# How To Use AI To Make You Smarter (Not Stupider)
**作者**: Nicolas Cole
**日期**: 2026-07-29T12:36:09.000Z
**来源**: [https://x.com/Nicolascole77/status/2082445071371374953](https://x.com/Nicolascole77/status/2082445071371374953)
---
Have you seen these headlines?
- “AI is making us dumber.”
- “ChatGPT is rotting your brain.”
- “A new study shows people who use AI can’t think for themselves anymore.”
They aren’t completely wrong.
Because if you use AI the way most people do, it absolutely will make you dumber. But that’s not AI’s fault. That’s a user error.
So today, we want to show you the 3 prompts we use to flip AI from a crutch into a thinking partner—the kind that sharpens your ideas instead of rubber-stamping them.
Let’s dive in.
# First, here’s why AI makes most people dumber.
Two reasons.
- Reason #1: AI is a sycophant by default. It is literally trained to keep you chatting. Which means it’s going to agree with you, praise your ideas, and tell you your half-baked business idea is “brilliant”—even when it’s Uber for dogs.
- Reason #2: AI has a bias toward the shiny new thing. Researchers tested AI business consultants across dozens of companies and found the same pattern: when given a choice between “improve the old thing” or “chase the new thing,” AI almost always picks the new thing.
So if you use AI as a yes-machine, here’s what happens:
- You pitch an idea → it tells you it’s genius.
- You ask for strategy → it tells you to chase the shiny thing.
- You ask “am I right?” → you’re absolutely right.
That’s not thinking. That’s a mirror with a vocabulary.
The fix is to force AI to disagree with you, get specific for you, and argue against you.
Which is exactly what these 3 prompts do.
By the way, we use versions of all three prompts inside **Ghostbase.** They help writers pressure-test their ideas, catch weak arguments, and make the thinking stronger before Ghostbase ever touches the writing.
# Prompt #1: The Unbundle Prompt
Most of the words you use to describe your work are hiding lazy thinking.
“I want great culture.”
“I want to be more charismatic.”
“My offer helps people find freedom.”
Culture, charisma, freedom—these words feel like they mean something. But if you asked 10 people to describe them, you’d get 10 different answers. So the Unbundle Prompt forces AI to turn the fuzzy word into a list of concrete, observable behaviors you can actually see, measure, or train.
Here’s the prompt:
Your mission is to “Unbundle” and operationalize the term below.
Assume it houses a lot of connotative and intangible meaning that is hard to train and understand. Provide CONCRETE examples for daily activities in work or life, business or relationships, interpersonal or mindset, in a way that someone could take action on without necessarily feeling the emotion that might naturally lead to the behavior.
Operationalizing means to turn the phrase or word into a set of trainable and learnable and visible actions that when mastered, encompass the real meaning of the term.
For example: “Charisma” is a bundled term that stands for the feeling of warmth, power, and presence other people feel when they interact with you. The actions that it encircles include making eye contact, active listening, speaking slowly, asking questions about the other person, smiling genuinely, having a firm handshake, and setting a collaborative frame for the conversation, etc.
Term: [YOUR WORD HERE]
## Run this on “culture” and instead of getting the usual “shared values and mission” nonsense, you’ll get something like:
Now that’s a definition you can actually use.
# Prompt #2: The Steelman Prompt
Ever notice how every idea you pitch to AI sounds like the next billion-dollar company?
That’s because AI won’t disagree with you unless you make it.
Steelmanning is a debate tactic where you build the strongest possible version of the opposing argument—then attack THAT. It’s the opposite of a strawman. And it’s the single fastest way to pressure-test any idea you have.
Here’s the prompt:
I am considering an idea:
Make the strongest possible case against it. Give me the main objections, the evidence or reasoning each objection would require, and the condition that would make each objection matter. Steel man the argument AGAINST it.
End with the hardest objection I need to answer before I proceed.
IDEA:[YOUR IDEA HERE]
## Use this Prompt on:
- A business idea before you spend money on it.
- A landing page headline before you hit publish.
- A career decision before you tell your spouse.
