MVP in the AI era: first close the smallest loop

A piece of advice you hear all the time in startups: resources are limited, don't overthink it, build an MVP first, a Minimum Viable Product. Make the most basic version, find customers, test the market, then improve bit by bit.

I've always found that logic very reasonable. But recently I came across a different view:

When AI makes building products easier and easier, should the starting point of a startup change too?

Before, you'd spend months building a product, and the ability to develop it was itself a barrier. You had some time to test the market and listen to feedback.

Now, you build some tools with AI in three days. You're delighted and feel very efficient. But then you look around and, huh, someone else has built it too, with roughly the same features.

You have AI, and so does everyone else. As building gets faster, the track gets crowded faster too.

If someone can copy my features tomorrow, why would customers keep using me?

And if the only answer is "because I did it first", that advantage may be shorter-lived than we think.

Founders should dare to think bigger

I think founders should dare to think bigger.

There were problems we used to look at and think: too complex, the team isn't big enough, development is too expensive, forget it, let's build something simple first. Now that AI has lowered part of the cost of building, can we take another look at the problems we once couldn't afford to touch?

Of course, that doesn't mean making the product complicated on purpose. Customers want the problem solved. Ideally you handle a mountain of complexity behind the scenes, and for them it's very simple to use.

Tesla makes me think of this direction. Its ambition was to redefine what an electric car should be: could people want one for its performance and driving experience, rather than accepting compromises for the sake of the environment? The first Roadster did borrow from a Lotus chassis, but the goal behind it was big. The first step was to enter with a high-priced sports car, then move gradually towards the mass market.

I think that distinction matters.

You can start somewhere very small, but keep a bigger direction in your head. AI startups are the same. If our imagination for AI is just adding a chat box or an auto-generated report to an existing product, isn't that a bit of a waste?

Many things used to be done a certain way because of limits on cost, people and technology. Now that those limits are changing, do we have a chance to rethink how the whole service should be delivered?

First, close the smallest loop

Does that mean we should shut ourselves away for two years and only meet customers once we have an amazing product? I think it's the opposite. The bigger your idea, the sooner you should find a real user and see whether the pain you want to solve actually exists.

Here I think the MVP can be understood differently: first, close the smallest loop. Put simply, get one real customer to actually solve a problem by using what you've built. The scope can be tiny, but the path from hitting the problem, to taking action, to seeing a result has to work end to end.

Say you're building an AI business analysis tool. The boss asks why sales dropped yesterday. You give him a beautiful report; he reads it, says "oh, I see", and that's the end of it. But if you can go further, help him find the problem, get the person responsible to act, then track whether things improved after the change, the value is completely different. Help one company solve one recurring problem first, see whether they keep using it and are willing to pay, then keep improving from real feedback.

And every time you finish a round, ask yourself: what did this round leave behind? Do we understand this industry better? Have we built up sound judgement and reusable processes? Can we do it better and faster next time? If you start from zero every time, you may be incredibly busy without building any advantage. Only when these capabilities accumulate slowly can you reach the day when others think it looks simple, and only find out how hard it is once they try.

The MVP still matters. Surviving first still matters.

But our dreams can be bigger, while the first step needs firmer footing. Find real demand as early as possible, close one loop, and let every step become the foundation for the next.

When everyone can build a product quickly, are we accumulating features, or accumulating capabilities that others find harder and harder to catch?