We’ve all heard the legend of the 10X developer. The engineer whose output and impact seem to exceed the work of ten others. It’s one of the great myths of software, and, like most good myths, it’s rooted in some truth.
But what if the next generation of great software doesn’t start with the best engineer at the keyboard? What if the first meaningful version comes from the person who understands the user problem so well that they can now prototype the solution themselves, with a little help from AI?
I’m not here to throw shade at developers. In fact, I have described at least one dev I worked with as the Mozart of software. Software engineers are amazing. This isn’t about titles, it’s about removing obstacles to building software. I believe we’re entering a moment where the first working version of the next great product may come from a product manager, founder, or technologist who knows the user and the problem deeply, and can now vibe code their way to a real, dirty, somewhat usable prototype.
What Is Vibe Coding?
The term vibe coding was popularized by Andrej Karpathy, a well-known AI researcher, who defined it like this:
“Vibe coding refers to a programming technique where a person describes a problem or desired outcome in natural language—often as a prompt—to an advanced AI tool or large language model (LLM) tuned for coding. The AI then generates the code, shifting the programmer’s role from manual coding to guiding, testing, and refining the AI-generated output.”
– Andrej Karpathy
Said simply: vibe coding is when you build software by describing what you want, not by writing every line of code yourself. The builder’s job shifts from typing to prompting, shaping, and testing. So, to do this well, we can imagine that the person writing the prompt needs to be able to deeply describe the problem to be solved, the desired user experience, and key outcomes. Hmm, this sounds like Product Management!
A Quick Look Back: Why Product Management Exists
If you’ve spent time around good product teams, you’ve probably heard sayings like, “Life is too short to build software that doesn’t solve real problems.” That quote comes from my longtime colleague and friend, Mike Bousquet, and it sits at the heart of why product management exists in the first place.
The discipline of product management grew up around frameworks like Pragmatic Marketing and Design Thinking, which emphasized deeply understanding user problems before jumping into solutions, and the Silicon Valley Product Group (SVPG) approach, which encourages empowered product teams, not feature factories.
In Design Thinking, the classic Venn diagram asks:
- Desirable: Do users want it?
- Viable: Will the business thrive?
- Feasible: Can we build it?
Traditionally, product managers have been strongest on desirable and viable, often relying on engineers to lead on feasible. But what happens when PMs (and founders, and curious technologists) get access to tools that let them build working software from an idea, even at the earliest prototype stage?
The AI-Powered Software Shift
We’re seeing an explosion of tools that enable this:
- Cursor and Replit (AI-powered IDEs)
- Bolt and Lovable (new ways to build with LLM support)
- Copilots everywhere, Claude, and ChatGPT
These tools don’t just autocomplete code. They let you describe the problem, set constraints, ask for alternatives, and iterate. They lower the barrier between ideas and working software. Or somewhat working software.
It’s not no-code. It’s not low-code. It’s something different: AI-assisted coding that speaks your language.
My Own Experience Vibe Coding
I’ve felt this shift firsthand.
Earlier this year, I tinkered with building a simple web app to sell my family’s extra MLB tickets for games we couldn’t attend (three kids, multiple sports teams, busy life). I’m not a developer by trade. I used ChatGPT to help me build the app. Was the experience perfect? No. But I got something working. Something real, with code and a UI that I could test and show to others. And I learned a lot along the way (StubHub and FB Marketplace for-the-win).
More recently, I used Cursor and Claude to prototype a Model Context Server (MCP) for my company Query, front-ending our normalized API gateway. This was a much more technical project, involving real architectural choices. No experienced developer would look at the code and say it was production-ready. But it worked. And even more importantly, it helped me learn faster, by testing ideas, seeing how they played out in code, and clarifying what was needed to deliver a complete solution to our users.
These are small examples. But they reflect a much bigger shift.
Why PMs (and Founders) Might Be the Right People to Build V1
The first version of software is rarely about scale. It’s about understanding. It’s about validating that you’re on the right track before you spend the time and money to make it production-ready.
This is the heart of Ash Muraya’s Running Lean and Lean Startup thinking: get the smallest, fastest version of your idea into the hands of users. Learn early. Adjust quickly.
Vibe coding, when used by people who know the user and the problem deeply, supercharges this process. It’s prototyping on steroids.
But Don’t Forget What Can Go Wrong
Of course, shipping software isn’t just about writing code that works once. It’s about building code that scales, that’s secure, that handles edge cases, that doesn’t fall apart when the next feature gets added.
AI-powered coding tools don’t always know best practices. They can hallucinate packages that don’t exist. In fact, one recent study found that 19.7% of recommended packages from code-generating LLMs weren’t real (“We Have a Package for You! A Comprehensive Analysis of Package Hallucinations by Code Generating LLMs”).
Vibe coding doesn’t replace engineering. It doesn’t replace architectural decisions. It doesn’t replace hard-won experience about scale, performance, security, and reliability.
But it can replace some of the slow, early loops where the idea itself is still unproven. It can replace the awful experience of teams of amazing people building something that doesn’t solve a real problem.
Future Vibes
The best product managers of the future will build software with AI. But they won’t do it alone. They’ll work alongside great engineers. They’ll collaborate with designers, architects, and security experts. In fact, the people of the future may turn out to be a blend of all these things. If we can outsource the tedium and the toil to our new AI friends, just imagine what we might be able to accomplish.
The promise of AI is not that it replaces skill—it’s that it can bring product sense and engineering capability closer together. Sometimes in a single person. Often in smaller, faster-moving teams.
This won’t be true for every project. And it certainly won’t make great engineering obsolete.
But it does open the door for something powerful.
Fewer handoffs. Faster iterations. Teams that can sketch in working software, test with users, and zero in on the right problems to solve, before they scale.
Don’t worry, they don’t let me write production code at Query. We have an amazing team, pushing the boundaries with new AI powered features in our product and enabling our customers AI use cases with Query as the data gateway, with an approach we call Federated Security.
It’s still early days. But the shape of what’s coming is getting clearer. And I believe it’s going to change how great software gets built. In the meantime, please don’t make fun of my horrible vibe code. I learned something. And that’s all that matters.
