The New MVP

Coding in English

Feb 10, 2025

Few years ago I was working with a small team on building a credit scoring system for a fintech product. It would assess loan applications to minimize default risk. It took about two months to build it. The funny thing is, when we showed it to the upper management, they were like "wow, this is so complex, this is what we're looking for!" – that pretty much summed up how things worked back then. The more complicated your system was, the better, because your competitors would have a hard time replicating it.

In July of last year, I had a cool concept for a landing page. I wanted to do it so bad I reached out to one of the best designers I know. He told me “It will take 3 weeks and will cost $2,400” – I wanted to do it because it had some fancy animations and design, but ain’t no way I’d pay $2,400 for a landing page. So I did it myself. Sure it wasn’t near the quality of something done by a designer, but I got something working and made peace with it.

Crazy to think this was just a few months ago.

Do you know how much that $2,400 landing page costs today?

$1, adjusted for AI-generated output. It's no longer complex.

Today you can record a short video of whatever complex animation you like, pass it to whatever AI tool, and you can get it done in under 5 minutes. And that sophisticated algorithm from earlier which previously took months can now be implemented in just a day.

Historically, the core intellectual property of a software product was fundamentally rooted in ‘how complex it was’ because it demonstrated how much effort went into research and product development. This complexity created the barrier needed for differentiation, such that it’s impossible for someone to wake up and decide they’d do the same exact thing without probably bleeding resources and time.

This has entirely changed. Today, we shifted from asking “is this idea feasible to implement” to “which ideas to implement” now that we know anything is possible. The barrier of entry to any sophisticated feature that used to signal competence is now almost zero.

Some people would think this is frightening, but let’s think about this: If we can now execute on complex ideas at a very high speed, then the real complexity lies in having original ideas to start with. The barrier is no longer technical – it's creative.

The new MVP

The way we used to implement a minimum viable product, i.e the simplest functional form of an idea, was this: We have an idea in mind, maybe a handful of features and a few designs, and we want to develop a simple prototype with a simple interface. It would typically take anything from a few weeks to a few months to implement this MVP - so this is the old model. This is how we used to think about implementing a prototype from scratch.

Today this has fundamentally changed. Now we can prototype at the speed of thought. With just a prompt, you can generate a complete prototype – including the entire codebase, database, and deployment, into a fully functional software package. What used to take weeks or months to build as an MVP can now be done in a few days.

This changes what defines an MVP, because no one is expecting *just* a working prototype anymore. Today the general expectation from an MVP is a mature product that’s working with almost zero flaws. Not only that but also the interface has to be so good because it no longer reflects what you had in mind when you started working – it actually now reflects how well you prompt and iterate with your AI. Because imagine having this tool that could program or design anything, and you still end up with a poor user interface – what does this tell about the product?

If you think deep enough in your prompts, it would be reflected on product quality and design. This is the new standard.

English is the new programming language

Since we’re shifting from programming languages and design languages to just pure English, this puts clear emphasis on the quality of your thoughts. Because now, more than ever, if you give your AI a bad prompt (i.e., bad English), this is what you're going to get out of it. Actually, we now have a clearer indicator of 'quality of thoughts' than any time before.

In the past, it was easy to blame poor results on team members or processes or systems, but now there are no intermediary layers, it's just you, your AI, and pure English. Even most communication within the team has been less about how to implement stuff and more about how to communicate it to AI. This wasn’t even a thing just one year ago.

Taste is now more relevant than ever

No wonder conversations about taste are recently common, because it’s probably one of the last things AI can take away from you. If your thoughts could be translated in real time into a working product, then it all comes down to what you want to see out in the world, what you want to create from nothing, and how you shape it to create value. So things that were not widely discussed before like taste, originality, creativity, are now the most important again.

Now that we already know that more than 90% of your software is probably written with AI, let’s not talk about the complexity of your software or about the feasibility of implementing it, because we know that’s possible, let’s now talk about what brings real value to your customers? what is the mixture that goes into your ‘building’ process? What’s unique enough about it that it is not replicable by a competitor?

What to do about it

I like the fact that lots of conversations are now shifting towards values rather than tech. For example, the gist of this post from yesterday by Sam Altman was that ‘agency, willfulness, and determination’ are the new differentiator.

It seems things like mindset, will, and taste were always ‘presumed’ as a backdrop to seemingly more important things like technical abilities. Now that technical abilities are moving to the backdrop, the conversation is becoming much more interesting and we are here for it.