I am not a developer. Yet in the past months I have built our diagnostic tool, a survey engine, and a personal health app, by describing what I wanted to an AI, in plain language, and arguing with the result until it worked.
None of these will ever be massive product breakthroughs. By any commercial measure, they are exactly the kind of thing Michael Malewicz, a professional designer and coder, had in mind when he wrote “You only have weeks left to vibe code” — his argument that vibe coding is reaching the end of its hype cycle. AI lets people without technical skills turn ideas into apps, and most of those apps will never sell, scale, or even be used by the person who built them.
“If you haven’t yet vibed your awesome idea, I suggest you hurry. Or just give up already.” (Michael Malewicz)
That critique is fair. There is a lot of slop. A lot of abandoned prototypes. A lot of people mistaking “I made something” for “I created value.” And I am probably one of them.
But I think his article misses the more important story.
For people who never had the technical skills to build, vibe coding was not just a shortcut to making unsellable apps. It was a doorway into experimentation. Ideas that used to stay trapped in notebooks, workshop slides, and conversations could suddenly become something visible and testable, without hiring developers or designers.
We learn by exploring and failing
Most of my apps and tools will never become products. But judging them only by whether they sell misses the point. People learn by making. Without becoming expert developers, they start to understand how the pieces fit together — code, data, interfaces, deployment, and the prompts that steer it all. Not with mastery, but with growing practical literacy to see the system more clearly. And that is important, because it is hard to imagine the future of a technology you have never really worked with. Once you understand a little of how the pieces connect, you start seeing new possibilities, better questions, and more realistic ideas.
Subsidized AI use might be coming to an end
This matters even more because the window may already be closing. Access to powerful models has been unusually cheap, partly because the major AI companies have been subsidizing usage to attract users, developers, and startups. That phase is ending: pricing is probably shifting toward metered, token-based models where serious use will become very expensive. I have started rationing Claude Fable and saving it for the moments that matter. If that trend continues, early experimentation becomes more valuable, not less. Those who used this subsidized, slightly delusional first wave to explore gained their practical literacy at a discount, before the next phase becomes more expensive and controlled by the big US frontier labs.
None of this means every AI-generated thing deserves attention. When everyone can produce, creation itself becomes less scarce, and even good ideas may now be too many. But that is a story about publishing, not about learning, and it deserves its own essay.
Co-intelligence stays
So yes, vibe coding may have come full circle as a product fantasy. The idea that everyone can build and therefore everyone can win was always naive. If everyone becomes better at producing, we do not all become equally successful. We mostly create more noise, more competition, and more pressure on attention.
But as a learning experience, it moved people forward. Most people still use AI at assistant level — a search engine, a summarizer, a polite helper for work they already do. The amateur builders did something different: they made AI a partner in work they could never do before. They may not beat the professional builder and will probably never ship a successful product at all. But they are way ahead of the crowd when it comes to the race for success in a world dominated by intelligent machines.
They have left the house. They have seen the terrain.
The app may fail. The literacy stays.