🎯 AI Machines: A Path to Mastery

Master artificial intelligence prompting and usage

---

📋 Table of Contents

• How to Elevate Basic Prompting into Expert-Level Interaction
  • The AIM Framework
  • The MAP Framework
  • The OCEAN Framework
• Step 1 & 2 (Week 1): Learning the Basics
  • Learn Machine English (AI Mechanics)
    • AI Predicts Language (Not Comprehends)
    • Uses Tokens/Vectors in Embedding Space
    • Generates Answer via Probability/Proximity
  • Use AIM Framework (Sharp Prompts)
    • A: Actor (Define Persona)
    • I: Input (Provide Context/Data)
    • M: Mission (Define Specific Goal)
  • Pick Your Instrument (Go Deep on One Model)
    • Examples: ChatGPT, Gemini, Claude
    • Learn Personality, Cadence, Limits
• Step 3: Build Context (The MAP Framework)
  • M: Memory (Conversation History/Continuity)
  • A: Assets (Attached Files, Data, Resources)
  • A: Actions (Tools model can call: Search, Write Code)
  • P: Prompt (Instruction itself)
• Step 4: Debug Your Thinking (Iterate and Refine)
  • Assume Fault is User's Prompting
  • Use Iterative Patterns/Cheat Codes
    • Chain of Thought ('Think step by step')
    • Verifier Pattern ('Ask clarifying questions')
    • Refinement Pattern ('Propose sharper input versions')
• Step 5: Steer to Experts
  • Avoid Generic, Superficial Outputs
  • Direct Model Towards Mastery/Depth
    • Specify Experts, Frameworks, and Research
    • Tip: Ask AI to list experts first
• Step 6 (Week 3): Verify Output
  • AI is Generative; Verify All Claims
  • 5 Verification Methods
    • Assumptions (List and rank confidence)
    • Source (Cite independent source)
    • Counter Evidence (Find credible disagreement)
    • Auditing (Recompute figures, show math/code)
    • Cross-Model Verification (Critique claims using other LLMs)
• Step 7 (Week 4): Developing Tastes (OCEAN Framework)
  • O: Original (Demand non-obvious ideas)
  • C: Concrete (Require names, examples, numbers)
  • E: Evident (Demand visible reasoning/evidence)
  • A: Assertive (Require a defensible stance/thesis)
  • N: Narrative (Guide the structure/flow like a story)

---

🎯 Discussions

How to Elevate Basic Prompting into Expert-Level Interaction

The AIM, MAP, and OCEAN frameworks elevate basic prompting—which often results in vague guesses and generic outputs—into expert-level interaction by introducing structure, context, and refinement (or "taste").

By applying the AIM and MAP frameworks, you can join the top 10% of AI users. Applying the OCEAN framework allows you to hit the absolute expert level.

What You'll Learn About the AIM Framework

The AIM framework transforms vague requests into targeted, computable instructions, helping the AI understand and compute your intent rather than merely trying to comprehend it.

AIM stands for:

• A (Actor): Telling the model who it is acting as. Instead of a generic query, you assign a persona, such as "the world's most sought-after résumé editor and business writer" who has reviewed thousands of résumés.
• I (Input): Giving the model the necessary context and data. This includes attaching resources like a resume and a job description for grounding the model.
• M (Mission): Defining precisely what you want it to do. This clarifies the goal, such as providing a bullet list of 10 specific ideas to improve clarity and measurable impact.

Using this three-part structure helps the model understand, compute, and reason with the prompt, often resulting in outputs that are at least five or ten times better than before.

What You'll Learn About the MAP Framework

The MAP framework provides the essential context that grounds the AI model, which otherwise resides in a massive mathematical space filled with billions of numbers. Context serves as "the map" that guides the AI on where to look and what matters.

MAP stands for:

• M (Memory): Providing continuity by including conversation history or notes from previous chat sessions.
• A (Assets): Attaching or copy-pasting files, data, or resources into the prompt to ground the model in reality.
• A (Actions): Defining tools the model can call to do work, such as searching the web, scanning a drive, or writing code.
• P (Prompt): The instruction itself.

The better the memory, assets, and external actions provided, the richer the context given to the AI, which results in better AI reasoning and response.

