How to Distill Any Expert into an AI Skill: Elon Musk SOP Guide

Distillation vs. Simple Copying

Distillation extracts the decision logic behind a person’s statements and turns it into a reusable tool, whereas copying merely records the raw quotes.

What to Distill

Core recurring propositions (underlying logic).

How those propositions are applied in concrete situations.

Counter‑intuitive views that oppose common intuition.

Open problems the person has not solved (opportunity space).

Step 1 – Build the Archive

Create an Obsidian vault with the following layout:

📁 20_People/
└── Elon_Musk/
    ├── 00_档案.md          ← profile card
    ├── Sources/            ← one file per raw source
    │   ├── 2026-传记-Elon-Musk.md
    │   ├── 2025-03-推特-火星殖民观点.md
    │   ├── 2025-TED-X演讲实录.md
    │   └── ...
    └── Distilled/          ← output of distillation
        ├── 01_核心主张.md
        ├── 02_决策方法论.md
        ├── 03_反直觉观点.md
        ├── 04_未解之题.md
        └── 05_我的选题映射.md

In 00_档案.md record why you are distilling, start date, source count, and links to the distilled files.

Step 2 – Collect Materials

Gather at least 20 high‑quality sources before running any AI prompts. Prioritise:

First‑hand written material (e.g., Walter Isaacson biography, Lex Fridman podcasts, TED talks, long Twitter threads).

Secondary sources that contain direct quotations (early biographies, Tesla/SpaceX letters, analyst podcasts).

Avoid low‑fidelity content such as generic media interviews or short video clips.

Insufficient material leads the model to hallucinate.

Step 3 – AI Distillation (6‑Step Prompt Chain)

Run the prompts in order, writing each result immediately to the corresponding file in Distilled/.

Step 1 – Core Propositions (Top 10)

以下是我 Obsidian 里关于 Elon Musk 的原始 Sources 内容（见附件/粘贴内容）。
请仅基于这些来源，输出他的核心主张 TOP 10。
每条必须：
1. 一句话总结主张
2. 附 2‑3 个来源引用（注明文件名或段落）
3. 指出这条主张和普通人常见做法的差异是什么
禁止脑补他可能说过的话；只输出来源里有明确依据的主张。

Step 2 – Decision Steps Behind Propositions

基于 Step 1 输出的主张，请针对其中 3 条（你觉得最可迁移的），描述他在具体情况下是怎么"用"这条主张的——即：触发条件是什么？他做了什么动作？得到了什么结果？同样只基于 Sources，不要脑补。

Step 3 – Repeated Phrases & Keywords

请统计 Sources 中，Elon Musk 反复使用（超过 3 次）的词汇、短语或句型。每个词汇/短语附上：出现场景 + 他用它想表达的核心意思。

Step 4 – Counter‑Intuitive Views

请从 Sources 中找出 5 条"违反大多数人直觉"的观点或决策。即：一般人看到他的做法会说"这怎么可能有效"，但后来被证明有效（或至少他有充分理由这么认为）。每条附来源 + 他的逻辑链。

Step 5 – Unsolved Problems

请从 Sources 中找出：
1. 他承认自己也不确定的领域（含糊表达 / 多次改口）
2. 他避而不谈或一笔带过的话题
3. 他的方法论在哪些场景下"按他自己的描述"是不适用的
这些"空白"最有价值，是你作为创作者可以继续深挖的地方。

Step 6 – Map to Your Content Topics

基于以上 5 步的蒸馏结果，结合我的账号定位（如：面向创业者的 AI 工作流），帮我生成 10 个可以用马斯克的思维框架来写的选题。
每个选题给出：
- 标题方向（引发好奇）
- 核心角度（马斯克的哪条逻辑）
- 目标读者痛点（他们为什么会点开）

Step 4 – Add Your Own Voice

After the AI outputs, write three personal sections in each distilled file:

Agreement & Supplement: Confirm a proposition with your own experience.

Disagreement & Doubt: Note limits or boundary conditions.

Further Exploration: Identify which unsolved problem you want to investigate next.

Common Pitfalls

Too few sources: AI will hallucinate. Collect ≥20 files first.

Not writing results to files: Knowledge stays only in chat logs and is lost.

Missing personal commentary: The output becomes a quote collection rather than a usable framework.

How to Use the Distilled Asset

Manual Lookup

Open 00_档案.md and jump to the relevant file in Distilled/ when you need a proposition, decision step, counter‑intuitive view, or topic idea.

Dialog‑Based Invocation

Paste one or two distilled files into a Claude/Cursor session and ask for a concrete angle, a cited proposition, or a hook, e.g.:

以下是我从 Elon Musk 的公开内容中蒸馏出的档案（主张 + 方法论 + 反直觉观点）。
现在我要写一篇关于「[你的选题]」的文章。
请基于档案给我：
1. 一个可以用他的视角切入的角度
2. 一条可引用的主张（注明来源）
3. 一个他做法与普通认知的反差点，用作开头 hook
只用档案里的内容，不要脑补。

Automated Context Template

Store the distilled files as a persistent system prompt (Claude Projects) or as a Cursor rule using @file syntax. Then every new conversation automatically has Musk’s framework available.

Packaging as a Reusable Skill

Create a minimal YAML‑style skill file, for example:

---
name: elon-musk-perspective
description: |
  用马斯克的思维框架分析问题。基于 Sources/ 调研。
  触发词：「马斯克会怎么看」「第一性原理」「白痴指数」「五步算法」
---
## 核心主张
[粘贴 01_核心主张.md 内容]
## 决策方法论
[粘贴 02_决策方法论.md 内容]
## 反直觉观点
[粘贴 03_反直觉观点.md 内容]
## 使用规则
- 每次引用主张，注明来源
- 如有用户批注，优先使用用户视角
- 高敏话题仅给出框架，不下结论

Place this file in Claude Projects or Cursor’s .cursor/rules/ directory; the skill becomes a permanent assistant.

Full Workflow Overview

Build Archive : Create 20_People/Elon_Musk/ with 00_档案.md. Clarify why you are distilling him.

Collect Materials : Gather ≥20 first‑hand sources (biography, podcasts, talks). Avoid starting the AI before reaching the threshold.

AI Prompt Chain : Run Steps 1‑6 sequentially, writing each result to the matching file in Distilled/.

Add Personal Insight : Insert ≥3 personal notes (agreement, disagreement, exploration). Without them the distillation is incomplete.

Skill Packaging : Compress core files into a single skill definition and test with a query such as “用他的框架看这件事”。

Writing Calls : Before each article, @‑reference the skill or manually look up the relevant distilled file.

Iterative Update : After publishing, feed reader feedback and new insights back into Distilled/; the skill improves over time.

Iterative Update Loop

Each time you publish, note where the framework succeeded, where readers raised objections, and where your own understanding shifted. Write those observations back into the appropriate Distilled/ file, then regenerate the skill file so future AI assistance reflects the refined knowledge.

Key Takeaway

Distillation turns a person’s mental model into an AI‑ready knowledge asset. By following the SOP—archive, source, prompt, annotate, package—you gain a reusable 24‑hour consultant that can be invoked manually, via chat, or as an automated skill, while keeping the final judgment firmly in your hands.