Everybody’s Wrong About "Vibe Coding"

“Vibe Coding” creates confusion, sparks division, and misses the revolution for the code! Let’s call it what it is: Prototyping and Building with AI tools. The vibes and the code are relatively irrelevant!

Everybody's wrong about "Vibe Coding" because the coding doesn't matter! What matters is the revolution in how we build and use software, driven by the new capabilities AI affords us.

The term “Vibe Coding” is anchored in the old world of engineering-driven product development, centres the conversation on the code and the developer, and emphasizes the powers and perils of generating and shipping AI-written code. That anchor holds us back from exploring the new possibility space we have entered into and is preventing us from talking about what’s really happening:

Today, anyone with access to generative AI software generation tools has the capability to build their own bespoke just-in-time throwaway software to meet their own needs and those of their community!

“Is there an app for that” is rapidly transforming to “I’ll build an app for that!” 

This shifts power away from the companies that build our digital products and services to the people who use them! After all, why pay for or use a product that doesn’t serve your needs if you can build your own?

That’s not just a new set of capabilities; that’s a new level of freedom!

“Vibe Coding” considered harmful?

When AI engineer Andrej Karpathy coined the term “Vibe Coding” in February 2025, he described chill voice-controlled coding sessions where the LLM does the work and the developer adopts a laissez-faire attitude towards errors:

“I'm building a project or webapp, but it's not really coding - I just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works.”

In the year since, colloquial use of the term has broadened the definition so much it's become more or less meaningless. Case in point: AI enthusiasts and AI-positive developers use the term as a broad umbrella to encompass any kind of AI-assisted coding, while AI sceptics use it in a more derogatory context, synonymous with irresponsible developers or inexperienced novices shipping bad AI-generated code riddled with errors and security issues to production causing catastrophic issues and eventually destroying everything.

In short, "Vibe Coding" is an unhelpful term and practically unuseful term referring to too many diverging practices and carrying with it too many diverging connotations, contexts, and meanings.

What we're talking about when we talk abut "Vibe Coding" is better described as Prototyping (using generative AI tools), Building (using generative AI tools), and Software Development (using generative AI tools).

More importantly, the term "Vibe Coding" is a distraction because the term puts the focus on the coding in a new context where the code and how it is built no longer matters! 

When anyone can build anything using generative AI tools, what matters is what the end-product does and how well it does it.

🔊 Please hold your righteous execrations until the end of the article.

You, Builder

AI tools lift the floor for product designers, builders, and anyone with an idea for a helpful piece of software, giving them new capabilities and the power to not just imagine but actually build software that works for them!

This radically shifts power from the people who write the code that makes software work to the people who use the software. The revolution in front of us is less about the code written than about how decisions about our digital ecosystem are made and who makes them.

Today, AI can spin up a functional prototype in minutes, meaning anyone with a product idea can build it, test it, and prove it in less time than it takes to write a PRD. With AI, code is cheap and the real value of software products and services is shifting from availability towards accessibility, usability, and individual utility.

Think about it: 

If you have an idea for a product or an app or a tool or just need to automate some tedious work, you can use AI coding tools to build a working prototype right now. Without coding skills! All you need is a clear idea of what you want your product to do and how to verify it actually does it. The technical implementation can be left to the AI.

Right now you have the capability to build software that works for you. You are the builder.

Have you lost your mind?!? Your objections preemptively answered

But wait! Won't the resulting code be full of errors? Possibly, but as long as it works for you, it probably doesn't matter.

What do you mean "it doesn't matter?!?" What about security? And privacy? And malicious code? That’s an infrastructure problem, and it is a solvable one. Meanwhile if you use an AI prototyping tool like Replit or Lovable, the necessary security and privacy layers are built in. That said, the big security and privacy issues raised in the “Vibe Coding” discourse center on enterprise software and commercial products. That line of reasoning is anchored in the old world. First, the revolution I’m seeing is people being able to build their own bespoke just-in-time throwaway software to meet their own needs. In most cases, these will be apps that run locally and have no major security or privacy risks. When those apps are hooked into external services, they can be built to shift security and privacy concerns to that external layer. Think custom app connecting to an external service through MCP with OAuth authentication and you get the idea. Bottom line, as long as the AI-generated app isn't bolted into enterprise systems or built in an environment that doesn’t prevent public passwords, keys, or other private information, it doesn't matter.

But it's code generated by AI, so it can't be trusted! You’re building a prototype, so it really does not matter. If you want to turn that prototype into a real product and ship it to other users, you are no longer building a prototype, you’re building a commercial product. This is where you bring in the necessary expertise to make sure it works properly, just like you would with any other commercial product.

But about people who Vibe Code at work and push AI slop to mission-critical enterprise software? That’s not an AI problem; that’s a culture and professionalism problem. Developers and engineers shipping untested code to production are not doing their job properly. Full stop. 

How Builders are changing the enterprise

It’s in exactly this last enterprise setting where the revolution is going to be felt the most: 

Until now, most software companies have been engineering-driven because in practical terms the coders had final say on what could and couldn't be built. A product designer could come up with the best idea in the world, but without engineering support and resources it would likely never make it past a Figma design. 

That past is now firmly in the past. Today, given access to the right AI tools, anyone can build, test, iterate, and prove the value of a new product idea without engineering support. And in companies that make the transition to a Builder-focused process that means every employee becomes the potential creator of the Next Big Thing! 

Empowered by the capabilities afforded by AI tools, subject matter experts can explore and iterate on software solutions that automate tedious or repetitive work, build better flows and processes, solve long-standing issues, and build new products at a fraction of the old cost in time, resources, and expertise. And by creating cross-functional builder teams that include knowledge workers, data and privacy experts, product designers, and engineering, the old bottlenecks can become a fond memory of a bygone era. 

This revolution in capabilities also brings new challenges for the enterprise, because like I said before, with AI everybody is a Builder, and when people realize they can just build their own software whenever they need it, the business of building and selling products and services is changed in revolutionary ways.

As we enter into this new reality, digital service companies have to choose a path in their effort to stay relevant and profitable. Here are three of many options:

1. Stay anchored to the old way of doing business by only providing the service through official products (and risk people going elsewhere)
2. Move with the flow by working with users to build products that are so good they don’t have reason to build their own or go elsewhere
3. Push ahead by providing the core service in a way that lets each user choose how they want to engage with it, either through official products, 3rd party products, or their own creations.

The future is not inevitable; it’s an opportunity space

Everybody's wrong about Vibe Coding because the vibes and the coding are relatively irrelevant. What matters is the revolution AI prototyping tools empower us by affording anyone the capability and freedom to build anything, for themselves and for the world. 

The soil of the opportunity space in front of us provides the conditions for human flourishing. AI provides the tools, and code the materials. What comes next is determined by what ideas we plant as seeds, and how well we tend them as they grow.

So, my friend: What future are we building together?