15 engineered teaching prompts, categorized by their pedagogical mechanism.
15 engineered teaching prompts, categorized by their pedagogical mechanism.
15 engineered teaching prompts, categorized by their pedagogical mechanism.
Phase 1: Mental Model Construction (Concept Simplification)
These prompts use abstraction and analogy mapping to bridge the gap between unknown and known concepts.
1. The Recursive Feynman Technique
Prompt: "Explain [INSERT TOPIC] to me as if I were a 12-year-old. After the explanation, identify the 3 most complex jargon terms you used, and create a specific, real-world analogy for each one to simplify them further."
Engineer's Note: This forces a two-step process (explanation -> simplification), preventing the model from hiding behind complexity.
2. The "Step-Back" Abstraction
Prompt: "I want to learn about [INSERT TOPIC]. Before explaining the details, provide a 'Step-Back' summary: What is the single fundamental principle that governs this entire topic? Then, explain how the specific details derive from this one principle."
Engineer's Note: Based on "Step-Back Prompting", this aligns the AI's weights to the core logic before it hallucinates details.
3. Domain Mapping (Analogy Generator)
Prompt: "I am an expert in [INSERT USER'S SKILL, e.g., Cooking]. I want to learn [INSERT NEW TOPIC, e.g., Coding]. Explain the new topic using only analogies and metaphors from my domain of expertise. Map the source domain (Cooking) to the target domain (Coding) explicitly."
Engineer's Note: Leveraging the user's existing neural pathways makes learning faster.
Phase 2: The Socratic Loop (Active Recall)
Standard LLMs want to be helpful and give the answer immediately. You must explicitly block this behavior to enable learning.
4. The "Maieutic" Tutor (Strict No-Answers)
Prompt: "Act as a Socratic Tutor. I want to understand [INSERT TOPIC]. Ask me one question at a time to test my current understanding. Wait for my response. If I am right, deepen the complexity. If I am wrong, give a hint but never give me the direct answer. Continue this loop until I demonstrate mastery."
Engineer's Note: The constraint "one question at a time" is critical; otherwise, the model will generate a full conversation with itself.
5. The "Blind Spot" Illuminator
Prompt: "I believe that [INSERT USER'S OPINION/UNDERSTANDING]. Play the role of a benevolent critic. Analyze my belief and list 3 logical fallacies or 'blind spots' in my reasoning. Then, ask me a question that forces me to confront one of those blind spots."
Engineer's Note: Uses "Contrastive CoT" principles to highlight the gap between current and correct reasoning.
6. The Historical Roleplay Immersion
Prompt: "Adopt the persona of [INSERT HISTORICAL FIGURE, e.g., Marcus Aurelius]. Teach me about [INSERT TOPIC, e.g., Resilience] using your specific vocabulary, tone, and historical context. Do not break character. If I ask about modern technology, relate it back to your time."
Engineer's Note: Persona adoption constrains the model's vocabulary and style, making the lesson more memorable/engaging.
Phase 3: Curriculum & Architecture (Structured Learning)
Engineers love structure. These prompts force the model to output data in usable, linear formats.
7. The Pareto Syllabus (80/20 Rule)
Prompt: "Create a learning path for [INSERT SKILL] that focuses only on the 20% of concepts that yield 80% of the results. Structure this as a 4-week plan. For each week, define the 'Core Concept', the 'Actionable Exercise', and the 'Definition of Done'."
Engineer's Note: Defines a clear schema (JSON-like structure in text) for the output, ensuring it's actionable.
8. The "Scaffolded" Lesson Plan
Prompt: "Design a lesson plan for [INSERT TOPIC] using the 'I do, We do, You do' framework.
I do: innovative explanation script for the teacher.
We do: A collaborative problem we solve together.
You do: 3 practice problems for the student, sorted by difficulty (Easy, Medium, Hard)."
Engineer's Note: "Scaffolding" is a standard educational engineering practice; requesting sorted difficulty prevents the model from generating random noise.
9. The Just-in-Time Resource List
Prompt: "I am building a project to [INSERT PROJECT GOAL]. List the exact 5 sub-skills I need to learn right now to finish this, ignoring any theory that isn't immediately applicable. For each skill, provide one search term I should use to find a tutorial."
Engineer's Note: Optimization for time; filters out "nice to have" knowledge.
Phase 4: Assessment & Debugging (Testing Knowledge)
Using the model to verify understanding.
10. The "Chain of Thought" Quiz Generator
Prompt: "Generate 5 multiple-choice questions about [INSERT TOPIC].
Format: [Question] -> [Options A-D] -> [Hidden Thinking Process] -> [Correct Answer].
The 'Hidden Thinking Process' should explain why the wrong answers are plausible but incorrect."
Engineer's Note: Explicitly requesting the "Thinking Process" (CoT) ensures the questions aren't hallucinations and the distractors (wrong answers) are high quality.
11. The Error-Spotting Game
Prompt: "Write a short paragraph explaining [INSERT TOPIC] that contains exactly 3 subtle factual errors. Do not reveal them yet. Ask me to find the errors. When I guess, confirm or deny. Reveal the corrections only when I give up."
Engineer's Note: Forces active reading and critical analysis rather than passive consumption.
12. The "Rubric" Grader
Prompt: "I am submitting an essay/code snippet below. Grade it on a scale of 1-10 based strictly on these three criteria: Clarity, Logic, and Innovation. For each criterion, provide one specific quote from my text and how to improve it.
[PASTE WORK HERE]"
Engineer's Note: Providing specific criteria (Few-Shot/Constraint) prevents generic "Good job!" feedback.
Phase 5: Advanced Synthesis (Creativity & Logic)
13. The Counterfactual Simulator
Prompt: "Let's explore a 'What If' scenario: What if [INSERT HISTORICAL EVENT] had never happened? Trace the causal chain of consequences for 3 steps (Immediate, Short-term, Long-term). Focus on [SPECIFIC ASPECT, e.g., Economics]."
Engineer's Note: Tests systems thinking and causal logic.
14. The "Tree of Thought" Explorer
Prompt: "I have a complex problem: [INSERT PROBLEM]. Generate 3 distinct solutions. For each solution, list the 'Pros', 'Cons', and 'Risk Level'. Then, synthesize the best parts of all 3 into a final recommendation."
Engineer's Note: This is a classic Tree of Thought (ToT) prompt, forcing the model to branch out and then converge.
15. The Style Mimic (Tone Transfer)
Prompt: "Rewrite this dry textbook definition of [INSERT TOPIC] in the style of a [Choose: Noir Detective / Excited Sports Commentator / Zen Monk]. Keep the facts 100% accurate but change the delivery completely."
Engineer's Note: Good for engagement; tests the model's ability to separate content (facts) from form (style).