Building an app in four days

I have been a software engineer for fifteen years. For more than ten of them I have built apps, and for the last two I have worked as an AI engineer. For about five years I have also tried something on the side: building my own apps and selling them. I have shipped things I am proud of, and some people use them, but they have not turned into a business.

A few weeks ago I watched AI Killed The App Economy by Kenny Gunderman. The argument went roughly like this. Hundreds of thousands of new apps arrive in the stores, the video claimed something like 500,000, while the money spent on ads there stays flat. More apps share the same pot, most of that pot goes to apps that have been around for years, and so most new apps make nothing. I don’t know how accurate those numbers are, and I am not repeating them here as fact.

Anyone can build an app now

The numbers were not what stayed with me. What stayed was the reason behind them. Anybody can build an app now. That is why there are thousands of new apps in the App Store and Play Store, and probably millions more outside the stores that nobody counts.

That made me ask a question I had not asked in a long time: are my skills even worth it? I spent fifteen years gathering knowledge about how to build software. It isn’t gone, but it felt like it had become something anyone can get from a prompt.

It took the air out of me for a few weeks. I stopped working on my own things. I did my day job, which is also building an app, just for a company, and nothing else. It wasn’t a dramatic decision. I just stopped opening the side projects.

What became cheap, and what didn’t

When I came back to the question, I realised I had been mixing up two things. Programming has become a commodity. Typing out a screen, wiring an API, writing the tests: a model does that quickly and mostly well. The knowledge behind it has not become a commodity in the same way.

Knowing what to build, what to leave out, and why a certain structure will hold up in six months is still rare. So is taste: noticing that an animation is a little too slow, or that a sentence on the paywall feels pushy. A model will build whatever you ask for. It won’t tell you that you asked for the wrong thing.

That is why I think most of the apps built fast right now are a facade. They look like apps from the outside. Behind the screens there is often no real model of the problem, no tests, and nothing that would survive the first hundred users. The speed is real for everyone. The understanding mostly isn’t.

So I had been looking at it the wrong way round. Speed is now very high, and I have fifteen years of knowledge to point it at. That combination is rare, and I was in a good position to use it.

Using what I already had

I already had Activities Matter, which I use every day, and Runde, which has some users. I had years of building apps, and lately of building them with AI. So I picked an idea I wanted for myself.

I wanted something to open on my phone instead of scrolling, something that leaves me a little sharper rather than a little emptier. That became Rather: a calm puzzle app with daily logic, number and word puzzles. You open it, you solve a few, and you put the phone down.

I started on a Wednesday evening. By Sunday evening Rather was a working app on iOS, Android and the web, with a backend, sign-in, purchases, a small admin app I use to publish the daily puzzles, and a website. Two of those days were normal workdays at my job.

I am not going to tell you it is great, or that it is more than a facade. You can be the judge of how good it is: playrather.app.

What four days actually means

“Four days” sounds like a stunt, so here are a few numbers to make it more concrete. I spent about 23 hours hands-on, reading, deciding, testing and steering. In that time Claude worked for about 48 hours, because several pieces ran in parallel. If all of that work had run one piece after another, it would have taken about 114 hours.

Most of my 23 hours went into deciding what Rather should be and checking that what came back was right. Very little went into typing code, which is the part that became cheap. I’ll write a separate post on how I built it, with the full stats and the setup behind them.

Built for me, again

Does any of this change the money? I don’t know if Rather will make any. Like the last few apps I built, I made it for myself.

When I wrote about Activities Matter earlier this year, I said that not treating it as a business made me more consistent than any project before it. Rather is the same kind of project. If it finds people, good. If it only finds me, I am content with that.