I wrote a Vibe app in just two hours (and regretted it the next day) | Towards Data Science

My daughter is three years old years. She broke her toy guitar and wants a new one. Instead of another toy guitar, I bought her a real ukulele. It’s small, lightweight and perfect for a little girl. Plus, it sounds much better than an annoying toy guitar.

She liked it. But I’m the one who played with it the most. I don’t know how to play the ukulele. I studied and it was fun.

One Saturday morning, at exactly 7:15, I came up with the idea for an app. Have you seen karaoke apps? You sing karaoke while the lyrics scroll down on the screen. I wanted to build one for ukulele players (for myself). They can see song lyrics and ukulele chords scrolling across the screen. They can sing along while playing the ukulele.

The best part: you can find ukulele chords for almost any song ever played on the planet. It doesn’t have to be “Country Roads” or “Somewhere Over the Rainbow” all the time. You can enter any unpopular song; the app can still generate chords for it. I want to create it using an AI agent that finds a song on the internet (mostly YouTube) and uses it to determine the chords. So, as long as the song you typed into the search bar is somewhere on the Internet, you will find the chords for it.

I sat down in front of my computer, opened Google AI Studio and told it what I wanted. My wife wanted me to join the family for breakfast, but I kept saying, “One minute please,” taking up even more time. By 8:30, I had not only developed the app, but deployed it, bought a domain, linked it to my app, and shared the link in a Reddit thread.

This is what it looked like.

Ukuflow is an AI app written in Vibe for those learning to play the ukulele. This is the home page

Ukuflow is an AI app written in Vibe for those learning to play the ukulele. This is the song page

I am a professional programmer. I’ve been programming for the last fifteen years of my life. Believe me, if I had to develop all this myself, it would take me a good six months. But I did it in about an hour, without even writing a line of code.

Honestly, the AI ​​is amazing.

Later that day, I checked my post on Reddit. I received a fantastic response. One redditor even said: “Sounds like the uke app I’ve always wanted!I slept well that day, feeling like I had finally created what the world wanted.

The next morning I checked the statistics in Google Analytics. I have about 200 people try my app. Amazing! I then checked the Billing tab in my Google Cloud Console. It was here that I felt that what I had built had failed.

I spent $52 to launch the app in just a day. The AI ​​agent is responsible for most of this. Every time a user visits the site, they can use this colorful feature to search for songs they thought they would never find. And this increased the cost dramatically.

My AI app coded Vibe lasted just one day and cost $52.

You might argue that this is normal for such an application. I agree. But there was more than one reason to close the app.

A business that fails at the division level doesn’t make money.

Before this, I wrote two blogs myself. I monetized them through display advertising and affiliate links. From this experience, I can say that in general you can earn about 10-20 dollars for every 1000 visits to your site. If you have something to sell, you can aim for $30-$50. This means that my 200 visitors would earn a maximum of $4. But at what cost? 52 dollars!

Unit economics is failing and I don’t like it. Even if I buy shares of a company, I look at how much it makes from each sale. This is something I learned from reading the works of the greatest investors the world has ever seen, such as Peter Lynch and Warren Buffett.

Some say that the cost of creating such an application would be recouped in 6 months. But keep in mind that $52 is the running cost of the app. Not the cost of its development. I developed it for free. The free Gemini credits were enough for everything I wanted to do in this app. However, operating costs will continue to accumulate.

Losing $48 ($52-$4) per day will ultimately result in losing $17k per year unless I find other opportunities to make additional sales.

Is the pain real?

Why am I deciding?

The ability to generate ukulele chords for literally any song on the Internet.

Of course, many people will want this application. I understood this from my pain. I loved playing the ukulele and as a beginner I would love to have something like this. LIKE A BEGINNER!

But what happens when I master it? I probably draw the chords myself from my experience. I don’t need an app for this. This means that the market I originally envisioned was even smaller. This is not some ukulele player; It’s especially difficult for beginners to find chords and coordinate singing and playing together. Professionals don’t need this.

Okay, what about the newbies? Won’t they use it enough? They would do it. But beginners usually tend to play a dozen songs that are widely considered “beginner-friendly ukulele songs.” If this were for my daughter, it would be something like “Baa-baa-black sheep” or “The wheels of the bus go round in circles.” For adults, probably the same old Country Roads. For beginners, popular songs are enough. Professionals know how to find chords for other songs.

But when I used AI to identify chords, I burned credits. That’s why $52. This is an expensive whistle that I hung on a really helpful Christmas tree.

The problem is real, but the solution is redundant.

Instead of a complex AI agent, a group of statically generated pages will do. And that’s exactly what I did now. If you visit ukuflow.com today, you will see a bunch of songs that you can select and sing along to. Made by a beginner for beginners.

Start small and repeat.

As someone who has worked on many SCRUM projects, I should have known better.

But I also think this is the smallest experiment I’ve done: it failed with only a $52 bill, not a $5,000 or $50,000 bill. I also made course adjustments to make the songs static rather than generating real-time AI. But I still want to talk about it, because now that AI allows anyone to build almost anything, more can end up down the rabbit hole.

Best place to start testing your critical assumption about your decision. In my ukulele app it would be so people would want to see lyrics and chords scrolling like in karaoke apps. Even a single song page could test this assumption. This is mine minimum realistic offer. People will still find it useful and give feedback. I can decide whether to expand it by adding more songs, adding more features like AI generation, or scrapping the idea altogether.

Final Thoughts

At one point, I was the only person in my organization who programmed in Python. Today I also mastered Java, JavaScript, TypeScript and some others. But you know what? Neither did my mom (who never wrote a line of code).

When AI programming came along, I instantly lost my competitive edge. Luckily, I didn’t rely solely on programming. This is my first lesson.

Ideas have always been cheaper. There was no execution. While execution seems to have gotten easier with the advent of AI, planning still hasn’t. To succeed as a vibe coder, or at least not fail dramatically, we need to be more flexible. We have to listen to end users much more than before, and then iterate and test quickly. This is an even more important lesson I learned after seeing my Vibe code app not working.