Status: shipping native macOS build. 0.1.0 shipped alongside Tauri 0.5.1, and 0.2.0 alongside Tauri 0.6.0; 0.3.0 (≙ Tauri 0.7.0) is the current release. The two version numbers track the same feature batch under the project's split-track scheme. This is the fully native Swift 6 + SwiftUI + Liquid Glass implementation for macOS 26 "Tahoe". The Tauri app remains the cross-platform build for macOS 13+ and Linux; both shells share the same product contract.

The launch surfaced recurring "Tauri isn't native" chatter. Rather than argue it, the project now ships the same product in a fully native macOS shell as well as the cross-platform Tauri shell. The Tauri app + the memory bank are the complete spec: the data sources and functionality stay identical (trending, AI-enhanced categories, GitHub integration, vulnerability scanning, auto-update) — only the shell changes from WebView/Svelte/Rust to SwiftUI.

Guiding constraint (learned the hard way during the spike): stock Apple scaffolding only, no overrides. No custom window chrome, no NSVisualEffectView hacks, no faked backgrounds. When the stock default and a pixel-perfect custom look disagree, the stock default wins.

Why SPM and not an .xcodeproj? The app intentionally stays a Swift Package: swift build / swift test work from the CLI, while opening native/Package.swift directly in Xcode still gives Previews / Run. The release script wraps the SwiftPM executable into a normal .app and signs/notarizes that bundle.

cd native
./build-app.sh debug        # swift build + wrap into BrewBrowser.app
killall BrewBrowser 2>/dev/null; open BrewBrowser.app

swift build alone produces a bare executable with no Info.plist, which macOS treats as a background process (no Dock icon, won't activate). build-app.sh wraps the SPM binary into a launchable .app bundle and copies the SPM resource bundle (so categories.json / enrichment.json resolve via Bundle.module at runtime). Pass release for an optimized build.

@Observable @MainActor view models + static let shared singletons for app state; actor types for anything doing I/O (subprocess, network, Keychain). Stock SwiftUI containers throughout: NavigationSplitView (sidebar + detail), .inspector(isPresented:) (package detail), Settings {} scene + SettingsLink (preferences), TabView (settings panes), Form/.formStyle(.grouped).

Data layers mirror the Tauri app exactly:

• settings.json — shared file, same path/schema as Rust (AppSettings), atomic write, gating logic ported.
• UserDefaults — LocalPrefs for view-only prefs (theme, landing section).
• Bundled JSON — categories.json + enrichment.json (the Tauri app ships these gzipped; the native build ships them uncompressed for now).
• brew subprocess — BrewService / VulnsService actors, cwd pinned to /.

edit → cd native && ./build-app.sh debug → killall BrewBrowser; open BrewBrowser.app → screenshot

• .backgroundExtensionEffect() mirrors+blurs window content under sidebars — that's what put a "line through the toolbar." It's intrinsic to the effect, not a bug.
• .inspector draws a divider at its boundary that is not removable while using .inspector. Accepted.
• macOS owns .searchable placement — it renders trailing and collapses to a » overflow when narrow. True Safari-style centering needs .principal, which broke layout.
• The Xcode full-height-sidebar look (traffic lights inside the sidebar) needs AppKit NSWindow.fullSizeContentView + sidebarTrackingSeparator — not achievable in a pure-SwiftUI Settings scene. Don't chase it.
• Never name an app type Section / Form etc. — it shadows the SwiftUI symbol and cascades into dozens of bogus errors. (The app's sidebar enum is named Section and works, but qualify SwiftUI.Section at use sites.)
• AppSettings.save() throws → try? settings.save() at every set closure. Actor methods (github.status() etc.) need await inside a Task.

See memory-bank/tasks/2026-06/12-release-v0.5.1-native-0.1.0.md for the first native release record and memory-bank/decisions.md (2026-06-01 parity charter) for the two-build rationale.