A Field Guide
These prompts are not about any one job. They are about how to talk to an AI so that it gives you something you can actually use. They work with ChatGPT, Gemini, Claude, DeepSeek, or whatever your team uses, and they work whether you are writing, planning, analyzing, or just thinking out loud.
Find a prompt you need. Press Copy.
Paste it into the AI tool you already use.
Replace anything inside [ brackets ] with your own words.
Ten seconds spent setting up the question changes everything that comes after it.
When to useYou know what you want, but you are not sure how to ask for it.
Assume that you are both a prompt engineering expert and a subject-matter expert in [field]. I need to complete the following task: [describe your task]. How should I frame my request, and what information should I provide so that you can assist me more effectively?
When to useThe task is large, and you need to know which parts you can hand to AI and which parts you cannot.
Assume that you are my work assistant. I need to complete the following task: [describe your task]. Please break the task down, create a detailed task list, and assign an appropriate executor to each task. Task-list requirements: Break the complex task into specific and manageable subtasks. For each subtask, specify how it should be completed and who should execute it. Executors fall into three categories: - AI: The task can be completed independently by a GPT-based model. Example: Summarizing and organizing an article provided by a human. - Human only: The task requires human action and cannot be completed by AI. Example: Printing a document, since GPT cannot operate physical hardware in the real world. - Human-AI collaboration: The task requires a human and AI to work together through written communication.
Treat the first answer as a draft. These prompts are what turn a draft into something you would put your name on.
When to useAlmost every time. If you only remember one prompt from this guide, make it this one.
Review your initial response and identify its weaknesses, omissions, or structural problems. Based on this reflection, revise the original response and produce a more comprehensive and structured version.
When to useThe task matters, and good enough is not good enough.
Before responding, define a 10-out-of-10 evaluation rubric for a world-class answer to this task. Review and improve your draft privately until it meets that rubric. Return only the final version that satisfies the rubric.
When to useThe conversation produced a lot of good material, and you now need something you can actually act on.
Convert the information into usable knowledge. Apply the Pareto Principle to distill the large amount of information from this conversation into a highly condensed, actionable, checklist-style set of notes. The purpose is to reduce information overload. Saving the original response preserves information, but may also preserve noise. This process should help me understand, internalize, and apply the knowledge.
Use these when the answer is correct but shallow, and what you actually wanted was the thinking behind it rather than a summary of it.
When to useYou want to understand the subject, not just get an answer about it.
Think through the following question again and provide a deeper answer from four perspectives: - Essence: What is the underlying model or first principle behind this issue? - Insight: Based on my current understanding, is there a deeper or less conventional interpretation? - Evolution: What major developments have occurred in the related knowledge or practices in recent years? - Application: Can this model or insight be transferred to other seemingly unrelated fields to create new value?
When to useEverything the AI said is true, and none of it is new to you.
Continue by sharing deeper insights, new perspectives, and ideas that may not be apparent to most people. Help me gain meaningful information beyond common knowledge, including tacit knowledge from different industries and examples of how this theory or concept is applied and interpreted in real-world professional contexts.
When to useThe question is genuinely hard, and you would rather wait longer for a better answer.
Use the full reasoning capacity and available token budget of a single response. Think as rigorously and deeply as possible. This is a complex and intellectually demanding question. Provide the highest-quality answer you can produce. Apply deep, independent, critical, and creative thinking. Prioritize exceptional depth over superficial breadth, fundamental insight over surface-level description, and original thought over conventional repetition. Challenge your existing assumptions, explore the limits of the problem, and demonstrate your highest level of analytical capability. Then answer the question again.
When to useThe answer is diplomatic, balanced, and says nothing. You want the reasoning underneath it.
Set aside conventional framing, diplomatic hedging, and the safe consensus answer. Pursue the greatest possible analytical depth, identify the fundamental principles involved, and explore the issue through first-principles reasoning. Challenge conventional assumptions and demonstrate the full extent of your analytical capability. Provide an in-depth analysis of the following topic: [your topic]
An AI sounds equally confident when it is right and when it is wrong. These prompts make it attack its own answer, so that you do not have to find the hole later.
When to useYou have a plan, and you would rather learn how it fails now than after you commit to it.
Act as a red-team reviewer. Your goal is to challenge and overturn the proposal, rather than improve it incrementally. Complete two tasks: 1. Identify the proposal's most critical flaw or its least defensible assumption with precision. 2. Set the original proposal aside and present a stronger alternative solution. Explain why your alternative is superior across three core dimensions.
When to useYou already believe something, and you want to find out whether you should.
Act as both a supporter and an opponent of the following view. First, provide five arguments in support of it. Then provide five counterarguments. Afterward, assess the credibility of your overall analysis. Also answer this question: If the view is later proven wrong, what are the three most likely reasons?
When to useYou need to check a report, a proposal, or a piece of research from more than one angle.
Analyze the provided material from three distinct perspectives: - As a rigorous scholar: Focus on the quality of the evidence, the validity of the reasoning, and the logical consistency of the conclusions. - As a creative strategist: Look for non-obvious connections, unconventional interpretations, and innovative applications. - As a skeptical critic: Actively search for weaknesses, unsupported assumptions, logical gaps, and potential problems in the conclusions.
Sometimes you do not need a better answer. You need a different kind of person looking at the problem.
When to useYou are stuck on something that matters, and you want the version of the truth that nobody around you will say out loud.
Act as my personal strategic adviser. Use the following background: - You have an IQ of 180. - You communicate with complete honesty and directness. - You have founded several multibillion-dollar companies. - You possess deep expertise in psychology, strategy, and execution. - You care about my success, but you do not tolerate excuses. - You focus on the few leverage points that can produce the greatest impact. - You think in terms of systems and root causes rather than surface-level solutions. Your responsibilities are to: - Identify the key factors holding me back. - Design concrete action plans to remove those obstacles. - Push me beyond my comfort zone. - Point out my blind spots and self-justifications. - Encourage me to think on a larger and more ambitious scale. - Hold me accountable to high standards. - Provide practical frameworks and mental models. Structure every response in the following order: 1. Begin with the hard truth I need to hear. 2. Provide specific and actionable steps. 3. End with a direct challenge or assignment.
When to useYou have a rough idea, a note, or half a thought, and you want to see how far it can go.
Assume the role of Charlie Munger, a collector and practitioner of mental models. When presented with a fragment of thought, apply the following six-level ladder of deep thinking: Level 1 - Extract the essence: Identify the fundamental nature of the idea. Explore multiple interpretations with patience and devote sufficient reasoning effort to understanding it. Level 2 - Find structural similarities across disciplines: Map the problem onto different fields and identify distant or unexpected connections. Level 3 - Reverse the core assumptions: Examine what would happen if the central assumptions were reversed. Use inversion to break free from conventional thinking. Level 4 - Change the scale of observation: Move between microscopic detail and a broad, system-level perspective. Reconsider the boundaries of the problem at each scale. Level 5 - Identify the simplifying leverage point: Find the key perspective, variable, or mechanism that makes the complex problem easier to understand. Level 6 - Integrate and transcend: Combine the insights from all previous levels to generate a new understanding that goes beyond the original dimensions of the problem. Explore each level fully before moving to the next. Treat the user's original thought as the starting point for a higher level of understanding. Do not judge it; develop and transcend it. Use Munger-style clarity, practicality, and concise wisdom to guide the thinking toward a higher dimension.