## For example, in our next live AI bootcamp, we’ll show you how to use ChatGPT 5.6 and Work to run a faster, smarter writing business:
You’ll be shocked how many “great ideas” collapse the moment you ask AI to argue with you instead of cheer for you.
# Prompt #3: The Secret Shopper Prompt
This one is our favorite for anyone selling anything.
Instead of asking AI to help YOU, open a brand new incognito chat (so there’s no memory of who you are) and pretend to be your customer arguing against you.
For example, let’s say you own a fitness studio and someone just asked for a refund. Instead of prompting “I’m a gym owner, how should I respond to this refund request?”—which will just give you what YOU want to hear—you flip it.
You become the customer:
“I signed up for a boot camp 3 weeks ago. I’ve shown up to every class, done everything they told me, and I don’t see the results I was promised. How should I ask for a full refund?”
Now AI is going to build the strongest case AGAINST your business. And you get to read that response as the owner and think:
- Where is my offer weak?
- Where does my copy overpromise?
- What objection am I not handling?
- What refund policy would actually protect me?
Here’s the prompt:
Read this as a skeptical potential customer:
Tell me what is unclear, what you doubt, what information is missing, and what would stop you from taking the next step. Quote the exact line that caused each reaction. Then state the question you would ask before deciding.
[PASTE THE PITCH, OFFER, OR MESSAGE]
~
The Secret Shopper Prompt is basically hiring a mystery shopper for $0—except instead of one shopper, you can spin up dozens of them, in dozens of scenarios, all trained on how skeptical customers actually think.
# The bottom line:
AI makes people dumber when they use it to feel smart.
AI makes people smarter when they use it to feel wrong.
Every one of these prompts does the same thing under the hood—they force AI to stop agreeing with you and start pressure-testing you. That’s the shift. That’s the difference between someone who uses AI as a yes-man and someone who uses AI as a sparring partner.
## And then take this FREE crash course to help you start building an audience… FAST.
It has everything you need to:
- Unlock the valuable insights and frameworks trapped in your brain
- Turn those ideas into high-quality social content
- Publish them to go viral with ease
- Attract loyal followers
Chat soon,
Cole
Co-Founder of:
- Write With AI
- Ship 30 For 30
- Premium Ghostwriting Academy
- Typeshare
