Maggie Appleton 写设计、人类学和智能体界面时，习惯先画清楚隐喻，再谈工具。把「Programmable Notes」整理成可阅读的中文笔记：问题在哪、界面默认会把人带去哪、落地时该改什么。原站导航、评论回响和广告已去掉。

The Context

Most “note-taking” or “knowledge management” software acts as a passive storage container. You create notes, shuffle them around into folders, add a few tags, and then they sit there. Waiting. Until you consciously remember to go looking for them.

This warehouse approach to notes requires us – the human agent in the system – to do a lot of cognitive labour if we want to use our notes. And using our notes is the entire point of any note-taking enterprise.

The Pattern

Most of us have an ulterior motive beyond writing notes – we want to remember, understand, and synthesise ideas, come to confident conclusions, generate new ideas informed by historical and social events, write coherent essays, or work through difficult problems.

Getting to any of these outcomes requires far more work than typing words into a storage container and retrieving them again. Most note-taking software doesn’t guide us towards any particular workflows or ways of thinking with our notes. We have to develop them ourselves.

References and Further Reading

We have to remember what kind of information we saved in the first place. We have to know what keywords or search methods will allow us to find it again. We have to develop ways to integrate old ideas with new ones. We have to prompt ourselves to expand on our notes, combine them, synthesise them into new realisations, and critique our conclusions.

Thankfully, there’s a long history of people finding ways to make these cognitive tasks easier. We can draw on well-established cultural practices and techniques; we can write Evergreen Notes and build Zettelkastens . We can practice knowledge synthesis and critical thinking . We can use spaced repetition to regularly resurface old ideas. We can write for an engaged and discerning public audience. We can even sign up for expensive master’s degrees that wil

值得单独记下的观察

• Do you actually have a point and have you made it?
• Does your introduction present a compelling problem?
• Are you assuming the audience is thicker than a brick?
• Is this littered with passive voice (to be, is, was, were…) and evasive language?
• Suggest similar search terms
• Brainstorm possible counterarguments
• Help me be more specific.

落地时建议先做的 5 件事

1. 先写清这个工具在强化思考还是在替人思考。
2. 默认交互不要只会谄媚：该追问、该给反例、该要求证据。
3. 智能体界面要露出推理步骤，而不是只给一个光滑答案。
4. 知识发布用可生长的笔记，而不是一次性营销长文。
5. 改产品前先画隐喻：用户以为自己在做什么，系统实际在做什么。

和智能体产品怎么接

龙虾PRO做 OpenClaw 落地时，最该从这类笔记里拿走的是「别把聊天框当唯一界面」。数字员工要能追问、能验证、能把过程摊开，而不是只负责说好话。

本文侧重智能体生命周期与技能边界。若你要动手验证，优先用 Playground 做小任务压测；权限与托管再单独规划，避免「先装一堆再治理」。 相关：Playground · 创建智能体 · 学习中心