I build practical, local-first tools that make AI workflows simpler, clearer, and easier to trust.

I enjoy turning rough ideas into small, polished products—especially developer utilities that combine thoughtful interfaces with transparent data boundaries.

A liquid-glass traffic-light island for monitoring background Codex CLI sessions on Windows.

• One persistent red/yellow/green signal capsule per CLI session.
• Hover expansion and best-effort Windows Terminal tab navigation.
• Structured local-event monitoring without prompt scraping or telemetry.
• One-command, no-admin Windows installation.

A native Windows floating orb for live Codex quota, daily token trends, and local Skill/Agent analytics.

• Liquid-glass WPF interface with a compact always-on-top orb.
• Local-first usage analytics with explicit attribution boundaries.
• One-command, no-admin Windows installation.
• Built with PowerShell, WPF, and Python.

• Useful automation that removes friction from real work.
• Local-first software with clear privacy boundaries.
• Interfaces that make complex systems feel approachable.
• Reproducible tools that explain what their numbers mean.

我喜欢把复杂的 AI 工作流做成简单、可靠、可解释的小工具。目前主要维护两个 Windows 原生 Codex 项目：用于监控后台 CLI 运行、授权与完成状态的 Codex Signal Island，以及展示周额度、每日 Token 和本机 Skill/Agent 归因的 Codex Quota Orb。

如果你也在做实用型 AI 工具、开发者体验或本地优先软件，欢迎关注我的项目。