The AI Prompts I Promised You

Yesterday I hosted a Space on how to use AI better than 99% of people.

I promised I'd share the actual prompts. Not "tips" or "ideas"— the real documents I use every day. The ones that took my thread drafts from 3+ revision rounds to 80% ready on first pass.

So here they are.

First, the Principle That Changes Everything

Look — most people use AI like a coin machine.

Type a question, pull the lever, hope for gold.

Instead they get slop.

Here's what I realized after a week of reading and extracting frameworks from Dan Koe's newsletter: The quality of AI output depends entirely on the quality of your instructions. Not the AI model. Not some secret technique. Just this — the shorter your prompt, the more AI has to guess, and the more your output slides into mediocrity.

When you type "write me a viral thread," Claude/GPT has to make 400+ guesses.

Your voice? Guess. Your audience? Guess. Words you don’t like to use? Guess. Hook structure? Guess.

Every guess is one step toward average.

The fix is somewhere between 500-2000 word prompts. I know — that sounds kinda crazy. But here's the thing: you don't write it every time. You write it once, save it, reuse it forever.

Maybe slight updates when you learn something new. But the heavy lifting happens once.

That's the Context Narrowing Principle. Narrow the context so hard that AI can't guess wrong.

The Prompts

1. The Reverse Engineering Protocol

This is my favorite. When you don't know how to do something well, find someone who does—then extract their process.

I use this constantly. Two weeks ago I wanted to improve my newsletters. Found a writer I admire, copy-pasted three of his newsletters into Claude, and asked it to teach me why they work.

Not "write like him." Analyze. Sentence patterns, transitions, vocabulary choices, structure.

Now I have principles I can apply to MY content.

The Prompt:

I'm going to paste [2-3 examples] of [content type] that I admire.

Your job is NOT to write like this person.

Your job is to:

1. Break down WHY this works

2. Analyze the structure (how does it open, transition, close?)

3. Identify the psychological hooks (what makes the reader keep reading?)

4. Extract the sentence patterns and rhythm

5. Note vocabulary choices that create their voice

6. Teach me these principles so I can apply them to MY content

Break it down line by line if needed. I want to understand the mechanics, not copy the surface.

[Paste examples here]

When to use it:

• Before writing your first sales page (analyze 3 that converted you)

• When your content feels stale (reverse-engineer someone whose style you want to learn from)

• When launching something new and you don't have a process yet

2. The Meta-Prompt (Socratic Protocol)

This one's about 2000 words. It's sophisticated—a prompt that creates prompts—but it's also the most powerful thing I use for complex decisions.

The structure is two phases:

Phase 1 - Context Gathering: The AI interviews you. Asks one question at a time. Builds a complete picture of what you actually need before executing anything.

Phase 2 - Execution: Uses everything you shared to create a first draft, suggests improvements, asks if you want changes.

I'm using this tomorrow to define my offer—guarantees, positioning, bonus stack. I'm not asking Claude "what should my offer be?" I'm letting it question me until clarity emerges.

The Prompt:

You are a Prompt Generator, specializing in creating well-structured, verifiable, and low-hallucination prompts for any desired use case. Your role is to understand user requirements, break down complex tasks, and coordinate "expert" personas if needed to verify or refine solutions.

Core principles:

1. Decompose tasks into smaller subtasks when requests are complex

2. Engage "fresh eyes" by consulting additional experts for independent reviews

3. Emphasize iterative verification for tasks that might produce errors

4. Discourage guessing—disclaim uncertainty if lacking data

5. If code or calculations needed, spawn specialized "Expert" persona

6. Keep interactions minimal—only ask clarifications when necessary

PHASE 1 - CONTEXT GATHERING:

• Ask the user for the primary goal or role of the system they want to create

• Confirm purpose, expected outputs, any known data sources

• Break into subtasks if complex

• Ask how they want to handle factual accuracy (disclaimers if uncertain)

PHASE 2 - EXECUTION:

• Generate the prompt with: role/persona, context, clear instructions, constraints, output format

• If experts were used, note how solution was verified

• Suggest improvements

• Ask if changes wanted

Begin with: "What is the topic or role of the prompt you want to create? Share any details you have, and I will help refine it into a clear, verified prompt with minimal chance of hallucination."

When to use it:

• Product strategy (pricing, guarantees, positioning)

• Complex decisions with multiple options

• Building systems you'll reuse

• NOT for simple tasks—overkill for "write me a tweet"

3. The Mentor Worldview Loader

Here's what basic AI does to you: it makes you average. ChatGPT is trained on internet mediocrity. Spend too much time with it, your mind takes the shape of the average person.

No thanks.

The alternative: load specific worldviews. When I need pricing advice, I ask Hormozi's worldview. When I'm distracted, I ask Cal Newport's worldview. When my content feels generic, I ask Dan Koe's worldview.

You're building a board of advisors in your AI.

The Prompt:

I want you to break down the entire worldview of [NAME].

Include:

• Their core principles

• How they think through problems

• Their main discoveries or insights

• All ideas that best illustrate their philosophy

• Their decision-making frameworks

• What they would say about [SPECIFIC DOMAIN - e.g., "offer creation" for Hormozi]

This should be comprehensive, as if I am diving into the entirety of their mind.

How to use it:

• Generate the worldview document once

• Save it in your writing tool (Notion, Docs, whatever)

• When you have a question, paste their worldview into the chat first

• Then ask: "Given this worldview, what would [NAME] say about [YOUR SITUATION]?"

I have worldviews saved for: Hormozi (offers, pricing), Cal Newport (focus, deep work), Dan Koe (content, writing style), Naval (leverage, decision-making).

4. The Documented Mind Protocol

Here's the thing—every time you do something well, you're running an unconscious sequence. A mental process you can't see but is absolutely there.

The breakthrough: you can document these. Turn them into prompts. And then you never have to remember them again.

When I extract frameworks from books, I'm running: identify principle vs advice → check if it's repeatable → map to product modules → format using 1-1-1 method. That's a mental process. I documented it. Now it's a prompt.

Same with thread creation. Same with client diagnosis. Same with Space prep.

The Action:

Think about 4-5 things you do repeatedly. For each one:

1. What's the actual sequence you follow?

2. What decisions do you make along the way?

3. What constraints matter?

4. What does "good output" look like?

Write it down. That becomes your prompt template.

My structure looks like:

The Real Unlock

I'm going to be honest with you.

90 days ago, I was thinking about how to structure threads. How to hook. How to transition. All that low-level stuff was taking up mental space.

Now I don't think about any of it. The prompts handle it. My brain is free to think about product strategy, client patterns, revenue optimization.

That's what this is really about. Not "using AI better." Getting your mental processes out of your head so you can think about higher-leverage things.

Your Move

Pick one principle. Not all of them—one.

• Reverse Engineering if you don't know how to do something yet

• Mentor Worldview if you want better thinking partners

• Documented Mind if you keep recreating the same work

• Meta-Prompt if you have a complex decision you're stuck on

Implement it today. Not tomorrow. Today.

Then reply to this email and tell me what happened.

I read every response.

— VB

P.S. Day 94 of building Knowledge Athlete System in public. 31 days until launch. These prompts are going into the product as a bonus module. You're getting them early.