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「On Opening Essays, Conference Talks, and Jam Jars」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。

Opening Essays and Other Pieces of Narrative Writing

People trying to get better at narrative, non-fiction, opinionated writing like essays, blog posts, and conference talks. Most of this advice won’t necessarily apply to fiction writing, academic writing, or technical documentation writing… but it might help.

Most of the narrative non-fiction opinion pieces I read online do not grip me in the first sentence. People often start pieces somewhere sensible, but boring. Like the chronological start of a story.

Tense, Paradoxical, and Problematic Openings

“I first became interested in data structures 17 years ago while pursuing my undergraduate degree. I was reminded of them last week, and decided to write up some notes on what I’ve learned since then.”

The beginning is almost never the most compelling or important part. It’s just the bit you thought of first, based on your subjective chronology.

Finding Tension, Paradoxes and Problems

Or far worse, people begin at The Beginning Of Time. Invoking paleolithic people is an overplayed way to convince us your topic is cosmically important.

“For tens of thousands of years, humans have relied on their fingers to do things in the world. Fingers have carved rocks, thrown spears, sewn clothes, and held Doritos. Fingers are essential to our survival. And now, I want to show you how to treat your fingers well by reviewing the top 10 mechanical keyboards on the market in 2024.”

Consequential Problems

Another bad way to start a piece of writing is with a statement of what you’re going to write about, followed by a definition.

“This post will outline the most common misconceptions in astronomy. Astronomy is the scientific study of celestial objects such as the stars and planets, and misconceptions are falsely held beliefs by a large number of people.”

Time, Place, Character, and Dramatic Action

“Bad” is obviously subjective and contextual here. I find up-front outlines and definitions dull for narrative non-fiction writing. But they’re standard practice for technical documentation and academic papers. Signposting what you’re going to write about is good, but starting with an exhaustive list of definitions is extremely boring.

To be clear, I make these mistakes too. I am a mediocre opener trying to become a good opener. Opening well isn’t just about snapping up someone’s attention and keeping them reading for a few lines. You can’t write good openings without having good ideas, good arguments, good structure, and good storytelling skills. Everything hangs off that starting point, so it’s worth learning how to nail it.

Starting in the Middle

In trying to become a better opener, I’ve collected explicit advice from a few books on writing: Stephen Pinker’s A Sense of Style , John McPhee’s Draft No.4 , and Joseph Williams’ Problems into PROBLEMS . Alongside implicit advice from studying lots of classic essayists like David Foster Wallace , Rebecca Solnit , David Graeber , and Joan Didion . I’m not going to be winning any originality awards for this lot – hard to pick a more cliché array.

If you dislike the way these authors write, it’s best to check out now.

值得单独记下的观察

• This talk is about X: a compelling problem with consequences for this audience, for which I have some suggested solutions
• We’re going to talk about A, B, C, and D, in that order

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