The Master Prompt in a changing scenario

Written by Alec Fearon on 21 September 2025 in Anapoly Diary
Transparency label: AI-heavy
<--- Testing a local AI
---> A more personalised way to learn with NotebookLM

A discussion with NotebookLM

References

What is the Master Prompt?

Its relationship with Goal-Directed Context Management

Application to a programme-like endeavour

What is the Master Prompt?

Give me a concise explanation of the purpose, scope and content of the master prompt advocated by Tiago Forte and Hayden Miyamoto in the selected sources.

Based on the sources, the master prompt is a comprehensive document that provides an AI with detailed context about a person, company, or organisation to dramatically improve the quality, relevance, and speed of its responses. Its purpose is to transform the AI from a generic tool into a strategic partner that understands your specific business, goals, and operational style, thereby unlocking its full potential.

Purpose

The primary goal of the master prompt is to provide the AI with deep, persistent context, eliminating the need to explain your business details in every new conversation. By doing this, you can:

Get more done, better and faster: The quality and productivity of your work can improve drastically, enabling you to accomplish in under an hour what might have previously taken weeks.
Democratise execution: It allows small businesses and even one-person companies to perform complex tasks, like strategic planning or creating sophisticated hiring processes, that were once only feasible for large corporations.
Enhance strategic thinking: The AI, armed with your business context, can act as a “thought partner,” helping you think through the strategic implications of decisions.
Automate repetitive tasks: It can be used to systematise and automate annoying, repetitive business tasks, such as creating Standard Operating Procedures (SOPs) or generating marketing materials.

Scope

The master prompt’s scope is broad, intended to be a “second brain for an organization” that contains a wide range of information. It can be applied across all major business functions, including marketing, sales, finance, and operations.

The methodology involves several layers of context:

Level 3 (User Preferences): Basic settings on how you want the AI to communicate, such as format (e.g., bullet points), tone (e.g., conservative), and requests (e.g., stating its level of certainty). In Claude, this is called ‘Personal Preferences’; in ChatGPT, it’s ‘Custom Instructions’.
Level 4 (Project Knowledge): For more specific tasks, you can upload project-specific documents into dedicated workspaces. The AI will draw on these files for any conversation within that project, acting like a fully onboarded team member.

The master prompt itself is the most comprehensive layer, a single document containing detailed information about the entire organisation.

Content

Hayden Miyamoto suggests creating the master prompt in a document (like a Google Doc) with several key sections. The AI can even help you generate the content for each section by asking you a series of questions.

The recommended content includes:

Personal Info: Your name, role, strengths, weaknesses, and how you want to use AI.
Company Info: When the company was established, number of employees, reporting structure, markets served, and ideal customer profile.
Products and Services: Descriptions of what you offer, including pricing, features, and benefits.
Market Information: A description of your differentiators and key competitors.
Team / People: Who is on your team, their roles, and their key performance indicators (KPIs).
Culture: The organisation’s core values, mission, and long-term goals (or “Big Hairy Audacious Goal”).
Custom Prompts/Protocols: You can also embed specific frameworks or “mini-programs” that the AI can execute when given a trigger word, such as “do AI SOP” or “do AI hiring”.

The master prompt also contains, what might be described as functions, enabling the AI to be instructed “Do “. Give me a concise explanation of what these functions are, how they fit into the master prompt, and how they can be used.

Based on the sources, the “functions” you’re referring to are essentially pre-defined protocols or “mini-programs” embedded within the master prompt document. They are triggered by a specific command, such as “do AI hiring” or “do AI SOP”, which instructs the AI to execute a complex, multi-step task according to a pre-written framework.

What They Are

These functions are not literal programming functions but rather detailed sets of instructions, frameworks, and protocols that you write directly into your master prompt. They, products, and culture has been laid out. For instance, Hayden Miyamoto mentions having specific prompts like “AI hiring”, “AI CMO”, and “AI SOP” as distinct sections at the end of his master prompt.

