A Founder Vibe-Coded His Way Into Swiss Jail (+5 Free Prompts to Build Smarter with AI)
What happens when you vibe-code your way into a police station?
That's exactly what happened to a founder named Sebastian at Davos this week. And this story is too insane not to share.
It's also a perfect lens into what AI-powered building looks like in 2025, the incredible possibilities, and the uncomfortable questions we haven't answered yet.
The Story
Sebastian built a physical hardware device, a wallet for smart contracts. Think of it as a secure, physical interface for blockchain transactions.
He built the entire software stack using Cursor, one of the leading AI coding assistants. Over the past year, he’s generated 25 billion tokens making him one of Cursor’s top users.
The prototype worked. He was taking it to Davos to pitch investors.
Here’s where it gets interesting.
The device was a black box with hot glue and wires sticking out. Now imagine what that looks like to a police officer especially one working security at the World Economic Forum.
Sebastian left the device on a table while he went to get a salmon roll. I’m not making this up.
Swiss police found it. They detained him. Put him in jail overnight.
And then came the question that changed everything.
The Question He Couldn't Answer
“Explain how this device works.”
Reasonable question. Sebastian built it. He should be able to explain it.
Except he couldn’t.
He’d vibe-coded the entire software stack. Cursor had decided to write it in Rust. Sebastian doesn’t know Rust. He couldn’t read his own codebase. He couldn’t explain how the architecture worked. He just knew it worked.
The police were skeptical. Understandably.
So they brought in an expert.
The Jail Cell Code Review
A technical architect named Chris was brought in to examine the device.
Chris sat down in the jail cell. He read through the codebase. He understood Rust. He could read the architecture.
And then he did something remarkable: he explained to Sebastian how his own product worked.
The verdict?
It works. It’s actually a pretty solid architecture.
Sebastian got his first real code review from an expert, in a Swiss jail cell, on code he didn’t write and couldn’t explain.
He was released. The device was legitimate. Just a really unfortunate combination of vibe coding and questionable hardware aesthetics.
What This Tells Us About 2025
This story is funny. But it’s also deeply significant.
We’ve crossed a threshold. You can now build serious, production-ready software in programming languages you don’t know, using architectures you didn’t design.
That’s brand new. That didn’t exist two years ago.
The implications are massive:
Building has been democratized. Non-engineers can ship real products. The barrier to entry for software creation has collapsed.
Speed has exploded. What took months now takes days. What took days now takes hours. 25 billion tokens in a year is a staggering amount of generated code.
The gap between building and understanding has widened. You can ship things you can’t explain. You can create systems you can’t debug. You can build products you can’t maintain.
That last point is the one we need to think about carefully.
The Hard Questions
Vibe coding raises questions we haven’t fully answered:
Debugging: What happens when vibe-coded products break? If you don’t understand the code, how do you fix it? Do you just vibe-code the fix too?
Accountability: Who’s responsible for code nobody understands? If the AI wrote it and you shipped it, who’s liable when it fails?
Maintenance: Software needs to be maintained, updated, and secured. How do you maintain code you can’t read?
Security: How do you audit code for vulnerabilities if you don’t understand what it’s doing? Sebastian’s code was solid but what if it wasn’t?
Scale: Vibe coding might work for prototypes. Does it work for production systems serving millions of users?
These aren’t hypothetical questions. They’re the questions every vibe-coding founder will face probably sooner than they expect.
The Opportunity
Despite the risks, I’m genuinely excited about what this enables.
We’re going to see a lot more Sebastians in 2026. Founders who can build faster than they can explain. People with ideas who no longer need engineering teams to create their first prototype.
The smartest ones will use vibe coding strategically: for prototypes, for MVPs, for testing ideas quickly. Then they’ll bring in people who understand the code before scaling.
The dangerous ones will ship vibe-coded products to production and hope nothing breaks.
I’ve put together 5 prompts to help you build smarter with AI whether you’re vibe coding or working alongside AI tools. They’re designed to help you maintain understanding even when AI is doing the heavy lifting.
Because the goal isn’t just to build. It’s to build things you can stand behind even in a Swiss jail cell.
The Bottom Line
Sebastian’s story has a happy ending. His code worked. His architecture was solid. He got validated by an expert in the most unexpected circumstances possible.
But not every vibe-coded product will pass that test.
The ability to build has been democratized. The ability to understand hasn’t caught up yet.
That gap is going to define the next wave of AI-powered entrepreneurship.
Build fast. But build smart. And maybe don’t bring black boxes with wires and hot glue to high-security zones.
What do you think? Have you vibe-coded anything? Would you ship something you couldn’t fully explain?
- Your Next Step
Let's Build Your Advantage
If you are ready to move beyond discussion and start implementing intelligent solutions that deliver a measurable impact, let's talk. I am selective about the projects I take on, focusing on partnerships where I can create significant, lasting value.
FAQs
Vibe coding is using AI tools like Cursor, GitHub Copilot, or Claude to generate code without deeply understanding the underlying language or architecture. You describe what you want, the AI writes the code, and you ship it often without being able to explain exactly how it works. It’s building by intent rather than by implementation knowledge.
As Sebastian’s story shows yes, it can produce working, well-architected code. But viability depends on context. For prototypes, MVPs, and testing ideas quickly, it’s incredibly powerful. For production systems that need maintenance, security audits, and long-term support, the lack of understanding becomes a serious liability.
The main risks are: inability to debug when things break, security vulnerabilities you can’t identify, maintenance challenges over time, accountability gaps when systems fail, and difficulty scaling or modifying the codebase. You’re essentially building on a foundation you don’t fully understand.
Use AI to explain the code it generates. Ask for architecture overviews. Request comments and documentation. Have the AI teach you what it built. Get code reviews from people who understand the language. Use vibe coding for speed, but invest in understanding before shipping to production.
Not entirely. It will democratize building and change who can create software. But understanding code being able to debug, secure, maintain, and scale systems will remain valuable. The engineers who thrive will be those who can work alongside AI, using it for speed while maintaining the judgment to know when human understanding is essential.