MagicFitter.
Photograph what you own and it puts outfits together by matching colours. There is no machine learning anywhere in it, which was the rule I set myself at the start.
What it is
An iOS app that holds your wardrobe. Add something by photographing it or pulling it out of your camera roll, give it a colour, and it lands in the closet.
From there it generates outfits, keeps the ones you save, schedules them against a calendar, and tracks what is in the laundry so it stops offering you a shirt you are not going to wear this week.
Why I made it
Standing in front of a wardrobe at eight in the morning is a bad moment to make a decision. I wanted the decision already made.
How I made it
SwiftUI with Core Data behind it. The matching runs over a fixed palette, beige, denim, olive green and the rest, and decides what goes together by rule rather than by prediction.
It has been going since 2023 and still gets commits. Inside Xcode the project is still called FashionAppOutfit, which tells you roughly how much of the original plan survived.
What I learned
Writing the rules by hand forced me to work out what the rules actually are. A model would have hidden that, and would probably have been right more often, which is a trade I'd make the other way round on a bigger app.