What You'll Learn About the OCEAN Framework

Once basic structure (AIM) and context (MAP) are established, the OCEAN framework is used to develop "tastes" and turn generic answers into insights that sound personalized and original. It encourages you to treat the AI like a sparring partner, pushing back and sharpening both your and the AI's thinking.

OCEAN stands for:

• O (Original): Pushing for nonobvious ideas and multiple angles. If the response lacks originality, you might ask for three angles no one else has considered.
• C (Concrete): Ensuring the response includes names, examples, and numbers that make sense. You push back until every claim is supported with a real example.
• E (Evident): Ensuring the reasoning is visible and supported by evidence. This involves asking the AI to show its logic in bullet points or provide evidence before delivering the final answer.
• A (Assertive): Requiring the response to take a stance that you can agree or disagree with. The AI is directed not to simply state what you want to hear, but to pick a side, defend its thesis, and address counterpoints.
• N (Narrative): Guiding the response to flow like a story, ensuring it is tight and follows a structure (e.g., hook, problem, insight, proof, actions).

By using OCEAN, basic, "junk food" AI outputs that sound like everyone else's are transformed into thoughtful, unique, and critically refined insights.

These frameworks collectively move the interaction away from relying on the AI's probabilistic "guessing machine"—which produces vague guesses for vague prompts—to a structured, contextualized, and iterative conversation where you and the AI are learning together. This transition moves you from passive consumer of AI output to active co-creator.

---

Step 1 & 2 (Week 1): Learning the Basics

The 7-Step Roadmap: Master AI in 30 Days is designed to help beginners master AI like the top 1%. The initial steps, Step 1 and Step 2, occur during Week 1 and focus on Learning the Basics by mastering communication with the AI and selecting a foundational tool.

What You'll Learn About Machine English (AI Mechanics)

The first week of the roadmap starts with learning "machine English". This skill is essential because most people interact with generative AI systems like ChatGPT or Gemini incorrectly, treating them like a person, which is described as a "huge mistake".

Why Machine English is Necessary:

• AI systems predict, not understand: Generative AI models do not actually understand human language; they predict the most likely next token based on probability and proximity within a massive mathematical embedding space.
• Vague inputs yield vague guesses: When a prompt is vague, the AI (described as a "guessing machine") produces vague guesses. Conversely, if a prompt is sharp and targeted, the AI will return sharp and targeted guesses.
• Goal: Machine English helps the AI compute your intent rather than simply trying to comprehend it.

What You'll Learn About AI Predicts Language (Not Comprehends)

Generative AI systems, such as ChatGPT or Gemini, "don't actually understand our language; they predict it". This prediction process is what sets these models apart from traditional search engines and dictates how you must communicate with them.

How the Prediction Works:

1. Tokenization and Vectors: The AI breaks down your text into smaller parts called tokens (e.g., "Humpty" or "Dumpty"). Each token is then converted into a list of numbers known as multi-dimensional vectors.
2. Embedding Space: These numbers are placed within a massive mathematical space called an embedding space. In this space, similar ideas or words tend to live closer together (e.g., "Humpty," "egg," "wall," and "fall"), while unrelated words (like "motorcycle" or "chocolate") are far apart.
3. Probabilistic Guessing: When generating an answer, the AI looks at the context and predicts the most likely next token based on probability and proximity within the embedding space. The answer is generated on the fly and not retrieved from "stored facts" or "memory".

The example of "Humpty Dumpty sat on a" illustrates this: just as your brain predicts "wall," the AI weighs all options ("party," "day," "chocolate") and selects the most likely outcome, "fall," to finish the line. This probabilistic nature is why AI "can feel so smart but also so alien".

What You'll Learn About Uses Tokens/Vectors in Embedding Space

To understand how AI generates language, you need to understand tokenization and embedding space, which are the mechanical foundations of how generative models process and predict text.

Tokenization: The AI breaks down text into smaller parts called tokens. A token might be a word or part of a word. For example, "Humpty Dumpty" is split into the tokens "Humpty" and "Dumpty".

Vectors: Each token is then converted into a multi-dimensional list of numbers known as vectors. These numbers represent the token in a way the AI can mathematically compute.

Embedding Space: These vectors are placed within a massive mathematical space called the embedding space. This space is where the AI "lives" and operates. In this space:

• Similar ideas tend to live closer together (e.g., "Humpty," "egg," "wall," and "fall" are nearby because they appear in similar contexts).
• Unrelated ideas are far apart (e.g., "motorcycle" or "chocolate" are far from the Humpty Dumpty cluster).

