Published on March 16, 2026
Back in 1999, Nick Swinmurn walked into a local shoe store with a camera. Not as a tourist. Not as a fashion blogger. Not even to buy shoes for himself. He just started taking pictures of the pairs on display. If you had been there, you probably would’ve thought, “What is this guy doing?”
That night, those photos appeared on a small, very 1999-style website. No warehouse. No inventory. No logistics operation in the background. When someone placed an order, Nick went back to the store, bought the shoes at full retail price, packed them up, and shipped them himself.
He wasn’t trying to build a billion-dollar company. He was trying to answer one simple question: are people willing to buy shoes online?
Ten years later, this little experiment, now known as Zappos, was acquired by Amazon for $1.2 billion. In 1999, the constraint was technology. Today, with AI writing code for us, the constraint is discipline.
Let’s get one thing straight from the start. If you hear people saying MVPs are dead, they are wrong. The core principles Eric Ries outlined in The Lean Startup still stand.
“That version of a new product which allows a team to collect the maximum amount of validated learning about customers with the least effort.”
An MVP was never meant to be half-broken or careless. It wasn’t about shipping something sloppy and calling it “lean.” It was about focus.
Not a stripped-down final product, but a deliberately small one.Not cheap, but cost-effective.All in service of one goal: validated learning.
For most of startup history, building digital products was expensive. You needed developers. You needed capital. Most of all, you needed time.
That friction acted as a natural filter. Prioritization wasn’t optional, it was survival. Founders had to ask: what is the absolute minimum development we need to validate this idea?
AI removed that filter.
The bottleneck shifted from building to decision-making. Modern AI tools can scaffold interfaces, generate backend logic, and connect integrations in a few prompts. Surface level it all works, until it doesn’t.
The effort needed to build something presentable has diminished. And when effort levels drop, restraint drops with it.
Ries defined an MVP as the version of a product that delivers maximum validated learning with the least effort. But what happens when effort becomes almost irrelevant?
When login with Google, role-based permissions, AI summaries, CSV exports, dark mode, and a full analytics dashboard can be scaffolded in a single prompt, what’s stopping us from adding them before we’ve validated our core set of features?
The question is no longer, “Can we build this?”
It’s, “Why not add it?”
AI didn’t change the definition of an MVP. It changed how easy it is to ignore it.
It’s tempting to load up the first version with features. As a developer, I’ve been guilty of this myself. When building becomes easy, adding “just one more thing” feels harmless. But ‘less is more’ exists for a reason.
Take the Zappos example. If your goal is to validate whether people are willing to buy shoes online, your entire focus should be on one thing: enabling someone to buy a pair of shoes.
Yes, you can add user accounts. Yes, you can build advanced filters. Yes, you can create a smart shoe finder. And any other seemingly amazing features that you can come up with.
But do you need any of that to test whether someone will pull out their credit card?
If those features break, the feedback becomes about bugs.“This filter doesn’t return any results.” “I got women’s shoes as a recommendation while I’m a man”
You end up fixing bugs and wasting time & money on features that were never essential to begin with. The real question remains unanswered. Do people want to buy shoes online?
Now imagine a simpler version: a clean list of shoes and a smooth checkout flow. Nothing else.
The feedback shifts. “Can you add filters for men’s and women’s shoes?” “Could you help me find shoes that match my style?”
That’s feedback driven by demand, not assumption. AI makes it easy to build everything upfront. Discipline is choosing not to.
There are two dominant ways founders are using AI today.
AI app builders like Base44 and Lovable dramatically lower the barrier to launch your own app. You can go from idea to product in days. They are excellent tools for rapid validation.
But that speed comes with trade-offs. Architecture is opaque. Scaling can be uncertain. Costs can rise unpredictably. And if the product takes off, migrating users to a custom-built system introduces friction and risk.
On the other side, AI coding tools like Cursor, ChatGPT, and Claude allow you to build on your own foundation from day one. You have control over architecture, scalability, and long-term maintainability. There are fewer limitations, but you carry more of the responsibility. You need a solid technical understanding. And still, you need discipline.
Both paths accelerate building. Neither replaces the need for focus.
AI didn’t replace MVP principles. It removed the time and effort constraints that enforced these principles.
The real advantage isn’t which AI tool you choose. It’s in how long you can resist building more than necessary.
Zappos didn’t start with infrastructure or the choice of which tool to use. They started with a question.
AI lets us build faster than ever before.
But the founders who win won’t be the ones who build the most.They’ll be the ones who learn the fastest.So why don't we start with the question?