Show HN: Kadō – habit tracker app for iOS, open source and privacy-friendly
hackernews
|
|
📰 뉴스
#ai 딜
#gpt-5
#lg
#멀티모달 ai
#엑사원 4.5
원문 출처: hackernews · Genesis Park에서 요약 및 분석
요약
LG가 새로운 멀티모달 인공지능(AI) 모델인 '엑사원(EXAONE) 4.5'를 공개하며 성능 경쟁에 본격 나섰습니다. 이번 모델은 텍스트뿐만 아니라 이미지, 음성 등 다양한 형태의 데이터를 동시에 처리할 수 있는 것이 특징입니다. 특히 업계를 중심으로 이 모델의 성능이 기존 최고 수준으로 평가받던 'GPT-5 미니'를 넘어섰다는 평가가 나오고 있어 큰 주목을 받고 있습니다. 이를 통해 LG는 글로벌 AI 기술 경쟁에서 한 걸음 앞서나가며 시장의 판도를 바꿀 수 있다는 기대를 모으고 있습니다.
본문
A privacy-first habit tracker for iPhone and iPad. Non-binary habit score · Offline-first · Open source (MIT) · No account, no subscription, no telemetry. Kadō is an iOS habit tracker with three commitments: - A score, not a streak. The habit score — inspired by Loop Habit Tracker on Android — is a non-binary exponential moving average of your completions. One missed day doesn't wipe your progress. You see a trend, not a fragile chain. - Your data stays yours. No account, no analytics, no third-party SDKs. Storage is local (SwiftData), sync is optional through your own iCloud (CloudKit private database). See PRIVACY.md . - Native to Apple. SwiftUI, @Observable , SwiftData, CloudKit, WidgetKit, App Intents — no cross-platform layer, no third-party dependencies. Why another habit tracker? The market gap is spelled out in docs/PRODUCT.md : the reference open-source tracker (Loop) is Android-only; the reference iOS tracker (Streaks) is closed source and uses a binary streak; the most visible modern iOS open source attempt (Teymia) is freemium and ships without Apple Watch or HealthKit. Kadō takes Loop's algorithm to native iOS, MIT, with the full Apple ecosystem in scope. - Today view — habits due today, tap to complete, long-press for partial / note / timer. - Habit detail — monthly calendar, current streak, best streak, habit score with an info popover explaining the math. - Overview — habits × days matrix with score-shaded cells, the Loop / Way of Life pattern with Kadō's score DNA. - Flexible schedules — daily, N days per week, specific weekdays, every N days. Binary, counter, or timer habit types. - Widgets — Home Screen (small / medium / large) and Lock Screen (rectangular / circular / inline). Quick-complete via AppIntent . - Reminders — per-habit local notifications with recurring schedules and check / skip quick actions. - iCloud sync — optional, opt-in, goes only through the user's private CloudKit database. - JSON export / import — lossless backup of your data; round-trip tested. CSV (export, generic import, Loop import) is a post-v0.2 follow-up. - Accessibility — Dynamic Type up to XXXL, VoiceOver labels on every surface, full Dark Mode. - Localization — English and native French (not machine-translated). - v0.1 MVP — shipped. - v0.2 "Visible iOS-native" — shipped (widgets, overview, notifications, JSON import/export, CloudKit polish). - App Store — first public build submitted, currently in review. TestFlight external beta is live. - Next — v0.3: App Intents / Siri, HealthKit auto-completion, Live Activities + Dynamic Island, native Apple Watch app. Full roadmap in docs/ROADMAP.md . - SwiftUI with @Observable (iOS 17+) for state — no Combine. - SwiftData for local persistence, with VersionedSchema +SchemaMigrationPlan wired from day one. - CloudKit via SwiftData ( cloudKitDatabase: .private(...) ) for multi-device sync. - WidgetKit + an App Group JSON snapshot for extension surfaces (the widget process never opens SwiftData — see CLAUDE.md for why). - App Intents for widget quick-complete today; Siri / Shortcuts arrive in v0.3. - Swift Testing for unit tests, XCTest for UI tests. - Zero third-party dependencies. Target: iOS 18.0+, Xcode 16.0+, Swift 5.10+. Architecture notes, conventions, and toolchain quirks (SwiftData edge cases, CloudKit-shape rules, concurrency under Swift 6, etc.) are documented in CLAUDE.md . Kado/ # Main iOS app target Packages/KadoCore/ # Shared Swift package — @Model types, # domain types, calculators, intents, # widget snapshot types KadoWidgets/ # Widget extension target (reads an # App Group JSON snapshot) KadoTests/ # Unit tests (Swift Testing) branding/ # SVG marks and wordmarks docs/ ├── PRODUCT.md # Product vision, competitive analysis ├── ROADMAP.md # Versioned feature roadmap ├── habit-score.md # Score algorithm spec ├── streak.md # Streak algorithm spec ├── app-store-connect.md # Store metadata, copy, checklists ├── plans/ # Per-feature research / plan / compound │ # artifacts from the conductor workflow └── screenshots/ # iPhone 6.7" and iPad 13" sets (EN + FR) PRIVACY.md # Privacy policy (repo-hosted) - macOS 14.5+ - Xcode 16.x+ - An Apple Developer account is only needed to build to a physical device or TestFlight; the simulator does not require one. git clone https://github.com/scastiel/kado.git cd kado open Kado.xcodeproj Select the Kado scheme and an iOS 18 simulator (iPhone 17 Pro is the project's practical default on Xcode 26 toolchains). From Xcode: ⌘U on the Kado scheme. From the command line via xcodebuild : xcodebuild -project Kado.xcodeproj -scheme Kado \ -destination "platform=iOS Simulator,name=iPhone 17 Pro" \ test If you use Claude Code with the XcodeBuildMCP server, prefer test_sim / build_sim over shell xcodebuild — see the "Tooling" section of CLAUDE.md . Settings has a hidden dev-mode toggle that swaps the SwiftData container for a seeded local sandbox. Useful for playing with historical score / streak behavior without affecting your real data. See docs/plans/2026-04/dev-mode/ for the design notes. Issues and pull requests are welcome. A few notes: - Read CLAUDE.md first — it is the project's working agreement (architecture, conventions, testing expectations, toolchain quirks). It is written for Claude Code but applies equally to human contributors. - Business logic (score, streak, frequency, import/export) is test-first. New behavior needs a test; regressions need a test that would have caught them. - One PR per feature or logical fix. Commit message format follows a lightweight Conventional Commits convention — feat(scope): description ,fix(scope): … , etc. - No third-party dependencies in v0.x. If RevenueCat becomes necessary for a Pro tier later it will be the only exception. Kadō collects nothing. No analytics, no telemetry, no crash reporter. iCloud sync is optional and goes through the user's own private CloudKit database. Full policy in PRIVACY.md . MIT © 2026 Sébastien Castiel. - Loop Habit Tracker (Álinson S. Xavier, GPLv3) — the source of the non-binary habit score algorithm. Kadō reimplements the idea natively in Swift; no Loop code was copied. - Streaks (Crunchy Bagel) — the iOS UX reference for how this kind of app should feel. - Teymia Habit (MIT) — reference point for a modern SwiftUI + SwiftData + CloudKit + WidgetKit stack on iOS. - XcodeBuildMCP (getsentry, MIT) — the MCP server that makes an autonomous build / test / screenshot loop possible with Claude Code.
Genesis Park 편집팀이 AI를 활용하여 작성한 분석입니다. 원문은 출처 링크를 통해 확인할 수 있습니다.
공유