By being part of the master prompt, these protocols have access to all the other context within the document. When a function is triggered:

The AI recognises the trigger word (e.g., “AI hiring”).
It accesses the detailed instructions for that specific protocol from the master prompt [ for marketing director”.

Execution: The AI, following the “AI hiring” protocol in the master prompt, automatically generates a comprehensive set of hiring documents. This includes:
- A detailed job description with responsibilities and performance metrics.
- A scorecard defining A-player vs. B-player performance.
- Recruitment materials like screening interview criteria and sample questions.
- Proposals for working interviews and homework assignments for candidates.
Impact: This process, which would typically take an HR director the better part of a week involving multiple meetings, can be accomplished in minutes. It democratises sophisticated business practices like “top grading,” making them accessible even to small companies that wouldn’t otherwise have the resources.

Another example is the “do AI SOP” function:

Trigger: The user provides a screenshot of a process flowchart and types “do AI SOP”.
Execution: The AI analyses the flowchart and writes out a detailed Standard Operating Procedure for every single step in the process.
Impact: This replaces what could be months of tedious documentation work, creating accountability and consistency within the organisation.

In essence, these embedded functions allow you to leverage a one-line command to unleash a powerful, context-aware workflow that dramatically improves productivity and the quality of the output.

Relationship with Goal-Directed Context Management

How do the ideas in goal-directed-context-management.md relate to those put forward by Tiago Forte and Hayden Miyamoto, in particular their levels of context and the master prompt?

Excellent question. The ideas in goal-directed-context-management.md and those put forward by Tiago Forte and Hayden Miyamoto represent two distinct yet highly compatible approaches to managing an AI’s context. Both frameworks recognise that providing context is the key to unlocking an AI’s full potential, moving it from a generic tool to a specialised, effective partner.

The core difference is one of perspective and application:

The Forte/Miyamoto Master Prompt method is an organisation-centric or user-centric framework. It focuses on creating a persistent, holistic “second brain” for a person or an entire organisation that the AI can access at all times.
The Goal-Directed Context Management (GDCM) framework is a project-centric or task-centric framework. It provides a formal methodology for managing the AI’s context dynamically throughout the lifecycle of a specific project or programme, ensuring the AI’s focus aligns with the current stage of work.

Here is a more detailed comparison of how their concepts relate, especially concerning the levels of context and the master prompt.

Relating the Levels of Context

The five levels of context described by Tiago Forte can be mapped directly onto the concepts within the GDCM framework.

Forte’s Context Level	Forte’s Description	GDCM Equivalent & Relation
Level 1: Training Data	The AI’s vast, pre-existing knowledge from its training on billions of documents.	This is the AI’s inherent “learned knowledge,” which both frameworks treat as a universal baseline that needs to be shaped and directed.
Level 2: System Prompt	The hidden, uneditable instructions from the AI’s creators that shape its core behaviour. Users can leverage “trigger words” to activate deeper thinking.	The GDCM framework does not explicitly name this layer but acknowledges that an AI’s behaviour is configured through instructions. The idea of using specific instructions to elicit desired behaviour is fundamental to GDCM’s concept of creating “AI artefacts” like bootstrap prompts and stage packs.
Level 3: User Preferences	Persistent settings (like ChatGPT’s ‘Custom Instructions’ or Claude’s ‘Personal Preferences’) that define a user’s preferred communication style (e.g., tone, format) across all chats.	This aligns with the GDCM concept of applying “a standard behaviour from the AI across all chats” using custom instructions. It represents a basic, global layer of context management.
Level 4: Project Knowledge	Uploading project-specific files into a dedicated workspace (like a Claude ‘Project’) so the AI acts like a “fully onboarded team member” for any conversation within that project.	This is directly analogous to the GDCM concepts of project instructions and project files. The GDCM framework formalises this by placing these files within a “project scaffolding,” which holds all the information the AI needs for its role. Both approaches use this level to silo context for specific workstreams.
Level 5: The Actual Prompt	The immediate, in-the-moment instruction you give the AI in the chat box. Forte notes that this prompt becomes “vastly more powerful” when the other four layers of context are in place.	This is the same as a “prompt” in the GDCM framework, which is used for controlling the AI during individual chats. The GDCM author agrees that a “precise prompt with the right background information gets us something much more useful”.