The embedding space is built from billions of parameters that have been trained on massive amounts of text data. The AI uses this space to compute which tokens are most likely to come next in a sequence, given the surrounding context.

What You'll Learn About Generates Answer via Probability/Proximity

Once the AI has placed tokens into the embedding space, it generates responses using probability and proximity—the core mechanism of how it predicts what comes next.

Proximity in Embedding Space: The AI examines which tokens are nearby in the embedding space based on the context provided in the prompt. Because similar concepts live closer together, the AI can quickly locate the most relevant candidates.

Probability Calculation: The AI then calculates the probability of each candidate token being the correct next word. The model has been trained to assign higher probabilities to sequences that commonly appear together in the training data.

Generated on the Fly: The answer is generated token by token, on the fly. The AI does not retrieve pre-stored answers or facts; rather, it builds the answer dynamically based on the probabilities and proximity in the embedding space.

For example, given "Humpty Dumpty had a great," the AI weighs all possible next tokens (e.g., "fall," "day," "party") and selects the one with the highest probability—in this case, "fall"—because that phrase commonly appears in the training data.

---

What You'll Learn About Using the AIM Framework (Sharp Prompts)

To create sharper, more structured prompts, the AIM framework is introduced. This framework transforms a simple request into a structure the model can understand, compute, and reason with. Using the AIM structure transforms prompts and can make results at least five or 10 times better than before.

What You'll Learn About A: Actor (Define Persona)

The first component of the AIM Framework is Actor, which involves telling the model what persona or role it should adopt. This assignment of a specific role or expertise area is crucial for grounding the AI's response in a relevant domain of knowledge.

Purpose: By defining an actor, you direct the AI to access the specific "area" of its embedding space that contains the knowledge, tone, and style associated with that persona. Instead of a generic or unfocused response, the AI tailors its output to match the expertise and perspective of the assigned role.

Example: Instead of a vague request like "fix my resume," you might instruct the AI: "You are the world's most sought-after résumé editor and business writer. You've reviewed thousands of résumés that led to interviews at top tech companies."

This instruction situates the AI within the expertise of a professional résumé editor, which helps it compute a more targeted, useful response.

What You'll Learn About I: Input (Provide Context/Data)

The second component of the AIM Framework is Input, which requires you to give the model the necessary context and data to ground its response in reality.

Purpose: Providing input ensures that the AI has the specific information it needs to generate a relevant and useful answer. Without sufficient input, the AI is forced to guess broadly, often resulting in generic or off-target responses.

Example: Following the actor assignment, you might specify the context: "I'm attaching my resume and the job description for a senior product manager role at a fintech company."

This grounding information allows the AI to focus its recommendations on the specific role and industry, making the output far more applicable to your actual situation.

What You'll Learn About M: Mission (Define Specific Goal)

The third component of the AIM Framework is Mission, which defines precisely what you want the model to do. This clarifies the specific goal of the interaction.

Purpose: The mission component ensures the AI understands the desired output format and the specific task it needs to accomplish. A clear mission prevents the AI from wandering or producing irrelevant content.

Example: Continuing the résumé improvement example, you might define the mission: "Review it and give me a bullet list of 10 specific ideas on how to improve clarity, measurable impact, and alignment with the role. The goal is to help me build the best resume that gets me hired."

This explicit mission statement ensures the AI knows exactly what deliverable you expect and what criteria to optimize for.

---

What You'll Learn About Pick Your Instrument (Go Deep on One Model)

Once you understand how to communicate using machine English, the next part of Week 1 is to "pick your instrument".

The Wrong Way vs. The Recommended Way:

• Most people start their AI journey incorrectly by skimming many tools, Googling the "top 50 AI tools," and jumping from one to the other, which is described as a "recipe for failure".
• The recommendation is to pick one foundational model and go deep.
• Learning one model deeply allows you to grasp the rhythm of the others faster, similar to how an experienced drummer picks up a guitar faster than a complete beginner because they understand practice and patterns.

What You'll Learn About Examples: ChatGPT, Gemini, Claude

While the specific choice "doesn't matter," suggested foundational models include:

• ChatGPT: Recommended for users who want the "most mature one".
• Gemini: Suggested for users who are "deep into Google stack and Google's ecosystem".
• Claude: Recommended for "more business and project-based AI".

