Getting AI to Write an App

Yesterday I went to a talk where Claude and Cursor were used to write a Flutter app for Android. The idea is quite simple. “Claude help me write a to-do app and a shopping list app” before following the instructions produced by AI to write the app.

When I watched this I thought “This reminds me of the laravel tutorial that I followed at least two or three times. If you find a tutorial you can write the code by hand by copying it line by line, and function by function.

With time and persistence you end up with an app that you have “written” by copying code and re-typing it manually.

In yesterday’s demonstration it was more simple than that. This is the app I want to make, tell me how to. After following ten or more steps the app was ready to run in Chrome but it took a little more trial and error to get it to work on the phone. The talk ran out of time before the app was running on the phone.

Of Two Minds

I’m in two minds about seeing such a demonstration. On the one hand I see that AI can write a tutorial for you to follow instructions to build the app that you want. On the other hand it requires a lot of energy, GPU processing, and water for cooling. In my eyes you could easily explore the Free and Open Source movement for an app that is close to what you want to do, and then tweak it.

In theory it’s fantastic that we can get AI to write apps for us because it gives us tremendous flexibility to do what we want to do. On the other hand I think that by creating too many apps that do the same thing as each other we’re wasting energy, water and time.

The Opportunity

For me the opportunity lies not in AI writing an entire app and helping us to debug why it isn’t working. For me the opportunity is in tweaking something that already exists and adding more flexibility.

Imagine that you have a wordpress blog, and that you want to migrate to Ghost. You might find the right tool but it may be old and out of date. For it to be usable you need to update some libraries and functions.

You run the app. You see the error, you ask how to fix it. Eventually, with time the errors are fixed and the app works as you intend it to. A stale project becomes active again. The advantage of this approach is that what you’re asking AI to help you with is minimal. Writing an entire app takes plenty of tokens. Identifying errors and writing simple functions takes fewer resources.

Opportunity

If AI can be used to write an entire application then it can definitely help with adding smaller functionality to a pre-existing app. I have already experimented with writing apps function by function for a website. Now, with the progress that has occurred over the last two years new opportunities present themselves.

Precursors

Years ago we had FrontPage 2000, Dreamweaver and other apps to help us create websites. We could also browse the web to find snippets of code that added the functionality we wanted. Now, instead of searching the web we can ask AI and it will provide us with help. It can also help with debugging.

I don’t think we should forget about tutorials, guides and open source projects that do what we want to do.

And Finally

What the talk made me think about, last night, is that AI can be used to write a course. We can tell it, I want to make this app, with that framework, and I want to be guided through the process and AI will. When you get an error message, instead of having to debug you can copy and paste the error and relevant info. It will help you problem solve.

And finally just because AI can write an app for you, doesn’t meant you should. It still takes time and resources.