How the Minimum Viable Product shrinks in the AI era. When a 6-month prototype cycle collapses to a day, the founder's job isn't planning — it's the frequency of attempts.
You probably have a business idea you keep hesitating on. "I'm not ready." "I need more funding." "I need an engineering team." A year passes like that. This essay takes that year and shrinks it to one night.
If you've never built an MVP, you can still follow — we'll go slowly. Today's examples are Airbnb and an AI startup hands-on event, but the principle works for any idea, in any era. Specific tools will change. The spine of this essay will not.
Here's the principle. Plans don't build businesses. Frequency of attempts builds businesses. A plan is letters on paper. An attempt is contact with the world. The market can only teach you through contact count.
Let me unpack MVP first. Minimum Viable Product. Meaning: make it as small as possible and put it in the world. Why small? Because big takes long, long delays feedback, delayed feedback means when your direction is wrong you've already gone too far.
In the AI era, the "minimum" in MVP shrinks dramatically. Past MVPs were still pretty large. Launching one homepage required a designer, a developer, a server admin, and weeks of time. Today AI writes code, the cloud hosts the server, APIs handle complex features. MVP size has shrunk by a full generation.
Classic example. Today Airbnb is a $93 billion global company, but its first MVP in late 2007 looked like this.
That's it. No hotel app, no payment system, no reviews. All they had was a problem to solve (rent is due) and an observed fact (conference attendees couldn't find hotel rooms). They combined the two using things already in the apartment.
The 2008 homepage didn't even have today's name. It was "Air Bed and Breakfast" — air mattresses and a morning meal. That is what an MVP looks like. Not perfect, not elegant, not scalable. But runnable tomorrow. Airbnb reached today's form by repeating this small experiment once, twice, ten times.
Notice the important thing. The first MVP was not fancy technology. They combined things that already existed (apartment, mattress, eggs). In today's AI era, even more already exists for you. Free AI models, free cloud credits, free dev environments, open APIs. The ingredients are already sitting there.
Think of the first day you ever threw a ball in a yard. Your direction was off, your force was uneven, it felt awkward. But after a few throws your hand learned. Distance came. Footwork settled. The only way the body learns is repetition. No amount of memorizing "elbow at 90 degrees, weight transfer back to front" makes the actual ball fly.
Business is identical. Markets can't be learned from books. You learn by throwing. In the past, making one ball took 6 months. Compare two founders — one throws once every 6 months, one throws once a day — after a year. The first threw 2 balls. The second threw 365. A 182× learning gap opens up.
Look at how time-to-prototype has collapsed across eras.
| Era | Time to build MVP | Attempts possible per year |
|---|---|---|
| 2007 (Airbnb founding) | ~6 months | 2 |
| 2015 (Lean startup era) | ~1 month | 12 |
| 2024 (AI era) | ~1 day | 365 |
The core of these numbers isn't speed. It's number of attempts. A founder who learns 2 times per year versus one who learns 365 times per year — after 18 months, they're not the same person anymore.
365 throws with no plan always teach more than 2 throws with a perfect plan.
When an idea hits, many people freeze because they don't know where to start. One question solves it.
"What would it take to make the smallest version of this tonight?"
The answer splits into three.
"I already have it" → run it tonight
Like a living room and air mattresses, if your ingredients are already on hand, start tonight. Polish doesn't matter. The goal is first contact with the world.
"I'm missing a little" → AI fills the gap
One or two ingredients missing. Don't code? Ask AI to code it. No design sense? AI image generators cover it. Need narration? Text-to-speech ends that. AI fills most missing ingredients.
"I'm missing a lot" → go to a hands-on event
When you don't even know where to start. Don't sit alone. Go to an AI demo event, a hackathon, a startup camp. Touching beats reading ten books. Events like AWS GenAI Loft run for one week at a time, including in Seoul. One visit rewires the picture in your head.
Task: an app that auto-summarizes long meetings.
Old way:
AI-era MVP:
In one 2-hour evening, you're at the point where you can ask three users "want to try this?" Collect feedback. Spend another 2 hours tomorrow evening based on the feedback. Improvements stack daily. The 7-month pattern can't match that rhythm.
Here's tonight's starting move.
Pull out a sheet of paper. Write big at the top:
Tonight, 2 hours —
what's the smallest thing I can show the world?
Two hours. Not a day — two hours. Tomorrow morning, show it to three people only. Family, friends, a group chat — anyone. Record their reactions. Interesting, weird, unnecessary, too complex — any reaction recorded is data.
Do this every night for a week. At the end of the week you have 7 data points. A far sturdier foundation than any single perfect plan.
Let me close.
MVP means Minimum Viable Product. In the AI era, the "minimum" has collapsed. What used to be a 6-month prototype is now an overnight build. Specific tools will keep changing. Today's AI models will have different names in three years. But the principle stays. Build small, ship fast, get feedback, build small again — the speed of that loop makes the founder.
Keep one question in your body — "What would it take to make the smallest version of this tonight?" That single question converts a year of hesitation into two hours. Not planning — frequency.
Good ideas don't win. The person who throws 365 times in a year wins. This truth runs straight from Airbnb's first mattress to today's AI startups, unchanged. Technology changes. Frequency doesn't.
Three words to carry: Small. Fast. Often.