What You'll Learn About Learning Personality, Cadence, Limits

By deeply exploring one model, you train your brain to see structures and patterns. Each model has:

• Personality: The tone, style, and general approach the model takes in its responses.
• Cadence: The rhythm and flow of how it generates text and structures information.
• Limits: The boundaries of what it can and cannot do effectively, including token limits, knowledge cutoffs, and areas of strength or weakness.

Week 1 Goal: By the end of Week 1, you should be comfortable writing a structured prompt using the AIM framework without conscious effort.

---

Step 3: Build Context (The MAP Framework)

To ensure smart outputs, you must feed the AI context, which acts as a map to navigate the complex mathematical space within the model. The framework for context is MAP.

Context is vital because without it, even the world's smartest AI will sound clueless. Using both the AIM and MAP frameworks puts you in the top 10% of AI users.

What You'll Learn About M: Memory (Conversation History/Continuity)

Memory refers to the continuity provided by the conversation history or notes that carry over from previous chat sessions with the AI. This component ensures that the AI has access to the ongoing narrative of your interaction.

Purpose: Memory helps the AI understand the broader context of your request by referencing what has been discussed previously. Without memory, each prompt is treated as an isolated event, forcing you to repeat information.

Implementation: To build continuity, you can:

• Repaste the thread from a previous conversation
• Ask the model to summarize the discussion so far before starting a new segment
• Use tools or features within the AI platform that maintain conversation history across sessions

What You'll Learn About A: Assets (Attached Files, Data, Resources)

Assets are the files, data, or resources that you attach or copy-paste into your prompt. These assets help to ground the model in reality by providing concrete, specific information.

Purpose: Assets ensure that the AI is working with actual data rather than making broad generalizations. By providing documents, spreadsheets, code snippets, or other resources, you give the AI a tangible foundation for its reasoning.

Examples:

• Attaching a PDF of a research paper
• Copy-pasting financial data or sales figures
• Including code samples that need review or debugging
• Providing your company's style guide or brand documentation

What You'll Learn About A: Actions (Tools model can call: Search, Write Code)

Actions are the tools that the model can call to do work beyond generating text. These capabilities expand the AI's utility from a text generator to an active assistant.

Purpose: Actions allow the AI to gather information, execute tasks, and interact with external systems, making it far more powerful and useful for complex workflows.

Examples of Actions:

• Search the web: Retrieve current information or verify facts
• Scan your drive: Access and analyze files from cloud storage
• Write code: Generate, test, or debug programming scripts
• Create documents: Generate formatted documents in various formats

What You'll Learn About P: Prompt (Instruction itself)

The Prompt is the instruction itself—the actual question or command you're giving to the AI. While this might seem straightforward, the quality of the prompt is dramatically improved when it's supported by Memory, Assets, and Actions.

Purpose: The prompt is the culmination of all the context you've provided. With rich memory, relevant assets, and available actions, even a simple prompt can yield sophisticated results.

Key Insight: The better you incorporate memory, assets, and external actions, the richer the context you give the AI, which leads to better AI reasoning and response.

---

Step 4: Debug Your Thinking (Iterate and Refine)

Achieving expert level requires debugging your thinking. When you don't get the desired answer, the problem is usually your thinking, not the AI. Prompting requires iterating, not just typing.

What You'll Learn About Assume Fault is User's Prompting

The fundamental principle of Step 4 is to assume the fault lies with your prompting, not with the AI model. This mindset shift is crucial for improvement.

Why This Assumption Matters:

• It forces you to critically examine your own communication and thought process
• It prevents you from dismissing the AI as "broken" or "stupid" and giving up
• It encourages you to learn the model's strengths and how to leverage them

What to Check: When output is weak, ask yourself:

• Did I provide the correct persona (Actor)?
• Did I give sufficient context (Input)?
• Was my goal clearly defined (Mission)?
• Did I include relevant memory, assets, and actions (MAP)?

You can also ask the AI to explain its logic and reasoning chain to learn how it thinks. This iteration process teaches you how to understand the model and teaches the model how to understand you.

What You'll Learn About Use Iterative Patterns/Cheat Codes

To improve your prompting through iteration, three specific patterns or "cheat codes" are provided. These patterns are proven techniques for refining AI output.

