Use a Thinking Framework in ChatGPT, Gemini & Other LLMs

If you're using Claude, the Claude Skill (.zip) does the work for you — upload once, and it auto-invokes when relevant. Read the Claude install guide →

This guide is for everyone else: ChatGPT users, Gemini users, Grok users, and anyone building on top of an LLM API. The plain .md file in your authority.md download works everywhere — here's how to use it.

What's in the .md file

Every framework you download from authority.md is a Markdown file with a specific structure:

• YAML frontmatter identifying the framework
• A core philosophy (2–3 sentences)
• Five named mental models with examples
• Decision heuristics
• Signature questions the person would ask
• What they would not do (anti-patterns)
• An activation prompt you can copy-paste

The file is designed so any LLM — ChatGPT, Gemini, Grok, or an agent framework — can use it as context and start reasoning through its lens.

Method 1: ChatGPT Custom GPTs

ChatGPT's Custom GPTs are purpose-built assistants you configure once and reuse. They're the best home for a thinking framework in the ChatGPT ecosystem.

1. Go to chatgpt.com and click Create a GPT
2. In the Configure tab:
   • Name: e.g. "Warren Buffett Thinking Framework"
   • Instructions: paste the contents of the .md file here
   • Knowledge: alternatively, upload the .md file as a knowledge document
3. Save and start using it

Using a Custom GPT vs knowledge upload

• Instructions paste — works well if the framework fits (most do). The GPT "thinks" through it on every message.
• Knowledge upload — better for very long frameworks, or if you want the AI to reference multiple files. Slightly less "always on".

Method 2: Paste into a single chat

If you just want to try a framework without setting up Custom GPTs, paste it into a fresh chat.

Each framework file ends with an Activation Prompt section. That's the shortest version — usually one paragraph — designed to be pasted directly into a chat to give the AI the framework's essentials.

Here is a thinking framework to apply for this conversation.
Read it carefully, internalise the mental models and decision rules,
then answer my question using this framework's approach.

[paste the .md file contents here]

My question is:

Then ask your question. The AI will use the framework for the rest of that chat.

Method 3: Agent frameworks and APIs

If you're building a product or agent on top of the OpenAI API, Gemini API, Grok API, or a framework like LangChain or CrewAI, the .md file content goes into your system prompt.

from openai import OpenAI

client = OpenAI()

with open("warren-buffett-framework.md", "r") as f:
    framework = f.read()

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[
        {"role": "system", "content": framework},
        {"role": "user", "content": "Should I invest in XYZ?"}
    ],
)

print(response.choices[0].message.content)

The framework becomes part of your agent's "personality" — every response is shaped by its mental models.

Gemini, Grok, and others

For Gemini, Grok, and other LLMs, paste the .md content directly into the system prompt (if the UI supports one) or include it at the top of your first message in a conversation.

Which method should I use?

Common mistakes

Pasting into ChatGPT's standard Custom Instructions

As noted above, Custom Instructions are character-limited and designed for user preferences ("I'm a vegetarian"), not rich documents. Use Custom GPTs or Project knowledge instead.

Using multiple competing frameworks in one chat

If you load both "Warren Buffett" and "Elon Musk" into the same Custom GPT, you're asking the AI to think two contradictory ways at once. It gets confused. Use one framework per GPT, or explicitly ask the AI to compare approaches.

Not starting a fresh session after upload

If you've added a framework to a Custom GPT's Knowledge, continuing an old conversation from before the upload won't see it. Start a fresh chat.

Next steps

Download a framework you actually want to try. If you haven't committed to one yet, start with a flagship — they're the most polished files in the catalogue.