# Value Speak

> 一个人 + AI 做上线的英语练习小程序：每天一句巴菲特、芒格等投资大师的原话，听、读、AI 打分、实战对话，边练英语边吸收价值投资的智慧。

Value Speak（开发期代号 Alpha Speak）是我一个人 + AI 开发并上线的第一个小程序。它把英语练习和投资智慧结合在一起：练习素材来自巴菲特、芒格等投资大师的原话——例如巴菲特致股东信，每句配中文翻译和出处，按主题（如清洁能源）组织。首页叫「Your Snowball」：连续打卡和累计天数像雪球一样滚大。

## 怎么练

1. **听**：先听标准朗读，让声音成为起点
2. **读**：跟读并录音
3. **评**：AI 对发音和流利度打分，给出反馈
4. **聊**：「实战对话」，围绕今天的句子和 AI 对话

## 怎么做出来的

- **工具**：AI Coding 编程工具、微信开发者工具、Python FastAPI、豆包大模型、语音合成、对象存储
- **五步流程**：定最小闭环 → 准备英语语料 → 写后端 → 写前端 → 部署上传

## 上线踩过的 4 个坑

1. 服务端口没有开启
2. 服务器登录密钥
3. AI 打分太慢：最初要等约 40 秒，排查优化后降到约 8 秒（本项目测试）
4. 微信模拟器里无法录音，需要真机调试

> AI 写代码很快，但上线靠的是把每个坑踩明白。完整过程见开发复盘视频。

---

## English

Value Speak (development codename Alpha Speak) is the first mini-program I built and shipped as one person + AI. It combines English practice with investing wisdom: the material is real quotes from Buffett, Munger and other great investors — for example Buffett's shareholder letters — each with a Chinese translation and its source, organised by theme (such as clean energy). The home screen is called “Your Snowball”: your streak and total days grow like a snowball.

## How you practise

1. **Listen**: hear the model reading first
2. **Read**: read aloud and record yourself
3. **Score**: AI scores pronunciation and fluency and gives feedback
4. **Chat**: a practice conversation with AI around today's sentence

## How it was built

- **Tools**: AI coding tools, WeChat DevTools, Python FastAPI, Doubao LLM, text-to-speech, object storage
- **Five steps**: define the minimal loop → prepare the English corpus → backend → frontend → deploy and upload

## Four pitfalls on the way to launch

1. The server port wasn't open
2. Server login keys
3. AI scoring was too slow: about 40 seconds at first, about 8 seconds after debugging (tested in this project)
4. The WeChat simulator can't record audio — you need a real device

> AI writes code fast, but shipping means understanding every pitfall. The full story is in the build video.