Agent systems · full-stack delivery · clear writing
I turn ambiguous work into tools, workflows, and products that can survive contact with real users.
我关心的不是 AI 看起来多聪明，而是它能不能把一件真实的事往前推进。
PERSONAL SITE · X / @luochenkafei · ALL REPOSITORIES
| 01 / AGENT WORKFLOWS Skills, harnesses, evaluation gates, and delivery habits that make coding agents useful beyond a demo. |
02 / REAL PRODUCTS Full-stack systems for orders, operations, documents, research, and the messy edges between roles. |
03 / PUBLIC THINKING Writing, courses, and visual explanations that turn hard-won project judgment into reusable knowledge. |
career-ops is the clearest public proof: an AI job-search system where I contribute across safety, compensation integrity, scanning reliability, PDF generation, system updates, and cross-CLI workflows. The project is now above 69K stars.
- Building agents that help with real work instead of performing intelligence
- Turning Codex workflows into reusable skills and production habits
- Shipping small experiments, then checking them against users, systems, and reality
Codex TypeScript JavaScript React Vue Python FastAPI Flask Shell
Ideas are cheap. Useful systems are not.
想法很便宜，真正有用的系统不是。