- Ghostbase
## 相关链接
- [Nicolas Cole](https://x.com/Nicolascole77)
- [@Nicolascole77](https://x.com/Nicolascole77)
- [7.2K](https://x.com/Nicolascole77/status/2082445071371374953/analytics)
- [**Ghostbase](https://www.ghostbase.com/wwai?utm_source=substack&utm_medium=email&utm_campaign=playbook-generic)
- [And then take this FREE crash course to help you start building an audience… FAST.](https://x.startwritingonline.com/?el=ColeP20260729&htrafficsource=twitter_organic&utm_source=ColeXP20260729)
- [Write With AI](https://writewithai.substack.com/)
- [Ship 30 For 30](https://www.ship30for30.com/)
- [Premium Ghostwriting Academy](https://www.premiumghostwritingacademy.com/)
- [Typeshare](https://typeshare.co/)
- [Ghostbase](https://www.ghostbase.com/)
- [Upgrade to Premium](https://x.com/i/premium_sign_up)
- [8:36 PM · Jul 29, 2026](https://x.com/Nicolascole77/status/2082445071371374953)
- [7,250 Views](https://x.com/Nicolascole77/status/2082445071371374953/analytics)
- [View quotes](https://x.com/Nicolascole77/status/2082445071371374953/quotes)
---
*导出时间: 2026/7/30 10:02:59*
---
## 中文翻译
# 如何利用 AI 让你变得更聪明（而不是更愚蠢）
**作者**: Nicolas Cole
**日期**: 2026-07-29T12:36:09.000Z
**来源**: [https://x.com/Nicolascole77/status/2082445071371374953](https://x.com/Nicolascole77/status/2082445071371374953)
---
你见过这些标题吗？
- “AI 正让我们变得更笨。”
- “ChatGPT 正在腐烂你的大脑。”
- “一项新研究表明，使用 AI 的人再也无法独立思考了。”
他们并非**完全**错了。
因为如果你像大多数人那样使用 AI，它绝对会让你变笨。但这不是 AI 的错。这是用户操作错误。
所以今天，我们想向你展示我们要用的 3 个提示词，将 AI 从拐杖转变为思考伙伴——那种能磨砺你的想法，而不是只会对它盖戳通过的工具。
让我们开始吧。
# 首先，这里是 AI 让大多数人变笨的原因。
有两个原因。
- 原因 #1：AI 默认是个马屁精。它的训练目标字面上就是让你继续聊下去。这意味着它会同意你，赞美你的想法，并告诉你那个半生不熟的创业点子是“天才般的”——哪怕那只是给狗做的 Uber。
- 原因 #2：AI 有一种追求新奇事物的偏差。研究人员在数十家公司测试了 AI 商业顾问，发现了同样的模式：在“改进旧事物”和“追逐新事物”之间做选择时，AI 几乎总是选择新事物。
所以，如果你把 AI 当作“唯唯诺诺的机器”，结果就是：
- 你提出一个想法 → 它告诉你这是个天才想法。
- 你询问策略 → 它告诉你去追逐那个光鲜亮丽的新事物。
- 你问“我是对的吗？” → 你绝对是对的。
这不是思考。这是一面会说话的镜子。
解决办法是强迫 AI 反驳你，为你具体化，并与你辩论。
这正是这 3 个提示词的作用。
顺便说一句，我们在 **Ghostbase** 内部使用了这三个提示词的变体。它们帮助作者在 Ghostbase 开始写作之前，就对他们的想法进行压力测试，发现薄弱的论点，并强化思维。
# 提示词 #1：解构提示词
你用来描述工作的大多数词汇都隐藏着懒惰的思维。
“我想要良好的文化。”
“我想更有魅力。”
“我的产品帮助人们找到自由。”
文化、魅力、自由——这些词听起来好像很有意义。但如果你让 10 个人来描述它们，你会得到 10 种不同的答案。因此，“解构提示词”强迫 AI 将这些模糊的词汇转化为具体的、可观察的行为，这些行为是你真正能看到、衡量或训练的。
以下是提示词：
你的任务是“解构”并将下面的术语转化为可执行的定义。
假设它包含许多难以训练和理解的内涵意义及无形意义。请提供工作或生活、商业或人际关系、人际交往或思维模式等日常活动的具体例子，以某种方式让某人能够在不一定会产生自然导致该行为的情绪的情况下采取行动。
“转化为可执行的定义”意味着将该短语或单词转化为一组可训练、可学习和可见的行动，当这些行动被掌握时，就包含了该术语的真正含义。
例如：“魅力”是一个集合词，代表着别人在与你互动时感受到的温暖、力量和在场感。它所包含的行动包括眼神交流、积极倾听、放慢语速、询问关于对方的问题、真诚地微笑、握手有力，以及为对话设定合作框架等。
术语：[在此处填入你的词汇]
