How AI and natural language interfaces are making traditional software obsolete, and why the future of building lies in the physical world, not digital products.
Last year, I was thinking about buying software that connects to your Gmail, scans all your emails, and extracts all the invoices to a nice database with all the attachments organized. This year, I wasn't even considering buying it. I sat down with Claude Code and we built it in 90 minutes. Obviously not perfect for everybody else, but perfect for me. Exactly what I needed, nothing I didn't need. It was done.
I'm not saying that company isn't still making money today, but I would never pay for them anymore. I would never consider buying this software. And I think this is what's happening to most software right now.
One Prompt Goes Seven Hours Deep
One prompt can take you very, very far. Meta's research shows that one prompt can accomplish seven hours of software engineer work. This is already proven and you can check the research yourself. And in a few months, it's going to be doubled. Actually, less than a few months. This report was current as of yesterday, but as I'm filming this, two models came out that will probably outperform the previous ones significantly. So in a few weeks, it's going to be 14 hours, and everybody will have access to this. I mean everybody.
We are no longer going to pay for those apps anymore. The apps will evolve into something else I'm pretty sure about this. But what we have already isn't going to last much longer.
Think about that invoice app. It probably took a significant amount of time to build. But now that time is probably divided by seven. The market already knows this. They call it the SaaS collapse. This week, $285 billion was wiped out of Adobe, Microsoft all the software stocks.
The Collapse of Product Moats
In 2017, research showed it took 18 months to copy features from another SaaS company. By 2024, that reduced to two and a half weeks a 98% collapse (!!). I bet that in 2026 it'll be even less, probably one week.
Your product no longer has a moat because anyone can reproduce it. I'm not saying we can build Monday.com in one week or less (though possibly, possibly). I'm not saying that because they have so many use cases, such a big company, customer service, and all that stuff. But we don't need it. We just need the software for our own use case, software that works for us. We don't need it to work for everybody else. And we're definitely not going to pay for it anymore because we can communicate with our computer and just make it ourselves.
Those giants—Salesforce, you name it—are no longer moats. They're the starting point.
The Eighth Layer: Natural Language
What's really happening now is that you can talk to your computer in your own language, and the computer can talk to the computer in computer language. You no longer need to know Python or anything. The computer knows it. You just need to know English or your own native language, whatever it is, and you can communicate with your computer.
This is obviously changing the ground for all of us. It's another layer in the OSI model: the eighth layer, above the application layer. The mass public is going to use this layer mostly: the native language layer.
I believe the next computer isn't going to have a user interface anymore. It's just going to have some sort of chatbot or communication method. maybe it will transcribe us and reply, whatever. Call it Claude Code, Codex, whatever you want to play around with doesn't really matter. But this layer is going to be the layer of interaction because human language is the best user interface. That's your new computer.
Think about this: your mom will be able to talk with the computer in her language and ask it to open a website, help her pay bills, do research, search online, open Netflix—whatever she does with her computer. She doesn't need to remember where all the buttons are or use the keyboard (probably maybe the keyboard, yes, but maybe also transcribe herself) and the computer will be able to assist her with everything. The computer will be able to generate applications from scratch for her own use case.
If I put it in a better proposition: we can talk directly with the terminal. That's it. The days of clicking buttons (I'm not sure they're completely gone because there's something about visuals that will always be important) but they're halfway gone. We're going to talk, we're going to write, and that's it.
The Builder Argument (And Why It Won't Last)
You could argue that if a company wants to develop their own software instead of using other SaaS companies, they'll need to hire a builder, and maybe the cost of the builder will be higher than just paying for the software. That's a good argument, but I think it's talking about today's world. The models are going to keep improving endlessly, exponentially.
Maybe in one or two years, you won't need a builder because the guy working in HR who has no clue about code whatsoever will be able to talk with the computer in his language and create whatever he needs. He'll be able to analyze spreadsheets, connect to APIs.. even though he doesn't know what he's doing, the computer knows what it's doing.
I think in a matter of months, this argument about needing a builder won't exist anymore.
