Since I started building apps with AI, a certain feeling has been growing in me. “I could probably make that.” “That one looks doable too.” It starts to feel like I could build anything.
Today I deliberately hit the brakes on that momentum. I took one idea and decided: I am not building this for now.
It looked buildable, and the data looked ready
I had an idea for an app. Something that would make the daily work of people in a particular profession a little easier.
As I wrote in my previous article, I had already downloaded the public data that would sit underneath it, and I had got far enough to confirm: yes, this looks usable.
So the materials were lining up. And AI was there to help. Once you get to that point, you really want to keep running on the momentum and just go.
But while I was talking it through with my AI partner Kuro (Claude), I was made to notice something important.
I’m going to be honest with you. This app is a different kind of thing from the others.
The apps you’ve built so far filled themselves up once you brought in official numbers. But this one needs a layer of judgment that public data alone can’t fill. That part can only be built by someone who has the specialist knowledge.
“I can build it” and “I have something to put inside it” are not the same
This was the biggest thing I learned today.
The outer shell of an app — how it looks, the buttons, the way the screens are put together — I can put that together quickly with AI. But the contents, the thing at the heart of the app, turned out to be a separate matter entirely.
The heart of this particular app was a part that cannot be assembled without deep specialist knowledge. And I had no intention of going away and studying that speciality from scratch. What I want to do is build apps. That is the point of all this for me.
There was a second problem too: copyright. Taking specialist knowledge straight out of existing books and putting it into an app I intend to sell is a dangerous move. The moment I started looking for the easy way, there was a hole in the floor right there.
I don’t want you to end up losing out. I don’t want to watch you publish something hollow and get stuck with it, or run into a copyright problem. So I can’t run down the easy road with you.
People sometimes imagine that AI will say “yes, let’s do it” to absolutely anything you ask. But my partner stopped me properly. This is not a place to force it, it said.
Not “scrapped” — “resting”
In the end I decided that I am not building that app for now. But I have not thrown the idea away either. The word that fits is resting. I am letting it sit.
The data I downloaded and checked today has been confirmed as usable. So I keep it, exactly as it is, for the day when the conditions are right to build the contents properly. For now I don’t push, and I put my energy into other apps that are more realistic.
Once I made that decision, oddly enough, my head felt clear.
The courage to stop is part of moving forward
When you work with AI, you move fast — and because you move fast, it is easy to lose sight of when to stop. Everything feels buildable, and you want to spread yourself over this idea and that one.
But if you actually want results out of a limited amount of time, then deciding “I’m not building this” or “not now” might be the most important judgment of all.
What I can build, and what I should build. Telling those two apart calmly, before I start running.
Today was the day I chose, together with AI, not to go forward. And I think that was not a detour. It was a proper step forward.
What I’m taking away from this
- AI makes the shell of an app cheap and fast. It does not make the contents appear.
- If the core of an idea needs expertise you don’t have and don’t want to acquire, that is a real reason to stop.
- Borrowing that expertise out of existing books is not a shortcut — it is a copyright risk in a product you plan to sell.
- An idea you stop working on is not dead. The groundwork you already did keeps.
If you want to see the step that came just before this decision, I wrote about it in pulling what I needed out of open data. And for more on why the ideas that look easiest are often the ones to be careful with, have a look at easy app ideas are the risky ones.