What You'll Learn About Chain of Thought ('Think step by step')

Chain of Thought is a prompting pattern that asks the model to show its reasoning process before delivering a final answer.

Instruction: "Think step by step, show your reasoning, then give me the final concise answer."

Purpose:

• Forces the AI to break down complex problems into logical steps
• Allows you to see where the reasoning might be going wrong
• Often produces more accurate results because the AI "checks its work" as it goes
• Helps you understand the AI's logic so you can refine future prompts

What You'll Learn About Verifier Pattern ('Ask clarifying questions')

The Verifier Pattern instructs the AI to seek clarification before attempting to answer, ensuring it fully understands what you need.

Instruction: "Ask me clarifying questions one at a time before attempting the answer again."

Purpose:

• Identifies gaps or ambiguities in your original prompt
• Ensures the AI understands the nuances of your request
• Helps you refine your thinking by forcing you to answer specific questions
• Creates a more collaborative, iterative process

What You'll Learn About Refinement Pattern ('Propose sharper input versions')

The Refinement Pattern asks the AI to improve your prompt itself, offering better versions for you to choose from.

Instruction: "Propose two sharper versions of the question and ask me which one I prefer."

Purpose:

• Leverages the AI's understanding of effective prompting
• Teaches you what makes a good prompt by showing you improved alternatives
• Allows you to choose the version that best matches your intent
• Accelerates your learning curve for prompt engineering

---

Step 5: Steer to Experts

AI often produces generic answers when prompted vaguely. To avoid mediocrity, you must direct the model toward the "sharper edges of its brain" by navigating it toward experts, frameworks, and depth.

What You'll Learn About Avoid Generic, Superficial Outputs

When you ask the AI a vague question without guidance, it tends to produce "generic, mid, superficial" outputs. This happens because the AI defaults to the most common, widely-applicable information in its training data—what might be described as the "center" of the embedding space.

The Problem:

• Generic outputs lack depth and nuance
• They sound like every other AI response
• They don't leverage the full knowledge embedded in the model
• They fail to provide the expert-level insights you need

The Solution: You must actively steer the AI away from generic responses and toward specific expertise.

What You'll Learn About Direct Model Towards Mastery/Depth

To access deeper, more sophisticated outputs, you need to guide the AI toward specific areas of expertise and mastery within its knowledge base.

What You'll Learn About Specify Experts, Frameworks, and Research

The most effective way to steer toward mastery is to explicitly reference specific experts, established frameworks, or relevant research in your prompt.

Example: Instead of asking a generic question like "How can I improve team creativity?", you could ask:

"Explain using ideas from Pixar's brain trust, Satya Nadella's strategy, and Harvard's research on psychological safety."

Why This Works:

• It directs the AI to specific areas of its knowledge that are rich with expertise
• It ensures the response synthesizes multiple authoritative perspectives
• It moves the output from generic advice to expert-level insight
• It prevents the AI from being just an "echo chamber" of common knowledge

What You'll Learn About Tip: Ask AI to list experts first

If you're not sure which experts, frameworks, or research to reference, you can ask the AI to suggest them first.

Two-Step Process:

1. Discovery: "List the top experts and frameworks related to [your topic]."
2. Synthesis: Take those experts and feed them back into a refined prompt: "Now explain [your question] using insights from [Expert A], [Expert B], and [Framework C]."

Benefits:

• You don't need to already know the field to get expert-level answers
• The AI helps you discover the most relevant authorities
• You can then guide the AI to synthesize those specific perspectives
• This creates a collaborative discovery process rather than passive consumption

---

Step 6 (Week 3): Verify Output

Because generative AI is designed to generate things, it can be confidently wrong. You must verify the output, critiquing rather than consuming the information.

What You'll Learn About AI is Generative; Verify All Claims

The fundamental nature of generative AI means it creates plausible-sounding text based on patterns, not fact-checking. This makes verification essential.

Why Verification is Critical:

• AI generates responses probabilistically—it predicts what "sounds right" based on patterns in training data
• The AI can be "confidently wrong"—presenting false information with authoritative tone
• There's no built-in fact-checking mechanism; the AI doesn't "know" if something is true
• You must shift from passive consumption to active critique

Mindset Shift: By the end of the third week, you should feel more in control of the output. You are not accepting the AI's word as final; you are verifying, questioning, and refining.

