LovisMind

第 01 章

把话说清楚：三个表达框架

表达不清往往不是口才问题，而是思考没有成形。三个场景各配一套可操作框架：提炼循环、承认边界、POP。

How to Be Articulate: Three Frameworks

Unclear speech is usually unclear thinking. One framework for each of three situations: distillation, admitting limits, and the POP response under pressure.

首发于 2026.07.24中 / EN
技术教程communicationarticulationframeworks
语言
字号
目录
篇内目录

这份资料要解决的问题

Vinh Giang 的核心判断只有一句：混乱的语言来自混乱的思考。所以「表达不清楚」很少是嘴的问题，而是想法还没有被整理成能说出口的形状。

他把日常最容易卡住的场景分成三类，各给一套框架：

  1. 你懂这个领域，但说不简单；
  2. 你不懂这个话题，又不想显得一无所知；
  3. 你明明知道答案，但被众人盯着时大脑一片空白。

第三类在职场最常见，也最容易被误解为能力问题。

时间轴

时间内容
0:00三种表达场景的划分
0:53「你没有表达问题」：说话与思考是同一件事
1:16领域专家为什么讲不明白：学过很多，讲过很少
2:51框架一的第一步：capture，把想法写下来
3:52每周写作如何倒逼思考成形
4:24第二步 structure：看见拼图才能重排拼图
5:00第三步 practice：对着镜头真的说一遍
5:57第四步 iterate：回听、删改、从头再来
6:57框架二：遇到不懂的话题怎么办
7:11三步：说出已知、说出未定、问一个推进的问题
8:14现场演示：不懂 AI 时如何仍然显得可靠
8:45「我不知道」什么时候可以直说
10:12框架三：压力下大脑空白的机制
11:17POP 的 P：pause，先停两秒
11:59现场示例：长时间的停顿也可以是自信的
13:10POP 的 O：one thought，只说一个点
13:45POP 的 P：proof，用一个例子撑住它

框架一：提炼循环（capture → structure → practice → iterate）

适用于「我很懂，但讲不清」。

先要接受一个事实：学习和讲解是两套独立的技能。工程师、医生、律师在专业上投入上千小时，讲解可能一次都没练过——所以「我讲不清，是不是我其实没懂」这个自我怀疑通常是错的。真正缺的是提炼（distillation）的训练。

Capture：写下来。 想法在脑子里是流动的、跳跃的，你没法加工一个看不见的东西。写的过程强迫你把感觉变成词，想法于是摆到桌面上，可以被观察。具体两步：先问自己「我最想讲清楚的一件事是什么」，再把「我现在会怎么讲」原样写下来——不用管顺序，第一遍本来就该是乱的。

Structure：重排。 看见拼图碎片之后，才谈得上拼。检查这份草稿：顺序对吗？是不是太长？有没有让人卡住的行话？能不能用一个类比替换一段解释？允许自己加碎片、调顺序，标准是「一个中学生也能听懂」。

Practice：说出来。 这一步区分了想得好的人和讲得好的人。打开手机录像，按新结构讲一遍。说出口会立刻暴露哪些地方不顺——作者写视频脚本时，也是靠读出声来发现逻辑断点的。

Iterate：回听。 看回放，问三个问题：清楚吗？简洁吗？连贯吗？找到最乱的那一段，回到 capture。有时候答案是把整段删掉重写。

四步是一个循环，不是流水线。循环转几轮，复杂的内容会被磨成简单的形状。

框架二：不懂的时候（已知 → 未定 → 提问）

适用于话题转到你不熟悉的领域。

多数人的两个本能都不好：假装懂，或者干巴巴地说「我不知道」。作者给的是三步：

  1. 说出你知道的部分。 哪怕是外围的、局部的理解。
  2. 说出你还在想的部分。 把边界摆在明面上，而不是藏起来。
  3. 问一个能把对话往前推的问题。

他在视频里演示的场景是「AI 会不会取代所有人的工作」：先承认 AI 在规则明确的重复性任务上已经很强，再说自己没想明白的边界在哪里——因为让它做困难任务时经常失败——然后把问题抛回给对方，区分「替代任务」和「替代岗位」。

这套做法的效果是：你没有回答问题，但你的回答依然显得有判断力，因为它展示了你知道自己知识的边界在哪。

关于「我不知道」。 作者的观点是需要一个前提：你此前已经展示过价值和可信度。在这个前提下说「我不知道」，会让人显得可信而不是无能——因为对方知道你不会为了面子编答案。反过来说，如果没有这个前提，这句话就只是暴露空白。

