One More Prompt: How Vibe Coding’s Casino Mechanics Cost Me a Billion Dollars

It was a beautiful summer day on my vacation. My family was waiting for me to go to the beach. I’d told them I’d join in ‘just five minutes’ – that was two hours ago. “Almost there,” I muttered. “Just one more try.” The code was almost right (for the last two hours). That’s when it hit me: I wasn’t debugging. I was gambling.

The Slot Machine in Your IDE

If you’ve spent any time “vibe coding“—that beautiful, chaotic dance of prompting AI until it spits out working code—you’ve felt it. That pull. That voice saying, “just one more prompt, baby.” When AI limits are actually coming to the rescue, letting you get your well-deserved sleep.

This isn’t accidental. Vibe coding has the exact psychological structure that makes slot machines, video games, and doomscrolling so addictive. Let me break down the mechanics:

Variable Reward Schedule: Sometimes the AI nails it on the first try. Sometimes it takes fifteen prompts. Sometimes it gives you something brilliant you didn’t even ask for. You never quite know what you’re going to get, and that unpredictability—psychologists call it a variable ratio reinforcement schedule—is the most addictive reward pattern we know. It’s the same principle behind slot machines and loot boxes.

The Near-Miss Effect: The code is almost there. It runs but has a bug. The logic is right but the syntax is off. These near-misses trigger the same neural pathways as almost winning at slots. Your brain interprets “almost there” as “so close that one more try will definitely work.” 

Why Professional Engineers Are (Partially) Immune

Here’s something fascinating: professional software engineers are actually less susceptible to vibe coding addiction than non-coders. At first, this seems backwards. Shouldn’t the people who understand code best be most excited by AI that writes code?

But think about what’s actually happening. For a professional engineer, the variable reward is: “Will the AI write code as well as I would have, or will it fail?” That’s not particularly thrilling. They know they can write the code themselves. The AI is a productivity tool, not magic.

For someone who can’t code? Holy shit. Every time the AI works, it’s a miracle. You go from zero ability to create software to suddenly building things. That’s not optimization—that’s transformation. That’s the rush I felt when I first started my engineering career, when I realized I could write some code and make characters move on a screen, when software felt like actual sorcery.

And right now, that moment of creation is being democratized to millions of people. The dopamine hit isn’t just about the variable reward—it’s about the infinite jump in capability. From “I can’t do this” to “I just built something.” That’s intoxicating.

Not only that, but most professional engineers have a career-long training of long debugging sessions, so it’s not a new drill for them.

The 100-Hour Week Isn’t What You Think

There has been a lot of discourse about founders now working insane hours—sleeping in WeWork offices, 100-hour weeks, and complete burnout. The narrative is usually about pressure from the AI boom, competition, and fear of being left behind.

But here’s the thing: founder pressure has always existed. That’s why most startups fail. It was true before AI, it’ll be true after. The startup journey has always been a roller coaster of existential terror and irrational optimism. The pressure, the competition, the feeling that the world is either against you or rooting for you—that’s not new.

But the vibe coding brings this to a whole new level of dopamine. When YC Combinator says that 90-something percent of their portfolio companies are vibe coding their products, suddenly those 100-hour weeks make sense. These founders aren’t just working hard—they’re hooked. They’re sitting there at 3 AM telling themselves, “one more prompt,” the same way I was. 

It’s a fundamentally different dynamic than the SaaS boom a few years ago. The tempo of methodically building your prototypes in the pre-AI era didn’t create the same variable reward dopamine hit.

The Conventional Wisdom I Followed (and Shouldn’t Have)

When I started Zencoder, I did the most logical thing – focused on professional engineers. Professional engineers have enormous workloads. When I was running a 1,200-person company, most of my ideas never shipped. The bigger the company got, the smaller the percentage of my ideas we could execute, because there were so many other priorities—security, test coverage (we had 30,000 tests), accessibility, analytics, and technical debt. I thought AI was an opportunity to unlock creativity at scale, helping companies ship faster. So I built for professional engineers and their complex repositories. 

Every business book tells you to define your ICP (ideal customer profile) and focus on it. From the day we launched, we received a ton of interest. Sign-ups poured in. But a sizable percentage of those sign-ups never actually used the product. They’d sign up, then never install our extension into their development environment. There were plenty of ICPs in the funnel to work with, so we looked at the data and thought: “A lot of these people probably aren’t professional developers. They’re ‘vibe coders’, we’ll support them eventually, but for now let’s not get distracted and stay focused.”

The Billion Dollar Blind Spot

I completely missed it. While we were building incredibly powerful tools for enterprise engineers, tools that genuinely helped them ship faster, navigate complex codebases, and reduce toil, we missed the entire emotional center of the market.

We weren’t optimizing for the dopamine cycle that was driving all those viral tweets, YouTube videos, and midnight coding sessions. We weren’t capturing the magic of transformation, that first moment when someone who couldn’t code suddenly could.

And here’s what makes this particularly painful: it wasn’t just a business mistake. It was a mission misalignment. Our mission is to unlock creativity. Why limit that mission to people who can already create software? We built for the people who needed us but didn’t emotionally crave us. Meanwhile, there was an entire audience craving the experience we could have delivered, and I’d originally written them off as “not the target segment.”

The Irony of Conventional Wisdom

Here’s what makes this especially ironic: earlier in my career, I’d read all the business books about starting with a niche. I understood the wisdom, and I agreed with the theoretical soundness of that logic. And then I completely ignored it and went with my heart. At Wrike, I recognized that everyone in the business world collaborates and manages work, so I developed products that were valuable to everyone, not just a narrow segment. I tried to boil the ocean. Every business school would have told me I was wrong. And it worked brilliantly.

But with Zencoder, in this specific decision, I somehow followed the business wisdom over my heart and my mission. Sometimes, the users who don’t fit your carefully defined customer profile are actually showing you where the real energy is. Sometimes the “distraction” is actually the signal. Traditional strategy—nail your segment, then expand—can blind you to emergent behaviors and emotional undercurrents that drive entire markets.

The Lesson

Now, I’m following my heart and building to truly unlock the magic without the fine print of “for this segment”.

Luckily for me, the AI market moves so fast that every day is a new opportunity. As “vibe coders” manage more complex environments, they become closer to the needs of professional engineers. As professional engineers shift towards AI-first workflows, they need an AI-first product instead of a code-first IDE. As LLMs and our agents improve, and the results become increasingly predictable, there’s an opportunity to bring a product to market that benefits both audiences. It’s time to follow my heart. As we work on our next launch, we’re taking everything we’ve learned from delivering enterprise-grade power and reliability and opening it up to everyone – from vibe coders to experienced developers. 

Ultimately, we are all on this journey together to create better software, no matter our background.