What You'll Learn About 5 Verification Methods

Five specific verification methods are provided to help you rigorously check AI output:

What You'll Learn About Assumptions (List and rank confidence)

This method involves asking the AI to explicitly state all the assumptions it made in generating its response, and to rank its confidence in each one.

Instruction: "List and rank every assumption you made."

Purpose:

• Makes hidden assumptions visible
• Allows you to evaluate whether those assumptions are valid for your situation
• Helps you understand the foundations of the AI's reasoning
• Reveals areas of uncertainty where the AI might be guessing

What You'll Learn About Source (Cite independent source)

This method requires the AI to provide specific citations for its claims, including verifiable details.

Instruction: "Cite two independent sources (including title, URL, and a quote) for each major claim."

Purpose:

• Forces the AI to ground claims in verifiable information
• Allows you to check the sources yourself
• Reveals when the AI is making unsupported assertions
• Provides transparency about where information comes from

What You'll Learn About Counter Evidence (Find credible disagreement)

This method pushes the AI to present opposing viewpoints or contradictory evidence, ensuring a balanced perspective.

Instruction: "Find a credible source that disagrees with your answer and explain the dependencies."

Purpose:

• Prevents one-sided or biased responses
• Shows you the full landscape of perspectives on a topic
• Helps you understand the nuances and debates in the field
• Demonstrates intellectual honesty and thoroughness

What You'll Learn About Auditing (Recompute figures, show math/code)

This method requires the AI to show its work, particularly for any numerical claims or calculations.

Instruction: "Recompute every figure and show your math or code."

Purpose:

• Verifies the accuracy of quantitative claims
• Allows you to check the methodology
• Reveals errors in calculation or reasoning
• Provides transparency in how numbers were derived

What You'll Learn About Cross-Model Verification (Critique claims using other LLMs)

This method involves using multiple AI models to verify each other's claims, creating a system of checks and balances.

Instruction: "Run the prompt in different models (ChatGPT, Gemini, Claude) and then ask one model to critique or verify the claims made by another."

Purpose:

• Leverages different training data and architectures to catch errors
• Identifies where models agree (higher confidence) and disagree (needs verification)
• Creates a more robust verification system than relying on a single model
• Helps separate factual knowledge from confident errors

---

Step 7 (Week 4): Developing Tastes (OCEAN Framework)

The final step focuses on making AI output sound like you, rather than sounding generic (like "junk food output"). You should treat the AI like a sparring partner—pushing back and sharpening both your and the AI's thinking.

The OCEAN Framework turns generic answers into tasteful insights. This framework is what elevates you to the absolute expert level of AI interaction.

What You'll Learn About O: Original (Demand non-obvious ideas)

Original is the first component of the OCEAN Framework, which is the methodology used for executing Step 7: Developing Tastes (Week 4). This step is focused on transforming generic AI output into personalized, insightful content by demanding novelty and multiple perspectives.

The Purpose of O: Original

The overarching goal of Step 7 is to move past the production of routine or "junk food output" and generate results that sound distinctly like you, reflecting your unique "tastes". The Original component specifically addresses the AI's tendency to produce conventional, predictable ideas.

By pushing for originality, you force the AI to explore less-traveled areas of its knowledge base and generate responses that stand out from typical answers. This is a key element of treating the AI as a "sparring partner"—arguing with it and pushing back to sharpen both your and the AI's thinking.

The Specific Instruction

To ensure the output is original, you must use a prompt that explicitly demands nonobvious thinking:

• The Command: "Give me three angles on this problem no one else has considered. Label one as risky and tell me which one you like the most and why."

This instruction serves multiple functions:

1. Multiple Angles: Asking for three angles forces the AI to explore diverse perspectives rather than settling on the first obvious answer.
2. Nonobvious Ideas: The explicit request for ideas "no one else has considered" pushes the AI away from generic, widely-known responses.
3. Risk Assessment: Asking the AI to label one angle as "risky" encourages it to think about unconventional or bold approaches.
4. AI Preference: Having the AI select and defend its preferred angle creates a more engaged, thoughtful response.

Context within the OCEAN Framework

Original is the first step in a sequence of refinements:

• O (Original): The user demands nonobvious ideas and multiple angles.
• C (Concrete): The user then grounds these original ideas in factual reality.
• E (Evident): The user next ensures the reasoning behind those concrete claims is visible and supported.
• A (Assertive): This pushes the AI to take a firm stance.
• N (Narrative): The final step guides the structure and flow.