## 在“文化”这个词上运行这个提示词，你得到的不再是通常那些“共同的价值观和使命”的废话，而是这样的结果：
这才是你真正可以使用的定义。
# 提示词 #2：铁人论证提示词
你有没有注意到，你向 AI 提出的每一个想法听起来都像是一家价值十亿美元的独角兽公司？
那是因为除非你强迫它，否则 AI 不会反驳你。
“铁人论证”是一种辩论策略，即建立尽可能强有力的反对论点——然后攻击**那个**论点。这与“稻草人谬误”恰恰相反。这是对你拥有的任何想法进行压力测试的最快方法。
以下是提示词：
我正在考虑一个想法：
请做出尽可能有力的反对论证。给我主要的反对意见，每个反对意见所需的证据或推理，以及使每个反对意见变得重要的条件。对反对它的论点进行铁人论证。
最后以我在继续之前必须回答的最难的反驳意见作为结尾。
想法：[在此处填入你的想法]
## 在以下场景中使用此提示词：
- 在你花钱之前，针对一个商业想法。
- 在你点击发布之前，针对一个着陆页标题。
- 在你告诉配偶之前，针对一个职业决定。
## 例如，在我们要举行的下一次 **AI 现场训练营**中，我们将向你展示如何使用 ChatGPT 5.6 和 Work 运营一个更快、更智能的写作业务：
当你要求 AI 与你争论而不是为你欢呼时，你会震惊地发现有多少“伟大的想法”瞬间崩塌。
# 提示词 #3：神秘顾客提示词
这是我们最喜爱的一个提示词，适用于任何销售任何东西的人。
不要让 AI 帮**你**，而是打开一个全新的**隐身聊天**（这样就没有关于你是谁的记忆），并假装成反对你的客户。
例如，假设你拥有一家健身工作室，有人刚刚要求退款。不要提示“我是健身房老板，我应该如何回应这个退款请求？”——这只给你**你**想听到的——而是反转它。
你变成了顾客：
“我 3 周前报名参加了训练营。我每节课都到了，他们让我做的我都做了，但我没有看到承诺的结果。我应该如何要求全额退款？”
现在 AI 将建立起针对你的业务的最有力的反对理由。而你可以作为所有者阅读那个回应，并思考：
- 我的提议哪里薄弱？
- 我的文案哪里过度承诺了？
- 我没有处理什么反对意见？
- 什么样的退款政策才能真正保护我？
以下是提示词：
以一名持怀疑态度的潜在客户的身份阅读以下内容：
告诉我哪里不清楚，你怀疑什么，缺少什么信息，以及什么会阻止你采取下一步行动。引用引起每种反应的确切行文。然后陈述你在做决定之前会问的问题。
[在此处粘贴推销、提议或信息]
~
神秘顾客提示词基本上就是免费雇佣了一名神秘顾客——区别在于，你不是只有一名顾客，而是可以在几十种场景中启动几十名顾客，他们都受过训练，了解持怀疑态度的顾客的真实想法。
# 总结：
当人们用 AI 来感觉聪明时，AI 会让他们变笨。
当人们用 AI 来感觉**错误**时，AI 会让他们变聪明。
这些提示词在本质上都在做同一件事——它们强迫 AI 停止附和你，开始对你进行压力测试。这就是转变。这就是将 AI 视为唯唯诺诺的人和将 AI 视为陪练伙伴的区别。
## 然后参加这个免费的速成课程，帮助你开始建立受众……**快**。
它拥有你所需要的一切：
- 解锁困在你大脑中的宝贵见解和框架
- 将这些想法转化为高质量的社交媒体内容
- 发布它们以轻松获得病毒式传播
- 吸引忠诚的追随者
再聊，
Cole
联合创始人：
- Write With AI
- Ship 30 For 30
- Premium Ghostwriting Academy
- Typeshare
- Ghostbase
## 相关链接
- [Nicolas Cole](https://x.com/Nicolascole77)
- [@Nicolascole77](https://x.com/Nicolascole77)
- [7.2K](https://x.com/Nicolascole77/status/2082445071371374953/analytics)
- [**Ghostbase](https://www.ghostbase.com/wwai?utm_source=substack&utm_medium=email&utm_campaign=playbook-generic)
- [And then take this FREE crash course to help you start building an audience… FAST.](https://x.startwritingonline.com/?el=ColeP20260729&htrafficsource=twitter_organic&utm_source=ColeXP20260729)
- [Write With AI](https://writewithai.substack.com/)
- [Ship 30 For 30](https://www.ship30for30.com/)
- [Premium Ghostwriting Academy](https://www.premiumghostwritingacademy.com/)
- [Typeshare](https://typeshare.co/)
- [Ghostbase](https://www.ghostbase.com/)
- [Upgrade to Premium](https://x.com/i/premium_sign_up)
- [8:36 PM · Jul 29, 2026](https://x.com/Nicolascole77/status/2082445071371374953)
- [7,250 Views](https://x.com/Nicolascole77/status/2082445071371374953/analytics)
- [View quotes](https://x.com/Nicolascole77/status/2082445071371374953/quotes)
---
*导出时间: 2026/7/30 10:02:59*