Product engineer and designer. I build software for Apple platforms, AI tools, and the infrastructure that keeps them useful in everyday work.
My work spans reading and dictionaries, Apple TV playback, native Swift AI runtimes, and agent-assisted device and network management. I care about the interface as much as the implementation: how a product reads, how it behaves, and how it can be maintained over time.
Website · Projects · Blog · Mastodon
- aDict — A dictionary and reading companion that brings Youdao, local MDict, and StarDict dictionaries into one lookup flow, with history and favorites tied to their original sources.
- eisonAI — An iOS / iPadOS app and Safari extension for exploring an article's structure before reading it in depth. The Safari popup uses Apple Intelligence or your own provider; the app also supports downloaded MLX models.
- Syncnext — An Apple TV media product I maintain across playback, interaction, content sources, and plugins. Public components include SyncnextHybrid for native AVKit integration and audio analysis, plus the plugin protocol and reference plugins.
| Project | What it does |
|---|---|
| localClash | Manages a local Mihomo runtime and exposes routing, configuration, and diagnostics through MCP for AI-assisted operation. localclash-luci provides the OpenWrt LuCI integration. |
| AIReasoningCore | A native Swift AI runtime for iOS and macOS, with an AnyLanguageModel adapter, conversation persistence, and app-integrated tools. |
| pi-ai-swift | The native Swift provider layer: model catalogs, provider requests, streaming, and authentication behind a shared interface. |
| WindowsAgent | A Go agent for Windows Graphics Capture screenshots and bounded observation tasks, with extensible Starlark packages. |
| codex-workflow-skills | Reusable skills for context recovery, evidence-led debugging, model consultation, and content handoff. |
Other work includes threadBridge, which connects Telegram threads to Codex sessions, and HLN Machine, a public reference pipeline for producing AI-assisted video shorts with operator-supplied models and assets.
I start with a real workflow, then follow it through the interface, data, runtime, and release process. When something breaks, I use logs, device behavior, and reproducible cases to find the cause. I keep responsibilities explicit and turn useful lessons into tools, documentation, or writing.
RonnieCC brings the projects and writing together. This profile links to selected public code and product pages; some apps have private source while their supporting components are public.
我是 Ronnie，做產品工程與設計，主要開發 Apple 平台軟件，也做 AI 工具和日常使用的基礎設施。
從 aDict 的查詞與閱讀、eisonAI 的文章結構，到 Syncnext 的 Apple TV 播放體驗，我關心產品是否清楚、好用，也關心它在真實裝置上能否穩定運作、持續維護。近期的工程工作包括 Swift AI runtime、localClash 網絡管理，以及讓 AI 能操作和檢查裝置的工具。
我會把開發、除錯和發布過程中值得留下的經驗，整理成程式、工作流程與文章。