A note on use of AI tools in the creation of Flyleaf

I started building Flyleaf in Summer of 2023. Back then, the world of software development looked quite different from what it looks now, although changes were already looming on the horizon.

Flyleaf is absolutely a labour of love for me. I wanted to build the read later app that matches my exacting standards for a great, Apple-like, native user interface that is clean, elegant, reachable and follows platform convention. I wanted to build a Mac-assed Mac app, and great counterparts on iPhone and iPad. SwiftUI was just about mature enough by that point to make building a real app with it feasible for a solo developer.

And so I did. I did not use any AI tools for building the app. No AI agent has ever seen the code base. I crafted the app icon manually in the awesome (back then indie app) Pixelmator, and designed the Website with Django. I probably did use ChatGPT to ask questions here and there about SwiftUI issues I was banging my head against or tips on how to build an App Store subscription system that actually works. But the code (and both the brilliance and the bugs therein) is all mine.

This post is not meant as a promise that this will stay like this forever. I have been using Claude Code at my day job for some time with mixed results. I am extremely worried about what these tools are going to do to my skills, to the software industry, and to the world as a whole. But at the same time, I can see that these tools, when used with intention, have the potential to help indie developers like me make their apps better, faster, without compromising the quality of the output.