By demanding originality from the start, you establish a foundation that prevents the rest of the OCEAN process from polishing generic ideas. This practice is part of the overarching idea that "you're not just training the model you are training you".

What You'll Learn About C: Concrete (Require names, examples, numbers)

Concrete is the second component of the OCEAN Framework, which is the systematic approach for implementing Step 7: Developing Tastes (Week 4). This component is essential for grounding the AI's original ideas in verifiable, tangible reality.

The Purpose of C: Concrete

The overall goal of Step 7 is to generate output that moves beyond "junk food" responses and creates material that sounds distinctly like you. The Concrete component specifically combats the AI's tendency to speak in abstractions and generalities.

By requiring names, examples, and numbers, you ensure that every claim the AI makes is backed by specific, verifiable details. This transforms vague assertions into substantive arguments that can be evaluated, verified, and trusted.

The Specific Instruction

To make the output concrete, you must use a prompt that demands specificity:

• The Command: "Back every claim with one real example—names, numbers, or a concrete case."

This instruction ensures that:

1. Names: The AI must cite specific people, companies, products, or entities rather than speaking in generalities.
2. Examples: Abstract principles must be illustrated with real-world instances.
3. Numbers: Quantitative claims must include actual data points, statistics, or measurements.

Context within the OCEAN Framework

C: Concrete builds directly on the foundation of originality and prepares for evidence and assertion:

• O (Original): The process begins by demanding nonobvious ideas.
• C (Concrete): The user then grounds these original ideas in factual reality by demanding specific examples and data points.
• E (Evident): The user next ensures the reasoning behind those concrete claims is visible and supported by evidence.
• A (Assertive): This pushes the AI to take a firm stance.
• N (Narrative): The final step guides the structure and flow of the complete, insightful argument.

By demanding concrete details, you act as a "sparring partner", actively pushing back against the AI's generic tendencies and ensuring that the final output is not only unique but also empirically solid. This process is part of the broader self-improvement mandate of Week 4: "you're not just training the model you are training you".

What You'll Learn About E: Evident (Demand visible reasoning/evidence)

Evident is the third component of the OCEAN Framework, which is the methodology used for executing Step 7: Developing Tastes (Week 4). This step is focused on transforming generic AI output into personalized, insightful content by demanding transparency and logical support.

The Purpose of E: Evident

The overarching goal of Step 7 is to move past the production of routine or "junk food output" and generate results that sound distinctly like you, reflecting your unique "tastes". The Evident component ensures that the sophisticated ideas generated through the OCEAN process are logically sound and transparently supported.

By requiring the AI to make its reasoning visible and confirm it has "enough evidence," you push the AI to move beyond a summary and explicitly demonstrate the path it took to reach its conclusion. This is a crucial element of treating the AI as a "sparring partner"—arguing with it and pushing back to sharpen both your and the AI's thinking.

The Specific Instruction

To ensure the output is evident, you provide a specific, multi-part prompt that demands transparency:

• The Command: "Show your logic in three bullets. Provide evidence before you provide final answer."

This instruction serves two critical functions:

1. Visibility of Logic: Requiring the logic in structured bullet points forces the AI to present its chain of thought clearly, allowing you to follow and critique the reasoning.
2. Evidence Precedes Conclusion: Demanding that the evidence be presented before the final answer prevents the AI from presenting an assertion without first establishing the basis of that claim.

Context within the OCEAN Framework

E: Evident is situated within the OCEAN sequence to ensure that the novel claims are thoroughly substantiated:

• O (Original): The process begins by demanding nonobvious ideas.
• C (Concrete): The user then grounds those ideas by requiring names, examples, and numbers to back every claim.
• E (Evident): Following the introduction of original and concrete claims, you demand visible reasoning and evidence to demonstrate the validity of those points.
• A (Assertive): This builds on the evidence by demanding that the AI take a specific, defensible stance.
• N (Narrative): Finally, all these elements are organized into a cohesive story or flow.

By consistently demanding visible reasoning and evidence, you ensure that the output is not just smart, but verifiable and critically robust. This active engagement during Week 4 emphasizes that "you're not just training the model you are training you".