他还提到一个附带好处：允许自己说「我不知道」之后，压力会显著下降。多数人走进会议时默认自己必须对所有问题有答案，这个默认本身就是紧张的来源。

框架三：POP（pause → one thought → proof）

适用于压力下大脑空白的时刻。

机制是这样的：被问到时，所有想法同时涌上来，大脑来不及挑一个起点，于是外部看起来是「卡住了」；为了避免尴尬，你开始边说边找，于是绕圈、跑题。作者用浏览器一次打开几十个标签页来类比——不是没有内容，是同时加载太多。

Pause。 先停两秒再回答。这与直觉相反，因为直觉要求立刻回应以显示专业。但作者的演示说明：只要你停顿得从容，对方读到的不是「慌了」，而是「在认真想」。他甚至会在工作坊里直接请求更多时间：「能不能给我两分钟想想，我会给你一个更有价值的回答。」前提是语气里有把握，而不是歉意。

One thought。 用一句话开头：「我脑子里冒出来的一个想法是……」它的作用是把注意力收束到一个点上，避免分析瘫痪；对听者而言，这句话还传达出「我有很多想法，但我刻意只挑一个讲」。

Proof。 说完这个点，立刻用一个例子撑住它：故事、结果、观察或数据。例子的作用是让那个点落地，也让整段回答有可信度。

可以立刻做的练习

作者给的每周挑战：挑一个你熟悉的主题，完整走一遍框架一。写下来，重排，然后真的对着镜头讲一遍，再回听。他强调多数人停在写作那一步——但没有说出口和回听，简化能力不会长出来。

理解检验

  1. 为什么「我讲不清楚」通常不代表「我没搞懂」？ 参考答案：学习和讲解是两种独立技能。专家在领域内投入大量学习时间，却很少练习讲解，缺的是提炼与口头输出的训练，而不是知识本身。

  2. 框架一的四步分别解决什么问题？ 参考答案：capture 让想法可见；structure 让顺序和颗粒度合理；practice 暴露只有说出口才会发现的断点；iterate 通过回听找出仍需简化的部分，再回到 capture 形成循环。

  3. 不懂某个话题时，三步回应分别是什么？ 参考答案：说出自己确定知道的部分；说出仍在思考、尚无结论的部分；提一个能把对话推进下去的问题。

  4. 直说「我不知道」为什么在这个框架里是安全的？ 参考答案：前提是此前已建立可信度。此时承认不知道会提升信任，因为对方确信你不会编造；反之，在尚未建立可信度时直说，只会暴露空白。

  5. POP 中「pause」为什么反直觉？停顿过久不会显得心虚吗？ 参考答案：直觉要求立刻回答以显得专业，但短促的抢答往往导致绕圈。从容的停顿会被读作认真思考；关键在于停顿的姿态是笃定而非慌张。

  6. 「one thought」对说者和听者各有什么作用？ 参考答案：对说者，收束注意力，避免在多个想法间瘫痪；对听者，传达出「想法很多但刻意只选一个」，显得有取舍。

与这本书其他内容的关系

同书后两章分别谈创业心态与 AI 工具，这一章是它们的前置能力：无论要讲一个商业判断，还是解释一套 AI 工作流，都要先能把想法提炼成别人听得懂的形状。

The problem this material addresses

Vinh Giang's central claim is one sentence: messy words come from messy thinking. Being inarticulate is therefore rarely a problem with the mouth; it is a problem with thoughts that have not yet been shaped into something sayable.

He splits the moments where people freeze into three situations, each with its own framework:

  1. You know the subject, but cannot explain it simply.
  2. You do not know the topic, and do not want to look clueless.
  3. You do know the answer, but the pressure of an audience turns your mind blank.

The third is the most common at work, and the most often mistaken for a competence problem.

Timeline

TimeTopic
0:00Three situations where articulation breaks down
0:53"You don't have a speaking problem": talking and thinking are the same act
1:16Why experts explain badly: lots of learning, little explaining
2:51Framework one, step one: capture — write it down
3:52How a weekly newsletter forces thoughts into shape
4:24Step two, structure: you can only assemble a puzzle you can see
5:00Step three, practice: say it out loud on camera
5:57Step four, iterate: listen back, cut, start again
6:57Framework two: when the topic is unfamiliar
7:11Name what you know, name what is unsettled, ask a question
8:14A live demonstration on not knowing AI well
8:45When it is safe to say "I don't know"
10:12Framework three: the mechanics of blanking under pressure
11:17POP's first P: pause
11:59A live example where a long pause reads as confidence
13:10POP's O: one thought
13:45POP's second P: proof

Framework one: the distillation loop (capture → structure → practice → iterate)

For "I know this well but cannot explain it".

Start with a fact: learning and explaining are separate skills. Engineers, doctors and lawyers spend thousands of hours learning and almost none explaining, so the self-doubt — "if I can't explain it, maybe I don't really understand it" — is usually misplaced. What is missing is the practice of distillation.

