We put this list together because enough people have asked — seriously, not just casually — whether vibe coding is the real thing or a term that got traction for six months and faded. We think it is real. And so do the ten people below, which matters a lot more than what we think. These are not casual observers. They are researchers, founders, and builders who have each, in some way, put something on the line for this idea. Karpathy coined the term. Masad built the infrastructure before the term existed. Levels proved the economics work. Altman, Huang, Graham, Brockman, Rauch, Willison, and Andreessen each made it harder to ignore in their own way. Read through even a few of them and you will leave with a much clearer picture of why this is not going away.
Quick version, if you are new to the term: vibe coding is when you use an AI tool to write most or all of the code, and you direct it in plain language — describing what you want rather than writing the syntax yourself. You are the one with the idea. The AI is the one doing the typing. The name came from Karpathy, who described the process as coding by feel rather than by formal instruction. That is the whole concept. Now here are the people who made it matter.
1. Andrej Karpathy
This is where the story starts. Karpathy coined the term “vibe coding” in February 2025 — describing a way of working where you hand most of the code-writing to an AI model and steer by intention rather than syntax. Not a shortcut. A genuine shift in what the job actually is.
What gives him credibility is not just the coinage. It is what he does after. He builds in public — full applications, AI tools he actually uses day to day — and shares the process openly. He is not theorising from a distance. He is in it.
That is a different kind of voice than someone who endorses something from the outside.
2. Amjad Masad
Naming something and building the platform that makes it possible are two very different contributions. Masad did the second one.
The CEO of Replit has been working toward a world where anyone can write software for the better part of a decade. Vibe coding did not change his thesis — it confirmed it, more completely than he could have planned for. Under his leadership, Replit became one of the first platforms where AI generation was the point, not a feature you bolt on after the fact.
His public writing is consistently optimistic about what this shift means for people without traditional engineering backgrounds. And you can tell he means it, because Replit exists and it works.
3. Pieter Levels
If you have been wondering whether vibe coding can actually pay the bills, Levels is your answer. Known online as levelsio, he has built Nomad List, Remote OK, and PhotoAI — real products, real revenue, no engineering team — almost entirely with AI tools and a clear idea of what he wanted to build.
He documents everything: the prompts, the build sessions, the revenue numbers. All of it, publicly. That transparency is the whole argument in one person. This is not just possible — it is repeatable.
Masad makes the case at the platform level. Levels makes it at the level of a single person with an idea and no engineering degree.
4. Simon Willison
Not everyone on this list is cheering loudly. Willison — the creator of Datasette — is doing something more useful than cheering. He writes carefully, at length, about what vibe coding actually looks like up close: the failures alongside the wins, the limits of current models, the mental shifts required to work effectively with AI tools over time.
Now, you might assume that kind of rigour signals scepticism. It does not. Willison is not arguing against vibe coding. He is arguing for understanding it properly. When someone who documents the downsides still says this matters, you pay attention in a different way.
5. Sam Altman
The CEO of OpenAI has said, in various forms and across many contexts, that most software will eventually be written by AI — and that the skill worth developing is not syntax but direction. His company built the models that most of the vibe coding ecosystem actually runs on.
His influence is structural as much as rhetorical. When the head of the most important AI lab makes a consistent claim about where development is going, it shapes how companies hire, how universities teach, and how individuals decide what to invest their time in. That is a different kind of weight than an opinion piece.
Huang talks in civilisations. Altman talks in systems. The scale is different from Levels or Willison, but the direction points the same way.
6. Jensen Huang
Huang returns to a single idea, often: everyone is a programmer now. Natural language, in his framing, is the new programming language. You describe what you want. The model executes it. His hardware is what makes that execution fast enough to matter at scale.
He talks at a scope that is easy to dismiss as too large — not individual workflows but what kind of society this produces. But it is worth sitting with, because this is how the most powerful voice in AI infrastructure thinks about the future of building things. His keynotes have put that framing in front of investors, universities, and governments. It sticks.
7. Paul Graham
Graham came at the question from the angle he approaches most things: historically, philosophically, and a little contrarily. His argument is that programming has always been about specifying your intentions clearly enough for a machine to act on them. English, when the machine is capable enough, is just another way of doing that. The form changed. The act did not.
That reframe gives vibe coding a grounding that goes well past the hype cycle. It is not a trend. It is a continuity. Graham has been writing in that vein long enough, and with enough track record of being early, that people in the startup world take it seriously when he says so.
8. Guillermo Rauch
Rauch built v0 — a tool that turns a plain-language description into working React components. That is vibe coding as a commercial product, shipped by the company behind Next.js and a significant piece of the modern JavaScript infrastructure stack.
This is the transition from argument to product. Graham makes the philosophical case. Rauch ships the thing you can open in a browser tab and use today. He has been direct in interviews about where frontend development is going: generated, iterated, deployed, with far less handwritten code than the industry is used to. Vercel is the financial bet that he is right.
9. Greg Brockman
OpenAI’s co-founder and president is in an unusual position on this list. He could argue for vibe coding entirely from the institutional side — “our models make this possible” — and it would carry weight just from who he is. He does not do that.
Brockman writes and speaks as someone who builds things himself. He has shared tools he made for his own use and talked about what the experience of building with AI actually changes — specifically, the relationship between having an idea and having a working thing. That personal dimension is not nothing. It makes the argument harder to file away as company promotion.
10. Marc Andreessen
Andreessen has been arguing for a decade that software reaches and reshapes every industry, eventually. Vibe coding is, in his framework, the step that makes that reach complete: if anyone can build software, the future he has been describing becomes substantially more plausible, and a lot sooner.
His firm has invested across the AI coding stack — tools, infrastructure, platforms. His public writing frames all of it not as a productivity improvement but as a redistribution of who gets to build things and at what scale. You do not have to agree with every part of his worldview to find that framing worth understanding.
These ten are not a consensus. They disagree on pace, on risk, on what the right mental model is for all of this. What they agree on is that the direction is real. And the way you know they mean it is not the things they said — it is the things they built, funded, and kept arguing for when it would have been easier to hedge. None of them said vibe coding was interesting and left it there.