What You'll Learn About A: Assertive (Require a defensible stance/thesis)

Assertive is the fourth component of the OCEAN Framework, which is the foundational methodology for implementing Step 7: Developing Tastes (Week 4). This component is crucial for moving AI output from generic summaries to critical, personal, and insightful material.

The Purpose of A: Assertive

The primary goal of Step 7 is to generate AI output that sounds distinctly "like you". By the time you reach this stage, you are transitioning from generating competent, verified output to injecting personal "tastes".

The Assertive component specifically addresses the AI's tendency to provide neutral, non-committal answers. By demanding a defensible stance, you ensure the output:

• Takes a Position: It must take a clear position that you "could agree or disagree with".
• Avoids Pleasing the User: It forces the AI to move beyond telling you "what I want to hear".

This process is a key element of treating the AI as a "sparring partner"—arguing with it and pushing back to sharpen the thinking of both the model and yourself.

The Specific Instruction

To make the AI output Assertive, you must use a prompt that compels the model to commit to a thesis and defend it:

• The Command: "Don't tell me what I want to hear. Pick a side, state your thesis, defend it, and then address the best counterpoint."

This rigorous instruction ensures that the final output possesses critical depth by requiring the AI to:

1. Pick a Side: Establish a clear thesis or position.
2. Defend It: Provide evidence and reasoning (drawing on the prior Evident step) to support the thesis.
3. Acknowledge Opposition: Address the "best counterpoint", demonstrating a well-rounded and defensible argument. This is where "real reasoning lives".

Context within the OCEAN Framework

As the fourth letter in OCEAN, the Assertive step builds directly upon the groundwork established by the preceding elements:

• O (Original): You demand nonobvious ideas.
• C (Concrete): You ensure those ideas are grounded with names, examples, and numbers.
• E (Evident): You require visible logic and evidence to support the claims.
• A (Assertive): With the original, concrete, and evident claims in hand, you push the AI to form a definitive, defensible thesis based on that material.
• N (Narrative): Finally, these well-defended insights are organized into a cohesive story.

By demanding an assertive stance, you ensure the content moves beyond mere information synthesis into the realm of critical insight, which is necessary to "turn generic answers into tasteful insights". This practice is part of the overarching idea that through this rigorous process, "you're not just training the model you are training you".

What You'll Learn About N: Narrative (Guide the structure/flow like a story)

Narrative is the fifth and final component of the OCEAN Framework, which is the systematic approach for implementing Step 7: Developing Tastes (Week 4). This step ensures that the rich, insightful content developed during Week 4 is presented effectively and cohesively.

The Purpose of N: Narrative

Step 7 is dedicated to generating output that moves beyond the "same junk food output everyone else gets" and instead creates material that sounds specifically "like you". After you have rigorously guided the AI to produce original, concrete, evident, and assertive claims, the Narrative component focuses on refining the presentation.

The goal is to ensure the final output "flows" well and is "tight". By demanding a narrative structure, you turn the compilation of facts and arguments into a compelling, digestible story. This is a vital part of injecting your "tastes" into the output, ensuring the content is not just accurate and insightful, but also beautifully composed.

The Specific Instruction

To implement the Narrative component, you take on the role of a director or editor, guiding the AI's structure:

• The Command: "Write it like a story: hook, problem, insight, proof, actions—whatever you want in that story."

This instruction gives you control over the storytelling elements of the output, such as the hook, problem, insight, proof, and actions. This ensures the final output is cohesive and maintains a high-quality flow.

Context within the OCEAN Framework

Narrative serves as the culminating step in the OCEAN Framework, bringing together all the critical elements developed previously:

1. O (Original): You demand nonobvious ideas.
2. C (Concrete): These ideas are grounded with names, examples, and numbers.
3. E (Evident): You ensure the logic is visible and supported by evidence.
4. A (Assertive): The AI is pushed to take a defensible stance/thesis and address a counterpoint.
5. N (Narrative): Finally, all the resulting original, concrete, evident, and assertive claims are woven together into a clear, compelling story.

By using the OCEAN framework and treating the AI as a "sparring partner" throughout Week 4, you transform "generic answers into tasteful insights". The final Narrative step ensures that your mastery—developed through every prompt, revision, and judgment—results in output that is not only factually superior but also structurally masterful. This rigorous process means that "you're not just training the model you are training you".

---

📚 Source