如何最终信任 Claude 的建议（使用 Karpathy 的 LLM Council method）

Claude just tells you what you want to hear. Every time. You can’t trust it.

Claude 只是告诉你想听的话。每次都是。你不能信任它。

So I built a skill that forces 5 AI advisors to argue about your question, anonymously review each other's work, and hand you a verdict you can actually trust.

所以我构建了一个技能，它迫使 5 个 AI 顾问就你的问题争论，匿名审查彼此的工作，并给你一个你真正可以信任的裁决。

In this article I'll show you exactly how to set it up.

在本文中，我将向您展示如何确切地设置它。

Here's the method and the exact skill I built, so you can run it yourself:

这是我构建的方法和确切的技能，您可以自己运行它：

P.S. If you want more AI workflows like this one delivered to your inbox every week, join 35k readers getting them free: aisolo.beehiiv.com/subscribe

P.S. 如果你想每周收到更多像这样的 AI 工作流送到你的收件箱，加入 35k 读者免费获取：aisolo.beehiiv.com/subscribe

Why one answer is the problem

为什么一个答案就是问题

Claude is incredibly agreeable.

Claude 极其顺从。

Ask it "should I launch this product?" and it'll find 5 reasons why you should.

问它 "我应该推出这个产品吗？" 它会找出5个你应该推出的理由。

Then ask it "is this product a bad idea?" and it'll find 5 reasons why it is.

然后问它“这个产品是个坏主意吗？”，它就会找出 5 个它是坏主意的理由。

Same product, different framing, completely opposite answers.

相同的产品，不同的框架，完全相反的答案。

You've probably noticed this.

你可能已经注意到了这一点。

You ask a question, get an answer that sounds smart, feel good, and move on.

你问一个问题，得到一个听起来聪明、感觉良好的答案，然后就继续前进了。

But that answer was shaped by how you asked it.

但是那个答案是由你提问的方式塑造的。

Your assumptions
你的假设
Your framing
你的框架
Your emotional lean
你的情感倾向

Claude picked up on all of it and told you what you wanted to hear.

Claude 捕捉到了所有这些，并告诉你想听的话。

That's fine for writing emails.

这对于写电子邮件来说还可以。

It's dangerous for making decisions.

这在做决策时是危险的。

The council breaks this by forcing 5 different thinking styles onto the same question.

这个理事会通过强制将5种不同的思考风格应用于同一个问题来打破这一点。

They can't all agree with you because they're not all looking at it from your angle.

它们不可能都同意你的观点，因为它们并非都从你的角度看待它。

Where this comes from

这来自哪里

Andrej Karpathy (co-founder of OpenAI + former head of AI at Tesla) built something called the LLM Council.

Andrej Karpathy（OpenAI 联合创始人 + 前 Tesla AI 负责人）创建了名为 the LLM Council 的东西。

The idea: in...