Skip to content

1. Native per-OS, macOS first, with a Tauri comparison gate

Date: 2026-06-27

Status

Accepted

Context

The app must be highly optimized, stable, and native-feeling. Speed-to-ship and contributor-pool size are explicitly not priorities. The team is willing to do more work for a better result.

Decisions

Native macOS SwiftUI Shell

Decision: Build a fully native macOS app (SwiftUI) through v1.0 (M0–M4).

Rationale: - Best possible native macOS result. - Enforces strict native controls and performance over bloated multi-gigabyte Electron apps.

Tauri Comparison Gate

Decision: Build a Tauri + Web UI spike that reuses the same Rust engine, measure it, and decide whether to pivot to Tauri for cross-platform or continue with native shells per OS (Linux → Windows).

Rationale: - The gate must be judged on "is Tauri close enough to justify cross-platform leverage," not "did Tauri beat native on macOS" (it won't).

Consequences

  • Requires discipline at the engine/shell boundary (ADR-0003) for the gate to be cheap.
  • Alternatives Considered:
  • Tauri-first (rejected: re-introduces webview + web-UI layering, and editor wouldn't feel native).
  • Qt + QScintilla (rejected: GPL-only license, non-native aesthetics — see ADR-0007).
  • Flutter (rejected: its value is identical-UI-everywhere, which is unwanted).