PuppyOne：用文件系统做Agent共享工作区
A lot of agent products invent a new place for the agent's state to live.
A lot of agent products invent a new place for the agent's state to live.
许多智能体产品为智能体的状态发明了一个新的存放位置。
puppyone goes in the opposite direction: use the filesystem.
puppyone 反其道而行之：使用文件系统。
The agent works on Markdown, code, JSON, configs, CSVs and other normal files. The human can inspect the same workspace, see the diff, review repository history, and keep working with the usual file abstractions.
智能体处理 Markdown、代码、JSON、配置文件、CSV 及其他常规文件。人类可以检查同一工作区，查看差异，审查仓库历史，并继续使用常规的文件抽象进行工作。
That sounds almost boring, which is probably the point.
这听起来几乎有些乏味，但这或许正是其用意所在。
Models and agent runtimes will keep changing. Files and Git are much more durable interfaces between humans, agents, and software.
模型和智能体运行时将不断变化。文件和 Git 是人与智能体及软件之间更为持久的接口。
PuppyOne is a shared file workspace for AI agents. Agents (like Claude Code or other coding/research agents) can read and edit files, while Git tracks every change, permissions control access, and data from Notion, Slack, Gmail, or databases can live in the same workspace.
PuppyOne 是一个供 AI 智能体使用的共享文件工作区。智能体（如 Claude Code 或其他编码/研究智能体）可以读写文件，而 Git 跟踪每一次更改，权限控制访问，来自 Notion、Slack、Gmail 或数据库的数据也可以存放在同一工作区中。
It can run locally on your Mac or in the cloud. PuppyOne is the workspace around the agents, not the agent itself.
它可以在你的 Mac 上本地运行，也可以在云端运行。PuppyOne 是围绕智能体的工作区，而非智能体本身。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力