I just built three apps in under a week. I’m already working on more. What made that possible might just kill apps as we know them.
In 2009, Apple’s marketing team coined “There’s an app for that” and changed the collective understanding of software. There were always Applications, but the iPhone’s App Store made them way more accessible, simplifying the management of them down to an icon on your homescreen.
The next decade was a golden age for apps. Indie developers made small fortunes. Airbnb, Uber and a slew of other startups made app development cool. These companies got bigger, and a bunch of other massive tech companies flooded the app market. Indie developers still exist, but they don’t get the love they used to. Especially from Apple.
In the mid to late 2010s, I worked as a producer/project manager on a team developing apps. I was a generalist, working with specialist UX and UI designers, iOS developers, Android developers (before most things moved to React Native for both platforms), back end developers, quality assurance developers and more. I helped keep projects on track and like to think I imparted some product direction into the projects too. It was pretty awesome.
I’ve often dabbled with web development on the side, but was always jealous of the mobile developers. I loved the apps that were around at the time, and some of the apps we were making were pretty cool too. It was undoubtedly a boom, and the mobile developers were in hot demand.
At the time, we often got asked a lot how much an app cost. The answer, factoring in the time and effort of a great team building a bespoke solution, was usually eye-watering for individuals, and justifiable only for the medium-large brands or startups with VC funding. But for a ballpark, we’re talking $100k+ in 2015-2020.
People asked this because they had ideas. Apps cost them a few dollars to buy, so they wondered if they could make one themselves. I was one of those people. I had my own ideas, and wanted to make them. But I couldn’t learn mobile development on top of a full-time job and a young family. So I made little side project WordPress sites and scratched my itch that way.
In 2019 (I think – everything pre-2020 is kind of a blur), I pitched an idea for a self storage app and got some funding from my company to make it. We spent 6 months part time and ~$100k on a great app that used machine learning to help catalogue your stuff. I still think it could have been a great business, but it didn’t work out.
I honestly think I could re-build that app today in less than a 9-5 work day. It sure wouldn’t be far off. And I haven’t even worked on apps as my day job since 2020.
Last year, “vibe coding” gained popularity among the tech nerd circles I still find myself in. Depending on who you spoke to, it could be a positive or negative term. But in the last couple of months, the vibe on AI has shifted. New Claude and ChatGPT models are not just a bit better at coding – they’re a lot better. Insanely better, even.
Genuinely excellent software developers are admitting on the record that they haven’t written code themselves for weeks or months. I was sceptical, of course. They’re developers – they know this stuff. What about me? Someone who admittedly has a better understanding of apps than most, but has never written a line of native code.
So I had a go at vibe coding.
A week later, I’ve built 3 native iOS apps, using Swift and SwiftUI, deploying to my iPhone and honestly having an incredible time. At least, I think they’re built on Swift. I only know that because that’s what the AI says. I’ve barely looked at a line of code.
I tried a few different tools, and got a lot of great ideas from this video. AI Studio is kind of amazing for an initial pass at a layout. ChatGPT is okay for images and ideas, but not development (though admittedly, I didn’t try Codex). Replit is great for building full web apps, but feels a bit janky for mobile apps once you try out what I eventually settled on: Claude Code. And don’t be as daunted by that as I was. There’s a UI version. And there’s a more work focused version called Cowork that seems to do a lot of the same stuff.
There are lots of things I know from the world of software development that make doing this easier. I know the jargon, from wireframes to prototypes, to git, to how to build and run an app in Xcode. But Claude can help you learn that jargon too. And I don’t think you’ll need to know it for that long.
I feel like my years of working in the field give me confidence that you could build in a few hours just playing around. And spending a few hours playing around is exactly what you’ll do. It’s incredibly addictive. The dopamine hit of making a thing that looks and works like something that used to cost more than a car is kind of incredible. And Claude will always ask what’s next, no matter how much you’ve taxed it on the last task. Unless you run out of credits, that is. Which you will.
You can go from idea to working app in a matter of an hour. But you shouldn’t stop there. It won’t be great. A few more hours of conversations, frustration, incredible breakthroughs and probably bug fixing (taking a screenshot and saying it doesn’t work is all the QA Claude needs most of the time) will lead to something much better.
I started by making a running app (inspired by this article). I made it in Replit first (using React Native), and then in trying to build it on my phone, the AI was telling me to install Node. That felt fiddly. I asked Claude if it could rebuild in Swift. It recommended just installing Node, which was probably the right call, but I persisted. We had a working Swift version literally minutes later.
It gets my run data from Apple HealthKit and presents it in a much nicer way than Apple ever has. I can set a yearly distance goal, and with every run I can see if I’m catching up or falling further behind. It makes me want to run more. It’s an incredible little standalone app, and I keep thinking of new features to add. Widgets, specific run info, run plans, and more. It’s great, and it’s just for me. I think I’ve probably spent 5 hours tops on it so far.
I’m also making a pizza recipe app, and have given that a lot more time. Remember, it’s addictive. I had a very rough idea for what I wanted, and working with AI Studio, Claude and ChatGPT, it’s become a much bigger project, and potentially something I could release to the world.
And I’ve made a proof of concept app for my day job that would act as a companion to a medical device we make. We’ve wanted to do one since I started (scoping an app was literally how I joined the company), but it wasn’t commercially viable. Maybe now it is.
It’s never been more fun to work on software. And I think a lot of “real” software developers agree. So why the doom and gloom up front then? Why are apps dead?
For all this work I’m doing with an LLM and a chat window, it seems abundantly clear to me that the near future will remove the other hurdles that I’m jumping over. Creating a UI and connecting a backend and publishing to the app store all feel a bit antiquated when you can just say “Hey Claude, use my HealthKit data to show me how much I should be running and come back to this any time I ask” or “Hey Claude, I’ve got pepperoni and pineapple, what pizza should I make?” It doesn’t just respond in text right now – it makes custom tables, documents, artifacts, and they’re getting better every time. They’ll be more functional than the best apps ever were in the blink of an eye. I can even imagine a world where Siri can create an app directly on your phone with just a single prompt.
Chat is the new OS and apps are genuinely whatever you and the AI can imagine. So I guess the new slogan is “There’s a chat for that”. It doesn’t have the same ring to it, but I see no way that it’s not coming. Maybe don’t hold your breath for a Siri version any time soon though.