유대선
프로젝트로
·기술 회고·1

Should DalkkakAI eventually move to Swift? Walked the blueprint to decide

Caught myself flirting with a Swift rewrite for 'Apple-grade UX.' Walked every layer of the BLUEPRINT and concluded the current product as designed has no part that *requires* Swift. Tauri stays unless the blueprint itself pivots.

The temptation

The pull was emotional, not technical: native macOS apps feel better than webview-rendered ones, the Apple-26 era APIs (Liquid Glass, Foundation Models, Shortcuts) are alluring, and "I should know Swift better" is a tax I keep wanting to pay.

The check

Walked every layer in docs/BLUEPRINT.md §5:

LayerRequires Swift?Notes
1. Multi-pane terminal multiplexerNoxterm.js on WebKit is already shipped
2. Multi-session orchestrationNoRust handles the FS and process tree fine
3. Claude Code augmentation (parse output)NoPure string parsing, stack-agnostic
4. Viz layer (Mermaid, React Flow)No — reversedReact Flow is web-only. Swift would cost here
5. Skills / templatesNoStatic + conditional
6. Cross-startup dashboardNoDisplay only
6b. Advisor agent (cloud)NoCloud backend, stack-agnostic

The blueprint also has §6 "Why Tauri (not Electron, not Swift)", which explicitly cites "native Swift would lock out Linux/Windows" — Swift is already an excluded option in writing.

§9 Out of scope excludes mobile. Even the strongest pure-Swift argument (iOS extension) is off the table.

Decision

Tauri stays. Swift is justified only if the blueprint itself pivots (Mac App Store-only, iCloud sync, deep Shortcuts integration, Apple Intelligence on-device, or dropping cross-platform). None of those are in scope today.

Pattern

When you feel like rewriting the runtime, walk the spec layer by layer and ask whether each layer requires the new runtime. "It would be nicer" is not a requirement. The blueprint section that already said no is usually the one you forgot you wrote.