Scene anchors
Swap "users find it inconvenient" for "what a specific person is going through at a specific moment." Abstract needs cannot produce a design; concrete situations can.
- 团队每个人对同一句需求的理解都不一样
- 讨论方案时全靠「我觉得」，没有判断依据
- 让 AI 生成界面，它给的是一个放之四海皆可的通用页面
The problem with abstract descriptions is not inaccuracy — it is that they are compatible with too many approaches. "Enter data faster" could mean voice input, a quick template, or automatic recognition, and each one fits the sentence. You finish building, only to find the user wanted a fourth.
Five elements of an anchor
How to use it to decide
Once the anchor is written, it becomes a ruler. Measure every proposed solution against it.
| 候选方案 | 拿锚点量一量 | 结论 |
|---|---|---|
| 语音输入 | 收摊时周围吵，而且他不好意思对着手机说金额 | 否 |
| 拍照自动识别 | 他手里是零钱不是小票，没东西可拍 | 否 |
| 大按钮快捷金额 | 单手能按，被打断也不丢，符合他的常见面额 | 是 |
| 自动保存草稿 | 直接解决「回头忘了记到哪」这个主卡点 | 是 |
A note for AI
Put the anchor in the prompt and what comes back immediately shifts from a generic template to a solution with real trade-offs.
先记住这个使用场景，后面所有设计决定都要以它为准： 谁：[具体的人，含年龄、设备、熟练度] 什么时候：[时间、地点、环境，以及手上还在忙什么] 想达成：[他要的那个状态，不是他要用的功能] 卡在哪：[具体哪一步出问题，出问题的后果是什么] 现在怎么办：[他目前的替代方案] 基于这个场景，请先告诉我： 1. 这个场景对界面提出了哪些硬约束（字号、点击区域、 单手操作、能否被打断等） 2. 你打算做的主任务是什么，为什么不是别的 3. 有哪些常见做法在这个场景下反而不适用 我确认之后你再写代码。
The third question is the most useful. It forces the AI to make a call like "voice input does not fit this scene," instead of piling on every feature it can think of.