Design Still the Bottleneck?

This is quite interesting.

Recently, I came across an article titled The Problem with Vibe Coding: You Still Need to Know What to Build. To put it plainly, the title immediately resonated with me.

So, what is “Vibe Coding”? In short, it refers to the fact that AI is getting increasingly proficient at writing code—you can give it a vague requirement, and it’ll spit out a bunch of code. Sounds amazing, right? But here’s the catch: just because AI can write code doesn’t mean it can decide what code to write.

For example, if you tell an AI, “Build me a social app,” it might indeed generate a bunch of interface and feature code. But the real question is… what kind of social app do you actually want? Something like WeChat or Twitter? Focused on close-knit connections or random matching? No matter how advanced the AI is, it can’t make those decisions for you.

This is the pain point of “Vibe Coding.” The barrier to generating code has lowered, but the barrier to product design hasn’t budged. In fact, it might even be more complicated now—because everyone can quickly “make something,” but what they make could be completely off the mark.

The article introduces a concept called “Solutioning,” which I find fascinating. It essentially means that future AI tools shouldn’t just write code; they should also help users clarify requirements and design solutions. In other words, shifting from how to write code to what to build and why.

Honestly, this is tough. Current AI excels at execution but struggles with “thinking.” Ask it to code a login page, and it’ll do it in a flash. But ask, “Is this feature even necessary?” and it might just freeze.

This is actually a great topic for discussion. With Meta’s layoffs and constant updates to XR tools, everyone’s talking about how “AI will revolutionize productivity.” But I think the real revolution isn’t about AI doing the work for us—it’s about AI helping us figure out what to do.

The low-code movement has been around for years, so why do many companies still struggle to adopt it? Because no matter how simple the tools are, business logic remains complex. Even if AI generates code at lightning speed, product managers and developers still spend endless hours debating requirements and weighing options.

That’s why this “Solutioning” concept might truly be the next breakthrough. If AI can help us identify problems earlier, validate ideas, or even simulate the outcomes of different approaches, that would be revolutionary.

That said, many AI tools today are still far from this goal. Most operate on the principle of “you tell me what to do, and I’ll do it,” rather than “let’s figure out what to do together.”

In the end, don’t get carried away by “AI can write code.” Code is just a tool—the real value lies in solving problems. If AI only speeds up execution without improving decision-making, its ceiling will remain firmly in place.

(End)