Vibe Coding Is Here: A Journalist Built a Working App Without Writing a Single Line of Code

A Journalist Built a Working App With an AI Prompt — And It Actually Worked

When The Verge's Allison Johnson came home to a dying yard, she did not call a landscaper or download an existing garden app. She opened Google Gemini, typed a lengthy prompt, walked away for five minutes, and came back to a functioning application.

That is not a typo. Five minutes. A working app. No developer. No code.

Writing for The Verge on June 13, 2026, Johnson documented what is being called "vibe coding" — a term for the emerging practice of describing what you want an AI model to build and letting it handle the actual programming. Her experience is worth paying attention to, not because a yard-tracking app is going to change the world, but because of what it signals for how software gets made and who gets to make it.

What Actually Happened

Johnson gave Gemini a detailed natural language prompt describing the app she wanted. When she returned, two things were waiting: a functional app running in a preview window, and an error message. The error read something like "Channel is unrecoverably broken and will be disposed" — which sounds alarming. But directly below the error was a button to fix the bug automatically.

She pressed it. The bug was fixed.

The process is not perfect. Johnson's account makes clear that vibe coding still involves iteration, occasional confusion, and moments where the AI produces something unexpected. But the core premise held: a person with no programming background described a software tool, and an AI built it.

That is a fundamentally different relationship between ideas and execution than anything that existed two years ago.

Why Business Teams Should Pay Close Attention

For most small and mid-sized businesses, the bottleneck has never been ideas. Teams know what tools they need. They know which internal processes are broken, which reports take too long to pull, which workflows require five manual steps that could theoretically be automated. The bottleneck has always been access — access to developer time, to technical resources, to budget for custom software.

Vibe coding chips away at that bottleneck in a real and immediate way.

A marketing manager who needs a custom tracking dashboard no longer has to wait in a development queue. An operations lead who wants a lightweight internal tool to manage vendor contacts does not need to justify a software budget to finance. A small business owner who has been running a process in a spreadsheet for three years can now describe what they actually want and get something functional in return.

This does not mean vibe-coded apps are production-ready enterprise software. They are not. Johnson's garden app is a personal tool, and the gap between a personal tool and a business-grade system is real. Security, scalability, integrations, compliance — none of that disappears because a prompt was well-written.

But for internal tools, prototypes, proof-of-concept builds, and lightweight automations? The calculus has shifted. Teams that understand AI tools for business are already experimenting with these workflows, and the early results are hard to dismiss.

The Deeper Shift: Execution Is No Longer the Hard Part

What Johnson's experiment illustrates is that the hard part of building software is increasingly the thinking, not the typing. Defining what you actually need, identifying the right inputs and outputs, understanding what success looks like — those are human problems that require human judgment. The AI handles the translation from idea to code.

This has implications for how businesses hire, how they structure projects, and how they think about automation for small business. The person who can write a clear, specific, well-reasoned prompt is now genuinely capable of shipping software. That is a new kind of skill, and organizations that recognize it early will move faster than those that do not.

It also raises honest questions about quality control. Bugs will happen. Johnson's Gemini session produced one in the first five minutes. The difference is that the fix was one click away. For now, that is good enough for many use cases. Whether it stays good enough as the stakes increase is a question worth tracking.

What This Means Right Now

The story Allison Johnson wrote for The Verge is a first-person account of a personal project. But embedded in it is a preview of how business software gets built in the next few years. Faster. Cheaper. By people who were never called developers.

For teams looking to move quickly without waiting on technical resources, platforms like WRRK.ai are worth exploring as a starting point for understanding what AI-assisted work looks like in practice.

The yard may still be dying. But the app works.

Original reporting by Allison Johnson, published at The Verge on June 13, 2026.

---

Start building smarter with AI at WRRK.ai.

Frequently Asked Questions

What is vibe coding and how does it work?

Vibe coding refers to the practice of describing a software application in plain language — without writing any code yourself — and letting an AI model generate the actual program based on your description. Tools like Google Gemini can interpret natural language prompts and produce working applications, including handling basic bug fixes automatically.

Can non-technical business owners really build apps with AI?

Yes, in limited but meaningful ways. Current AI coding tools are well-suited for internal tools, personal productivity apps, and lightweight automations. They are not yet a replacement for professional software development when security, scalability, or complex integrations are required. But for everyday business problems, the barrier to entry has dropped significantly.

Is vibe-coded software reliable enough for business use?

It depends on the use case. For internal tools, prototypes, and low-stakes automations, vibe-coded apps can be reliable enough to deliver real value. For customer-facing products or systems handling sensitive data, additional review, testing, and professional oversight remain important. The technology is improving rapidly, and the threshold for what qualifies as "reliable enough" is moving with it.