Apps Are Dead, Long Live AI Apps ⭐

I was lucky enough to grow up at the dawn of the personal computing revolution and I experienced the first home computers in the late 1970s and 1980s as each was released in turn. Those were heady times. I was young and captivated by this technology, and I dreamed about what I might do with my own computer.

Today, personal computing is common, and thanks to the rise of smartphones, especially, this thing that was once futuristic and even a curiosity seems almost old-fashioned. But back then, it was different. There was no formal market for computers or the software they could run. There were competing platforms from many companies, each mostly incompatible with the others, and it wasn’t clear which, if any, would succeed. And so the first home computer users were largely on their own.

This wasn’t a problem for the enthusiasts like myself who immediately gravitated to these computers before there was much if any good software to use. We could simply experiment with the BASIC that came with each computer and write our own apps. I embraced this immediately, and my first two computers, the Entertainment Computer System (ECS) add-on for Intellivision and the Commodore 64, both had a strong emphasis on graphics and sound. And so doing things like making a balloon sprite move across the screen above text, an example provided in the Commodore 64 user guide, was a nice peek at the possibilities. Which, at the time, seemed endless.

Not surprisingly, businesses sprung up to supply software of varying quality levels as well, many started by enthusiasts who were older than I was at the time. And as the market for home computers shifted to what we now call personal computers, or PCs, that software got ever more professional and useful. The market for enthusiasts who wanted to create their own software saw a brief resurgence with tools like HyperCard on the Mac and Visual Basic on the PC. But by the mid-to-late 1990s, that era was officially over. PCs and the software that ran on them had become quite sophisticated.

Given my old-school credentials, I would sometimes fret that those growing up in the 1990s and 2000s didn’t have that DIY grounding. But this evolution is both natural and inevitable. Just as car owners no longer needed to be mechanics or understand how manual shift transmissions worked over time as that market matured, PC owners no longer needed to understand why or how software worked. These things were tools that they could use. And they did.

Well, at least in first-world countries they did. PCs aren’t just sophisticated, they’re expensive. And as third- and second-world countries evolved, their citizens interacted with technology in ways that were quite different from what I would have considered normal. Likewise, reliance on connectivity was often problematic in these places, and that impacted their relationship with technology.

These trends continue to this day, but over the past 10 to 20 years, technology companies have made various efforts to lessen the technology gap. It’s not worth reciting all the major examples, but one of them in the form of Raspberry Pi jump-started a new era for creators and makers that harkens back to the home computer era I lived through. There’s a bigger emphasis on hardware this time, too, in the form of sensors and other add-ons one can access programmatically. I appreciate Raspberry Pi for giving this new generation an appreciation for computer science and, of course, doing so affordably and opening it up to a much wider audience. So where I might have made fun little programs in Microsoft BASIC on a C-64 and saved them to a Commodore Datasette, kids growing up today can use a much more powerful system with a much more powerful programming language and interact with interesting hardware sensors and devices. How fun.

As for the rest of us, we’re sitting here using our phones, tablets, and PCs as the tools they are. Most of us do not know any programming language, so we can’t create our own apps despite generations of attempts at low- and no-code automation solutions. We’ve witnessed and endured the rise of Big Tech and the resulting enshittification of their offerings. And we now live in what I think is a golden age of so-called Little Tech alternatives that can replace those apps and services we no longer find to be useful or valuable. Things are constantly changing.

Back in October, I wrote that AI would mean the end of apps. What I meant was that existing apps, things like Microsoft Word, Paint, and the like, are evolving to provide back-end interfaces to their internal capabilities that can be accessed by AI models and agents. If you think through what that means, this is in some ways a continuation of a trend that started with web apps, where so-called native apps that are locked to a single platform are giving way to cross-compatible apps that work everywhere. That’s a big threat to the mobile app stores that now dominate app distribution. But it’s also freeing in the sense that it makes the hardware less important and less of a restriction. If the apps we use just work everywhere, we can use whatever hardware we want.

But AI triggering the end of apps isn’t the full story.

Increasingly, moving like a tsunami in the way all things AI-related seem to move, we’re seeing the rise of a new generation of software solutions that also speaks to that home computer enthusiast era, albeit without the need to learn a programming language or be technical in any way. That is, with AI we can suddenly describe what it is that we want, our intent, and AI will do that thing for us. Sometimes, it involves programmatically accessing existing apps–like Word and Paint, as noted above–but now, suddenly, it’s starting to mean not using those apps in any way at all. What we’re getting is something new and, intriguingly, something that is personal to us.

This is not vibe-coding, one of the most widely misunderstood terms of this short but fast-moving AI era. But it is what most people thought vibe-coding was (or is). And this shift marks, I think, another nail in the coffin for apps as we now know them, meaning apps that run on some personal computing hardware device, and often only on that one device, made by businesses and other professional developers. This emerging new era of what I will for now call AI apps, lacking a better term, is something else entirely.

Consider one of the examples I always use when I try to explain when I might help us save time and money.

