মারি daughter wanted a new instrument after breaking her toy guitar, so I bought her a real ukulele. I ended up playing it most and wanted to learn. One Saturday at 7:15 AM, I had an idea: build a karaoke-style app for ukulele players that shows scrolling lyrics and chords.

I envisioned using an AI agent to fetch songs from YouTube and generate chords for any track, even obscure ones. I used Google AI Studio to describe the app, skipped breakfast to keep working, and by 8:30 AM had developed, deployed, bought a domain, and shared it on Reddit—all without writing a single line of code. I called it Ukuflow, a vibe-coded AI-powered app for ukulele learners.

The response was positive; one Reddit user said it was exactly what they’d wanted. Around 200 people tried it that day. But the next morning, Google Analytics showed the traffic, and Google Cloud Console revealed the cost: $52 in just 24 hours, mostly from the AI agent’s song-search feature.

Though I expected some cost, the unit economics were brutal—200 visits might earn $4 max from ads, far below the $52 spend. Even as a programmer with 15 years of experience, I realized the app wasn’t viable. I shut it down after one day, regretting the vibe-coded rush despite the initial joy of creation.