I build practical AI agents, developer tools, and automation products.
我是一名 AI 应用与 Agent 产品实践者，正在学习如何把 AI 原型变成真实用户可以理解、使用和信任的产品。
- Building local-first AI agent and personal knowledge workflows
- Turning AI prototypes into tested, deployable products
- Exploring AI-assisted product QA and multi-agent collaboration
- Contributing narrow, evidence-backed improvements to open-source agent tools
- Teaching beginners to use AI through real, observable tasks
A bilingual, beginner-friendly repository for completing a first pull request with automated validation.
An early public snapshot of a local-first personal AI secretary with tiered memory, search, and daily workflows. The private system has evolved substantially; this repository is being prepared for a cleaner public release.
An experimental Playwright-based tool for testing AI products through real user journeys and producing screenshots, findings, and release-readiness evidence.
A local-first macOS prototype for dispatching windows and workspaces in multi-agent workflows.
An AI-assisted video-processing toolkit for inspecting footage, transcribing speech, cutting clips, and generating subtitles.
An early experiment in collecting and filtering AI news, papers, and open-source projects into a personal daily briefing.
I use a watch-first, maintainer-safe workflow: understand the issue, reproduce it, keep the change narrow, provide evidence, and avoid unnecessary repository noise.
- agent-inspect #72 - merged
- agent-inspect #73 - merged
- mcp-audit #66 - merged
- fathom #157 - merged
- Omnigent #1666 - open; CJK explicit-math rendering with before/after evidence
AI agents Product engineering Enterprise automation
Developer tools Product QA Open-source collaboration
- Be honest about what is tested, what is inferred, and what remains unfinished.
- Prefer small, verifiable deliveries over broad claims.
- Use AI to increase implementation speed without giving up human judgment.
- Build tools that ordinary people and real organizations can actually use.
Most repositories on this account are experiments, learning records, or contribution forks. The six projects above are the intended starting point.