Each January, I write an article about the previous year’s PC sales. When I do this, I go back to my archives and find the Excel spreadsheet I used the previous January to create the chart that accompanies this article, and I make a new copy of it. I add the next year’s PC sales data to the spreadsheet, and then I have to create the chart that’s based on this data. This is not advanced Excel usage by any stretch. But I never use Excel, except for this one instance each year. And each year it’s like I’ve never used it before. I struggle to make the new chart, and sometimes I can’t even get it right.

Today, you don’t need to know how to use Excel to make this kind of chart. You could simply feed the data to ChatGPT or whatever AI you use and tell it to make the chart. If the chart it makes is not quite what you want, you can just tell it to change the chart, perhaps by making the columns a bit wider or different colors. Whatever. You have this idea, this intent, you explain it to the AI, and, over the course of some conversation, it makes the thing you want. You don’t need to learn or master Excel to make it happen.

Did the AI use Excel for that? No. Do you care that it didn’t use Excel? Of course not. All you want is the thing, in this case the chart, which is a graphic you can use anywhere. AI didn’t completely replace Excel here, but it could. And it will. And when it does, Excel will go away. First for the general population. And then, too, for those who consider themselves Excel experts. Their skills will no longer be required.

So that’s one thing. In the short term, Excel will semantically make its capabilities available outside the app to AI. In the long term, Excel is not required. Only the AI will be required. We will use AI, not Excel. A future generation will be amused to discover that Excel even existed. Just as we today are amused to know that you had to be an expert mechanic to even consider owning a car. How cute.

But there’s another thing. And it is tied to that experience that is not exactly vibe coding, at least not as originally envisioned by the person (a professional programmer) who coined that term. And this is what I will call AI apps for now. Again, lacking a better term.

Also lacking in imagination, I will stick to a simple example. Let’s say you are a child in school and you have a math test coming up. You could use AI to develop flash cards, quizzes, or whatever interactive experiences one could need, and you could use those one-off apps, lacking a better term, to help you study for that test. When the day arrives for you to take the test, you will likely never reference that “app” again. But you may use the experience of making it as background for some future task, be it another test in school or whatever else.

You might also make that app available to others. Others who might use it as-is, which would be unlikely for something that specific. But also others who might change it to meet some other need, like a different test or task. It might be used like a template, in other words. A guide. A starter app. Something that will be exchanged, just as enthusiasts in the 1980s might have exchanged BASIC source code.

As AI improves, one’s ability to make these apps—again, I don’t like that word for this, but it is what it is—will only get more sophisticated. So instead of making flashcards or whatever for a math test, you might—as a child or an adult—make something more sophisticated. Something that is specially tailored to some workflow, some need that you have that is maybe unique to you. A one-off app, in other words, just for you. A truly personal app.

I’m no futurist: There are examples of this everywhere right now. My point is only that this capability will only get more sophisticated. And while we have all these stupid terms for the technology that will make this happen, from agents and inference to MCPs and whatever else, we will not care about any of that, just as we do not care about how our cars work internally. We just want to get something done. We will just do it.

That something could be the full workflow from my Excel example, where I need to not just get the data to add each January, but make that chart and then write and publish my article to the web. Today, I use multiple apps and services for that. In the near future, I will use just one, and that one app will be some AI app that knows a lot about me, the way I work, and what it is I do. And when it makes a mistake or just doesn’t quite do what I want, we will converse and we will fix the problem together.

These apps that AI is and will create will not be stuck in or limited to some app store. They will not be stuck on or limited to a specific PC, phone, or other device. They will be universal and available everywhere, not native (which should be considered a negative now, not a positive). If I create a fun little game on my phone, I will be able to play it on my PC or TV. If I create a little app that tracks the news, it will alert me wherever I am using whatever devices I own, and it will not be locked to one thing in one place.

There are prototypical examples of this that pre-date AI, from websites in browsers to the extensions that make browsers better. There are the so-called super apps on mobile that run sub-apps, or mini-apps inside them. There are extensible apps like Microsoft Teams that support add-ons that connect that one app to an unknowable number of other services. This kind of thing is common and all around us already. And AI will exponentially expand its usefulness while ending our reliance on the very specific apps we’re so used to today.

In short, AI does mean the end to apps as we now know them. But it also means the beginning of a new kind of app. A new kind of app that anyone will be able to make, and not just enthusiasts who take the time to learn technical topics and skills. This is the very point of personal technology, to my mind, the literal democratization of capabilities that were once controlled and only used by scientists, engineers, and other elites.

It could mean the end to abusive Big Tech monopolies, though you can be sure they will do what they can to thwart any progress they can’t control, as always. We’ll see. Maybe we can keep the enthusiast spirit of the Commodore 64 and Raspberry Pi alive in this new era without succumbing to the terrible 30 percent fees and other restrictions that undermined the mobile era.

I do hope so. Because this is a future that excites me as much as those first home computers did 50 years ago.