Capture: write it down. A thought in your head is fluid and jumpy, and you cannot shape something you cannot see. Writing forces feeling into words and puts the thought on the table where it can be examined. Two steps: ask "what is the one thing I want to explain better?", then write out exactly how you would explain it today. Order does not matter; the first pass is supposed to be messy.

Structure: rearrange. Only once the pieces are visible can you assemble them. Check the draft: is the order right, is it too long, is there jargon someone will trip on, could one analogy replace a paragraph? Add pieces, reorder them, and hold the bar at "a teenager could follow this".

Practice: say it out loud. This is what separates good thinkers from good speakers. Record yourself on a phone and deliver it with the new structure. Speaking exposes the parts that do not flow; Giang says he finds the logical breaks in his own scripts the moment he reads them aloud.

Iterate: listen back. Ask three questions of the recording: was it clear, concise, coherent? Find the messiest section and return to capture. Sometimes the answer is to throw the whole thing out and start over.

The four steps form a loop, not an assembly line. Several turns grind something complex into a simple shape.

Framework two: when you do not know (known → unsettled → question)

For the moment a conversation moves into unfamiliar territory.

Both common instincts fail: pretending to know, or flatly saying "I don't know". Giang's alternative is three moves:

  1. Name what you do know, even if it is peripheral or partial.
  2. Name what you are still working out, putting the boundary in the open rather than hiding it.
  3. Ask a question that moves the conversation forward.

His demonstration is the question of whether AI will take everyone's jobs: acknowledge that AI is already strong on narrow, rule-based, repetitive work; name the open question — where is the limit, given that it often fails on difficult tasks; then hand the question back, asking whether the other person sees AI replacing tasks or replacing people in roles.

The effect: you have not answered the question, yet you still sound thoughtful, because you have shown that you know where your knowledge ends.

On saying "I don't know". Giang's view comes with a precondition: you must already have demonstrated value and credibility earlier. With that established, admitting ignorance makes you human and trustworthy, because the other person knows you will not invent an answer. Without it, the same sentence merely exposes a gap.

He adds a side benefit: once you permit yourself to say it, the pressure drops sharply. Most people walk into a meeting assuming they must have an answer to everything, and that assumption is itself a source of tension.

Framework three: POP (pause → one thought → proof)

For blanking under pressure.

The mechanism: all your thoughts fire at once, the brain cannot pick a starting point, and from the outside you look frozen. To escape the silence you start talking while still searching, and so you circle and ramble. Giang compares it to opening every browser tab at once — the content is there, but too much loads simultaneously.

Pause. Wait a couple of seconds before answering. This runs against instinct, which says a fast reply proves competence, but a rushed reply is what produces the circling. His demonstration shows that a composed pause reads as careful thought, not panic. In one workshop he even asks for more time outright: "Would you mind talking among yourselves for a couple of minutes while I really think about this?" The condition is that the request is delivered with confidence rather than apology.

One thought. Open with "The one thought that comes to mind is…". It narrows attention to a single point and prevents analysis paralysis; for the listener it also signals that you have many thoughts and are deliberately choosing one.

Proof. Immediately back the point with an example — a story, a result, an observation or a statistic. The example grounds the claim and carries the credibility of the answer.

A practice you can start this week

Giang's challenge: pick a topic you know well and run it through framework one end to end. Write it, restructure it, then actually say it on camera, then listen back. He stresses that most people stop at the writing stage — but without speaking and reviewing, the skill of simplification never develops.

Comprehension check

  1. Why does "I can't explain it" usually not mean "I don't understand it"? Answer: Learning and explaining are separate skills. Experts invest heavily in learning and rarely practice explaining, so what is missing is distillation and delivery, not knowledge.

  2. What does each step of framework one accomplish? Answer: Capture makes the thought visible; structure fixes order and granularity; practice exposes breaks that only appear when spoken; iterate finds what still needs simplifying and returns to capture, closing the loop.

  3. What are the three moves when you do not know a topic? Answer: State what you know for certain, state what is still unsettled, then ask a question that moves the conversation forward.

  4. Why is saying "I don't know" safe within this framework? Answer: Because credibility has already been established. At that point the admission increases trust — the listener knows you will not fabricate — whereas before credibility exists it only exposes a gap.

  5. Why is pausing counterintuitive, and why does a long pause not read as insecurity? Answer: Instinct says answer immediately to look competent, but rushed answers circle. A composed pause is read as thinking; the key is that it is delivered with confidence rather than anxiety.

  6. What does "one thought" do for the speaker and for the listener? Answer: For the speaker it narrows attention and prevents paralysis across many competing points; for the listener it signals deliberate selection, which reads as judgement.

Relation to the rest of this book

The other two chapters cover an entrepreneurial mindset and a set of AI working practices. This chapter is the prerequisite for both: whether you are explaining a business judgement or walking someone through an AI workflow, the idea has to be distilled into a shape another person can follow.

来源