Think about how important it was to learn C# five years ago and now it's no longer important. Nobody needs to know it. I'm not saying nobody needs to understand how it works but learning computer languages has become something we don't really need to do anymore because the computer handles this part and will continue to handle this part very, very well.
There will be some point when the models will be able to one-shot Monday.com, Salesforce. they'll be able to one-shot it. Maybe not the entire website, not all the use cases, but the core functionalities, definitely.
The Biggest Technical Gap Ever
We're going to live in a world where everybody creates their own apps, sets up their own Chrome tasks, and works with a computer in their own language.
But now let's talk about the gap. What we're seeing right now is a huge gap between the early adopters, and most of society. This gap is huge. It's maybe the biggest technical gap ever made. I don't think I'm overexaggerating here.
And it's happening so fast. People who thought they knew everything two weeks ago, if they're not caught up, they already don't know everything that's going on now.
This gap is the issue we're facing right now, where you think, ""Okay, people still need software, people still need SaaS."" But eventually, the entire public is going to catch up.
How to Build Sustainable Software
Now, I'm not here to say don't make money from software, that nobody's going to make money from software. No, I think you can. But I'm here to talk about how we can build software that will be sustainable for the future. That's an important question to ask nowadays, right? Because when the entire public catches up to everything happening right now, it's going to be the end of the current software companies. It's going to be the end because we don't need to Vibe-code the business. We don't need to serve millions of users. We need to serve one: ourselves.
What's Going to Survive?
Imagine if you were to take a person with today's capabilities, 2026 capabilities, and send them back in time to 2020. They're going to crush it. They're going to ship software in quality and speed like never before. I bet they're going to demolish all the other competitors, all the other companies.
AI raised the bar so high and so fast that it's hard to even get a hold of it. If coding is now so easy and accessible for everybody, how can you still build something meaningful? How can you protect your business, your software? How can you create a moat?
I think we have two options here. Maybe more.
Option one: Complexity is going to go significantly (& exponentially) higher, and we will ship more complex software faster with higher quality. That's what's going to differentiate you. I think it's logical that our software will know how to do significantly more stuff.
Option two: Because the application layer is going away from most of the public, and the application layer used to make most of our money, we now have this new CLI natural language layer (I don't know what to call it). This is going to be the new application layer. When the models improve, your product is going to improve.
This means that if you're building software that has UI, the UI is going to go away. People will be able to create their own UI. So your software isn't about the UI anymore. Your software should be about the next layer, which is interaction between humans and machines. Your software should be language-first, language-native-first. It should write that layer, that top layer.
In a sense, your valuable software (software you can actually sell for a long period of time) is going to become an orchestration of agents, of LLM chains, of whatever you're going to call it. Plugins for LLMs, you can call it. Like superpowers for Claude Code. This type of software is the next generation of software, imo.
Satya Nadella, the founder of Microsoft, said it a few years ago and really made a big earthquake. He said that SaaS is gone and everybody's going to create their own SaaS. It took a few years, but this is what is definitely possible now. Only we are doing this right now, but soon the late bloomers will catch up. You bet they will.
My Plan: Back to the Physical World
I'm going to share with you my plan. I'm not seeing myself looking for new products anymore in the digital world. I'm just going to talk with my LLM, my large language model, and I'm just going to create whatever I need because it's super fast and very efficient for me.
And that brings me back for my businesses to the real world, to the physical world. Because over here in the physical world, we have so many issues that still need to be solved. And we can use the power of LLM (the power of AI) to leverage and create more solutions in the physical world.
Look, for example, at robotics: $8.8 billion in funding just last quarter. The software index might be down 20%, but the hardware index is up 16%. Someone called it "atoms over bits." I think this is a great place to look into because the money seems to be moving from bits to atoms.
The Third Option
I would urge you: instead of shipping amazing software, even though I offered you two options, there's a third option. Build beautiful stuff for the physical world.
I think the days of clicking buttons are over. The days of paying someone else for their software are pretty much done. And the real opportunities for builders, for entrepreneurs, for everyone, are right here in the real, real, real world.
I think Earth, the street, is waiting for us to show up.
Good luck.