OS
orchestration
osstack
Multi-agent software development orchestration: routing, tracking, review gates, and delivery rhythm for complex builds.
01 / Workbench
The working system behind the work.
我的项目大多围绕一个问题展开：当一个人拥有越来越多 AI、脚本、文档、浏览器、代码仓库和自动化流程时， 怎样让它们不变成噪音，而是成为一个可调度、可记忆、可迭代的工作系统？
CX
workflow
Codex / Hermes workflows
From idea to code, tests, documents, assets, and launch surfaces, with each step turned into repeatable operational memory.
KB
memory
Local knowledge systems
Experiments in making files, notes, metadata, and context searchable enough to become a useful external mind.
CP
creation
Creative production lines
Design, writing, presentations, publishing, and research pipelines that move from spark to shipped artifact with less friction.
02 / Principles
Tools should make people sharper.
我喜欢的工具不是替人思考，而是让人更清楚地思考。好的系统应该减少摩擦，不减少人的判断； 应该放大创造力，不把创造力包装成模板。
Agency over automation
Automation should serve intent, not steal the steering wheel.
Memory over prompts
Long-lived context matters more than one brilliant instruction.
Tools as companions
A good tool appears naturally inside ordinary work and respects taste.
Shipping as learning
Things begin to teach back only when they can run, fail, and be used.
03 / Selected Projects
Infrastructure, not ornaments.
These projects are not isolated demos. They are pieces of a larger personal operating layer: systems that remember, route, compose, and help ideas survive contact with real work.
osstack
Project-manager layer for agentic software development. It turns scattered AI coding sessions into a managed delivery lifecycle.
Agent Skills Library
Reusable workflows for design, code review, research, documents, spreadsheets, presentations, publishing, and launch work.
Personal Knowledge Infrastructure
Making local files, notes, metadata, and context legible enough for AI collaborators to work with the actual shape of a life.
Creative Production Pipelines
Systems for turning ideas into pages, posts, decks, product surfaces, and visual assets without losing the original taste.
04 / Notes
Notes on building with AI.
我会记录一些还没完全定型的东西：AI 协作、个人工具、产品直觉、系统设计，以及那些在实际构建中慢慢变清楚的问题。
The interesting part is not faster tasks. It is a new way to organize intent, memory, judgment, and action.
Field notes from the edge of personal infrastructure, agentic workflows, and the strange comfort of tools that remember.
Agentic Workflows
How teams of tools move from instruction to verified work.
Personal Infrastructure
Local-first systems, external memory, and the shape of a durable workspace.
Product Taste
Why the best interfaces feel quiet, capable, and alive.
05 / About
A builder of systems that remember.
我对工具的兴趣，最终都落在同一个问题上：怎样让一个人更自由、更清醒、更有能力地完成复杂创造。
我写代码，也设计工作流；我关心系统边界，也关心使用它的人会不会感到被理解。现在我主要在探索 AI agents、 个人知识基础设施、自动化生产线，以及它们如何变成一种新的个人操作系统。
对我来说，AI 最有意思的地方不只是“更快完成任务”，而是它开始改变我们组织意图、记忆、判断和行动的方式。 好的工具应该像一个可靠的合作者：记得上下文，尊重人的品味，能推进事情，也知道什么时候把选择权交还给人。
06 / Contact
Work with me.
如果你也在思考 AI 如何进入真实工作流，而不是停留在演示、聊天框或一次性脚本里，我们可能会聊得很快。