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「Making Programming Visual, Spatial, and Learnable」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。

Inventing the End User

Nearly every computer programme created in the last 68 years was made by typing linear characters into a text editor. ASCII symbols arranged from left to right, top to bottom, are the universal medium of code.

Every language from FORTRAN to Python to Haskell works this way.

Draft in Progress

It’s then unsurprising every major IDE IDE stands for intergrated development environment. Essentially a souped up text editor with lots of helpful functionality for programmers baked in. and code editor is designed to optimise working in text; VS Code, Vim, PyCharm, WebStorm, xCode, Atom, Sublime Text are all text-first environments.

This might seem so obvious it’s not worth pointing out. Textual programming is the water we swim in. We express computation logic in written text because that must naturally and obviously the best medium for it. At least, this is what we tell ourselves from the vantage point of people who have only ever lived in a textual programming world.

What is Visual Programming?

What, exactly, might be the alternative to programming in pure text? Am I suggesting we interpretive dance our way through a function that prints out the Fibonacci sequence?

Not exactly. Although I would like to see someone try.

Visual Programming for Learnable Programming

What I want to explore here is the relative lack of visuals, graphics, and spatial design in our current programming environments. While we’ve crafted widely accessible graphical user interfaces (GUIs) for everyone else, programmers remain in the land of typed commands. As the mediators between meatspace and binary logic, we don’t get to click buttons, drag sliders, pan canvases, or select options from dropdown menus while we build. Instead, we type.

There are plenty of good reasons for this – both historical and logical – that we’ll get into. Understanding how we ended up in this purely textual place is a necessary first step if we want to question it.

Programming into the Void

In the early 1980’s a wide new divide opened up. With the rise of personal computing and the arrival of accessible machines like the Macintosh, we created a clear distinction between “programmers” and “end users.” Before that point, the people who designed and wrote computer programmes were primarily the same people who used computer programmes. You couldn’t do much of one without the other.

And all of these people were exclusively using text to talk to their machines. Because it was the only show in town.

Raising the Programming Bar

The aesthetic of bright green blocky text on black terminals isn’t just a fun, retro media trope – our pre-GUI interfaces geuinely looked like this.

Before the 1980’s showed up with its neon leg warmers and big hair, most computers didn’t even have screens . You blindly typed commands onto a keyboard that punched holes into paper.

值得单独记下的观察

• Read the vocabulary of a language by making clear what each keyword and function does
• Follow the flow of what happens at each stage in executing a programme
• See the state as the programme changes variables over time
• Create by reacting to how the programme behaves
• Create by abstracting from the simple to the complex

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。

本文侧重全链路风控方法论。落地时请用自身业务单据做回放验证，不要把示例阈值直接当生产策略。 相关：风控体检 · 方案资源