I’ve always been a builder. Whether it was my own startup or working within a 140-year-old company, the goal was always the same: make tech useful for normal humans.
Lately, the buzz around “vibe coding” has doubled since I started doing it summer 2025. The main reason is because AI coding agents are significantly better and can actually give a non-technical builder superpowers. I decided to put this to the test in my new series, where I give myself just 10 minutes to live-vibe-code an app. No dry runs, no scripts, no edits.
“Vibe Coding” in Google Trends
The "10-Minute" Plan
I chose a "softball" project for the first episode: a simple countdown timer.
My requirements were simple:
A fun and playful visual style.
The ability to count down into negative numbers.
Numbers that turn red when time is up.
A chime notification when hitting zero.
I used Claude Code as my primary tool, following my own first rule of vibe coding:
Pro Tip: You don’t need a perfect prompt. You just need to start a conversation.
The Reality Check
I started the clock at [02:53], and Claude quickly guided me through the design process, asking smart questions about architecture and visual style that I hadn't even considered. Admittedly, this is my first time using Claude Code to build from scratch since a few months ago, and this “brainstorming” skill blew me away. It recommended a single HTML file approach—simple, clean, and portable [08:15].
By the 8-minute mark, Claude didn’t even start coding yet; it was still writing documentation [10:38]. Writing the design doc and implementing the actual code took longer than my 10-minute window allowed.
Spoiler alert: The "10-minute" app actually took closer to 30 minutes to complete [13:38]. But the build process is extremely comprehensive.
The Lesson: Compound Learning, Not Just Consumption
Most people overestimate what they can do in a week but underestimate what daily compounding can do in a year.
What I learned from this experiment:
Vibe coding isn't "magic." You still need to make product decisions. AI is a guide, but you are the strategist.
Time management is different with AI. The brainstorming is fast, but the "implementation" phase still requires patience, tweaking, bug fixing, and sometimes endless iterations.
Expectation vs. Reality: 10 minutes is enough to prototype, but 30 minutes is what it takes to get a (very simple) functional, bug-free product [13:38].
Does it work?
In the end, the app was exactly what I asked for. It has a playful theme, custom presets, and yes—it blinks red and chimes when it hits zero [14:31].