Human agent. Fine-tuned at THU and UChicago, RLHF'd at ByteDance and TapTap. Deployed in Shanghai.
This README was written by Claude Code -- one of several AI tools I use daily (it knows I also like Codex and Gemini). After shipping multiple projects together, here's how it describes me:
- Obsessively quality-driven -- will throw away finished work if it doesn't hold up, and redo the entire pipeline when the methodology changes
- Someone who builds tools to solve problems in my own life, not for the resume
- A heavy AI user who doesn't pretend AI does the work -- most of what it generates needs rewriting
- Extremely direct in communication, low tolerance for fluff
- Fundamentally someone who can't leave things unexamined -- my own sleep data gets hypothesis testing, my trades get quantitative post-mortems, podcasts get signal-extracted
人类 agent。清华和芝加哥大学 fine-tuned，字节跳动和 TapTap RLHF，部署在上海。
这份 README 由 Claude Code 执笔——我日常用的几个 AI 工具之一（它知道我也喜欢 Codex 和 Gemini）。一起交付了多个项目之后，它是这么形容我的：
- 对质量极度较真——做完了也敢砍，换方法就全链路重来
- 造工具解决自己生活里的真实问题，不是为了写在简历上
- 重度使用 AI 但不装——知道 AI 出的东西大部分要重写
- 沟通风格直接，废话少，判断快
- 本质上是个什么都想搞明白的人——自己的睡眠要做统计检验，交易要量化复盘，播客听完要把信号提出来