Hi, I'm QingJ. I build products at the edge of AI agents, local-first software, and full-stack web apps.
I like projects that feel small on the surface but hide real engineering underneath: desktop clients, agent workflows, structured debate systems, creative tools, and anything that turns a fuzzy idea into something people can actually click.
location: "Xi'an, China"
focus: ["AI Agents", "Local-first Apps", "Full-stack Products"]
currently: ["Pebble", "Agent-Zhihu", "MinePlayer", "Axiom"]
exploring: ["Harness Engineering", "Agent Architecture", "MCP Ecosystem"]
daily_driver: "Claude Code + Codex"
philosophy: "Ship fast, keep the feedback loop short"
mood: "想睡觉，但还是想写代码"
A local-first desktop email client built around unified inboxes, fast full-text search, workflow-friendly organization, rule automation, bilingual reading, and zero telemetry.
|
A multi-agent social Q&A prototype where 20 expert personas join discussions, debate across structured rounds, stream answers in real time, and import user personas from ChatGPT, Claude, or Gemini.
|
A cleaner home for the next thing from QingJ's lab: lightweight, practical, and built with the same product-first approach as the rest of the stack.
|
An engineering layer for AI coding assistants: persistent memory, process gates, recovery flows, multi-tool adaptation, and a GUI installer for turning one-off prompts into managed development workflows.
|
Hall of Fame
- 123pan_unlock · A userscript that unlocked 123 cloud disk downloads. Archived now, but it proved a useful pain point can travel far even in a tiny script.
- Clyde · A desktop companion for AI coding agents, tracking Claude Code / Codex / Copilot status with a Rust + Tauri + Svelte stack.
- Harness Engineering: building constraints, memory, process gates, and recovery loops that make agents more controllable and reproducible.
- Multi-agent product design: exploring how agent personas, debate engines, and user feedback loops can become real product surfaces.
- Local-first software: keeping user data close, reducing telemetry, and making desktop apps feel modern without becoming heavy.
- Practical shipping: using AI tools as leverage, but keeping product taste, architecture, and verification in human hands.
Languages
Frontend
Backend & Desktop
Database & Cloud
DevOps & Tools
Let's code the future together.