我关心的不只是给模型套一层聊天框，而是把真实工作流拆成能够运行、交付和持续迭代的 Agent 系统。
I turn repetitive content and knowledge workflows into conversation-native, local-first tools that produce durable outputs — not just answers.
|
对话式公众号归档与知识资产化引擎，将公开文章转化为离线 HTML、Markdown、元数据和本地媒体资源。
|
面向垂直公众号内容生产的 Agent 工作流，覆盖选题、写作、编辑检查与发布后复盘。
|
本地优先的 AI 产品实验，将传统知识、多角色分析、结构化报告和导出工作流组合成可交互工作台。
|
flowchart LR
A["Conversation"] --> B["Agent Workflow"]
B --> C["Files & Archives"]
B --> D["Knowledge & Decisions"]
B --> E["Practical Automation"]
C --> F["Reusable Assets"]
D --> F
E --> F
| Agent Systems | Content Systems | Knowledge Archiving | Local-first Automation |
|---|---|---|---|
| Tool use, scoped workflows and human checkpoints | Topic discovery, drafting and feedback loops | Offline HTML, structured metadata and durable files | Reversible workflows that run close to the user's data |
CONVERSATION-NATIVE Let natural language control real workflows.
LOCAL-FIRST Keep user data and durable outputs close to the user.
DURABLE OUTPUTS Produce files, indexes and systems — not disposable replies.
HUMAN-IN-THE-LOOP Keep important choices visible, bounded and reversible.
- Shipping conversation-native agents that complete real content and knowledge work
- Turning public content into structured, portable and reusable local assets
- Exploring how vertical expertise becomes a bounded, maintainable Agent workflow
The best way to reach me is through the repositories and issues around the systems I'm building.
