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「Historical Trails」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

Whenever I’m on a punishingly long flight and feeling the ache of being crunched in a seat for a dozen hours, I soothe myself by opening up the live flight map. The tiny airplane trail is reassuring. We are making progress. We are on track. We will one day be free of this plane. Yes of course I wrote this opening while on a 14-hour flight.

A reassuring in-flight progress tracking map – Source: Magalie L'Abbé When I’m on a journey through a dark swamp of information, I want that same sense of reassurance that flight map gives me. I want to know:

The Pattern

Most interfaces are decent at the first two. We have lots of patterns that tell me where I am right now, like active items, focus states, and progress bars. Even though these are usually lying to us, they still make us feel better. Users will wait longer and be less annoyed if you just give them a fake progress bar. that indicate our current “position” in a process or navigation space.

Sign up screen for Cal.com showing I'm on step 1 of 5 Loading screen for getting a credit score on Monzo We also have good patterns for guiding me towards a known destination. Patterns like steppers and clear copywriting on buttons telling me what’s coming next.

References and Further Reading

Clearly labelled steps on N26 – Source: Mobbin Delivery steps on DoorDash – Source: Mobbin It’s a bit trickier when neither the interface nor I knows the final destination, as happens in information gathering, research projects, and open-ended learning, but we’ll save that problem for another time.

But very few interfaces do a brilliant job of the third need: showing me where I’ve been and how I got here. We can do it at small scales. Patterns like breadcrumbs help us navigate through website or wiki with lots of subpages, but those don’t scale well beyond ~7 steps.

值得单独记下的观察

• Where I’m going (and how long it’s going to take to get there)
• Where I’ve been and how I got here

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