Relating the Master Prompt to GDCM Artefacts

The Master Prompt as advocated by Forte and Miyamoto is a comprehensive, single-document compendium containing all critical information about an organisation: its people, products, culture, processes, and more.

Within the GDCM framework, the Master Prompt serves as a perfect real-world example of the persistent context within the “contextual scaffolding”. It’s the enduring reference material that anchors the AI’s understanding of the organisation across all tasks.

However, the GDCM framework further organises and refines this idea by dividing the context into more granular components for managing large, complex projects:

Persistent Context: This is the GDCM equivalent of the Master Prompt. It holds the enduring, high-level project information like objectives, scope, quality standards, and references to core documents like HR manuals. An organisation’s Master Prompt would be a primary source for this persistent context.
Current Context: This is a key concept in GDCM that Forte and Miyamoto’s model doesn’t explicitly separate. The GDCM framework emphasises the need to load only ethe stage-specific working information into the AI’s active memory to prevent “distraction or drift”. For example, during a project’s “Design & Planning” stage, the current context would include the content outline and latest draft, but exclude exploratory materials from the earlier “Initiation” stage. This prevents the AI from getting confused by outdated or irrelevant information.
AI Artefacts: The GDCM framework introduces a formal set of documents, or “AI artefacts,” designed to manage the AI’s role throughout a project’s lifecycle. The Master Prompt could be seen as a foundational “AI Startup Pack”. The GDCM model then adds stage-specific “stage packs” that dynamically reconfigure the AI for different phases of work, ensuring it always has the right information at the right time.

Conclusion: Complementary Frameworks

The Forte/Miyamoto Master Prompt method and the Goal-Directed Context Management framework are not competing ideas; they are complementary.

The Master Prompt provides the “what”: a rich, persistent body of organisational knowledge that gives the AI a deep, foundational understanding. It is excellent for ongoing, daily use and for democratising execution across a business.
The GDCM framework provides the “how”: a structured methodology for deploying that knowledge strategically over the lifecycle of a complex, goal-directed project. It introduces the discipline of dynamically managing the AI’s focus with “persistent” versus “current” context, ensuring the AI remains an effective team member from project initiation to closure.

In practice, one could create a Master Prompt using the Forte/Miyamoto method and then use the principles of GDCM to extract and structure subsets of that information into “stage packs” and “current context” to manage the AI’s collaboration on a specific, multi-stage project.

Application to a programme-like endeavour

As presented, the Forte/Miyamoto Master Prompt is being used within a well-established enterprise. Consider an early-stage, startup enterprise that is working to turn a busiess idea into a prototype business model. The Master Prompt method needs to be adapted for this scenario, which has programme-like (rather than steady state, business-as-useal) charateristics. The Goal-Directed Context Management framework can guide that adaptation. Suggest how to combine the two approaches for this purpose.

Of course. This is an excellent scenario for combining the strengths of these two frameworks. The Forte/Miyamoto Master Prompt method provides a superb structure for capturing an organisation’s identity, while the Goal-Directed Context Management (GDCM) framework offers the perfect methodology for managing an evolving, programme-like endeavour such as building a startup.

For an early-stage startup, the “Master Prompt” isn’t a static document describing a well-established business; it’s a dynamic, living document that captures the current best understanding of the business as it evolves. The GDCM framework provides the structure for managing this evolution phase by phase.

Here is a suggested approach for combining the two frameworks for this purpose.

Create a “Proto Master Prompt” Using the Miyamoto Template

Start by creating the initial version of your Master Prompt in a Google Doc, following the structure Hayden Miyamoto advocates. However, treat it not as a final record but as a “Proto Master Prompt” — a collection of hypotheses and initial decisions. Use the AI as a “thought partner” and “business coach” to help you flesh out each section, just as Tiago Forte did in the demonstration.

