Ivan Trajanovski

Choosing a stack by building in it

Building the same MVP in SwiftUI, Flutter and Kotlin Multiplatform before committing to one.

Period
2026
Scope
3 products · 7 builds
Stack
SwiftUI, Flutter, Kotlin Multiplatform
Status
Lumi live, FORQ in progress

Problem

"Which framework should I use?" is easy to argue about and hard to answer in the abstract. The honest answer depends on the app: how native it needs to feel, whether Android matters yet, and which platform features it leans on.

I wrote about this in Building an app before deciding its tech stack. This is what that looks like in practice.

Approach

When the stack isn't obvious, I build a small spike of the core screen in each candidate, usually in a day each. Each spike starts from the matching starter kit, so the spike is only the new part.

FORQ, a recipe manager. I built spikes in Kotlin Multiplatform, SwiftUI and Flutter. The deciding feature was import: paste a link from a recipe blog or a social post and get a clean, structured recipe. That pipeline is mostly networking and parsing, and it runs cheapest-first:

  1. Look for schema.org Recipe data in the page. It's free, it runs on the device, and it covers most recipe blogs.
  2. For social posts, fetch the caption, where creators actually put the recipe, and let the AI structure it.
  3. For other pages, send the page text to the AI.
  4. Fall back to an editable draft with the title and image, so the user never hits a dead end.

The Flutter version carried this best across both platforms, so it's the one being built out.

ReadMate, a book tracker. An older Flutter app, spiked in Kotlin Multiplatform to see what a native-UI rewrite would feel like.

Lumi, a reading companion. Lumi shipped first in SwiftUI. Once it was live, I rebuilt it in Flutter to reach Android, keeping the same bundle ID, design system, subscription setup and data keys, so existing users carried over. It also kept a native iOS home-screen widget written in Swift.

Outcome

  • Lumi is live on the App Store, with the Flutter rebuild ready for Android
  • FORQ picked Flutter on evidence, not on habit
  • the Kotlin Multiplatform spikes were cheap, useful and deliberately thrown away

A spike you throw away isn't wasted. It's the cheapest way to turn an opinion into a decision.

← Work