English · 中文
I build and break automated trading systems. One question runs through everything I write: how do you tell when AI-generated code is lying to you?
Four repositories, four versions of that same question:
- A backtest looks great — is it real, or is it overfitting with costs and fills left out?
- An AI-written trading bot's logs are all green — is it actually doing anything?
- The "money flow" figure in an exchange app — what is it actually measuring?
- An AI agent pulls macro data on a schedule — how do you know it isn't making things up?
| Repository | In one line |
|---|---|
| backtest-honesty | 14 checks to run before you trust a positive backtest |
| live-trading-bot-reliability | 7 layers of defense and 8 real pitfalls of AI-written trading bots |
| blackbox-indicator-reverse | A six-step method for reverse-engineering black-box indicator definitions |
| macro-radar | An AI-agent-driven macro sentinel with deterministic delivery |
我用vibe-coding做自动化交易系统，学习研究并分拆它们。 写的所有内容都围绕一个问题：怎么判断 AI 产出的东西是不是在骗你？
四个仓库，就是这同一个问题的四个版本：
- 回测数字很漂亮 —— 那是真的，还是过拟合、成本没算、填价不对？
- AI 写的交易 bot 日志全绿 —— 它到底在干活吗？
- 交易所 App 上的「资金流向」 —— 它量的到底是什么？
- AI agent 定时抓宏观数据 —— 怎么知道它不是编的？
| 仓库 | 一句话 |
|---|---|
| backtest-honesty | 回测出正收益前必查的 14 项 |
| live-trading-bot-reliability | AI 写的交易 bot 的 7 层防御与 8 个真实坑 |
| blackbox-indicator-reverse | 反推黑盒指标口径的六步对账法 |
| macro-radar | AI agent 驱动、但强制确定性投递的宏观哨兵 |
- Email: c405471810@gmail.com