The initial content will be aspirational and subject to change:

Personal Info: Founder’s strengths, weaknesses, and goals for using AI. This grounds the project in the founder’s vision.
Company Info (Hypothesised): Aspirational launch date, target markets, and a deeply considered Ideal Customer Profile (ICP). Getting specific on the ICP is crucial for differentiation and profitability, even if it evolves over time.
Products/Services (Proposed): Descriptions of the initial product idea, its features, and the benefits it’s intended to deliver.
Market Information (Initial Analysis): Hypotheses about differentiators and a first-pass analysis of potential competitors.
Culture (Foundational): The mission, core values, and the “Big Hairy Audacious Goal” (BHAG) that will guide the startup’s development.

This “Proto Master Prompt” becomes the foundational document—the persistent context in GDCM terminology. It represents the startup’s single source of truth at any given moment.

Adopt a Programme Approach to Business Model Development

Because turning an idea into a business model involves high uncertainty, it is best managed as a programme, not a single project. The GDCM framework outlines a typical programme lifecycle for a knowledge-based product (like a business model) which can be adapted for your startup:

Concept: Frame the initial idea.
Feasibility: Evaluate the idea’s viability and select a preferred approach.
Definition: Specify the requirements for a Minimum Viable Product (MVP) or prototype.
Development: Create the prototype.
Acceptance: Validate the prototype with the target market.
Operation & Maintenance: Launch the business model and begin iterating.

Each of these phases will be treated as a distinct project.

Use GDCM Artefacts to Manage the AI Through Each Phase

For each phase of the programme, you will use “AI artefacts” as described in the GDCM framework to manage the AI’s focus. This is where the two models integrate powerfuly.

Phase Example: Feasibility

Let’s say you are in the Feasibility phase. Your goal is to validate your ICP and initial product idea.

Project Scaffolding: Create a dedicated workspace for this phase (like a Claude ‘Project’ or a custom GPT).
Persistent Context: The “Proto Master Prompt” is your primary reference file. It ensures the AI always has the overarching vision in mind.
Current Context / Project Knowledge: This is where you load the stage-specific information into the AI’s active context to prevent it from getting distracted by irrelevant details. For the Feasibility phase, this would include:
- Market research data.
- Summaries of interviews with potential customers.
- Analysis of competitor products you’ve reviewed.
- The “stage pack” for this phase.
AI Stage Pack: This is a set of instructions, specific to the Feasibility phase, telling the AI its role. The prompt might be: “You are a market validation expert for an early-stage startup. Your goal is to rigorously test the assumptions in our Proto Master Prompt against the market data provided in the current context. Be critical and conservative in your analysis*. For each assumption about our ICP and product, state your level of confidence and identify the weakest points in our logic**.”*

Iterate on the Master Prompt After Each Phase

The output of each programme phase will be new knowledge. For example, the Feasibility phase might reveal that your hypothesised ICP was wrong, but a different, more promising segment emerged from your research.

Your final step in each phase is to update the “Proto Master Prompt” with these validated learnings. This is the critical feedback loop:

The ICP section is rewritten to reflect the new target audience.
The “Products/Services” section is updated to describe features that resonate with this new audience.
The “Market Information” section is refined with a clearer understanding of your true competitors and differentiators.

By doing this, the Master Prompt evolves from a set of hypotheses into a strategic plan grounded in real-world evidence. The AI, having access to this ever-improving persistent context, becomes an increasingly powerful and knowledgeable strategic partner, ready to help you tackle the next phase of development with a deeper understanding of your business.

This combined approach allows a startup to leverage the Master Prompt’s power for deep contextual understanding while using the GDCM’s structured, phased methodology to navigate the uncertainty inherent in building a new business.

The Master Prompt in a changing scenario

References

Contents

What is the Master Prompt?

Purpose

Scope

Content

What They Are

Relationship with Goal-Directed Context Management

Application to a programme-like endeavour