Why We Went Native

Mint started as an Electron app. It was the pragmatic choice — cross-platform compatibility, a familiar web stack, fast prototyping. The prototype worked well. And then we deleted it. ### The Prototype Worked. That Was the Problem. The Electron version had a smooth mind-map canvas, natural drag-and-drop, clean code. As a demo, it was compelling. As a daily-use utility, it was not. Three things made the decision for us: 1. **280 MB on disk.** For a file organizer. The irony of a cleanup tool that wastes more space than most of the files it manages was hard to ignore. 2. **4-second launch time.** A utility app needs to feel instant. When you reach for a tool, it should already be there. Four seconds is long enough to lose context and close the window. 3. **Sandboxed file system access.** Every operation required workarounds — moving files across volumes needed temporary permissions, watching directories for changes was unreliable. We were fighting the OS instead of working with it. ### The Question That Decided It Would we use this app ourselves? Would we keep a 280 MB file organizer running in the background? The answer was no. If we would not use it, we should not ship it. ### What "Going Native" Actually Meant Rewriting in SwiftUI was not just a technology swap. It changed what the app could be: - **0.3-second launch.** Mint opens before you finish the thought of opening it. - **12 MB on disk.** Lighter than most of the files it organizes. - **38 MB RAM at idle.** Leave it running all day and never notice. - **Direct file system access.** No sandboxing workarounds, no permission hacks. Mint works with macOS, not against it. - **Folder-scoped automatic organization.** Lightweight scheduled checks run only for folders the user explicitly configures. - **Apple Vision and native language analysis.** OCR plus on-device content analysis, without native-module sprawl or fragile bridges. These numbers are not abstract benchmarks. They are the difference between an app that feels like part of your Mac and an app that feels bolted on. ### What Transferred Not everything was lost. The mind-map layout algorithm — pure math, force simulation, collision detection — translated directly from TypeScript to Swift. The data model transferred. The design decisions transferred. We had already figured out the visual hierarchy, the interaction patterns, the information architecture. What did not transfer: the rendering layer and the state management. Those had to be rewritten from scratch. ### The Takeaway The Electron prototype was not wasted time. It was a proof of concept that taught us what the product needed to be. The native app was better because we had already made the design mistakes elsewhere. Sometimes the right engineering decision is to throw away working code — not because it is bad, but because "good enough" is not good enough for something people will use every day.