0 XP

Prompts Are Thinking Tools

The quality of your AI output depends on the quality of your thinking — not magic words.

💡
Key insight: Prompts are the result of clear thinking, not the cause of it. If your thinking is muddy, no prompt trick will save you.

The Real Problem

Most people treat prompts like magic spells — they think if they find the right words, AI will produce perfect results. But the truth is simpler:

💥 Unclear Thinking

"Help me with my history project"

You haven't decided what you need. The AI guesses randomly.

✓ Clear Thinking

"Break down the causes of WW1 into political, economic, and social factors. For each, give me one primary source I could cite."

You know exactly what you need. The AI delivers precisely.

Different Tasks Need Different Thinking Modes

Just like you wouldn't use the same approach for a math problem and an essay, different AI tasks need different thinking modes. That's what the 7 frameworks teach you.

Why do "prompt templates" often fail?

The 7 Prompt Superpowers

Seven different thinking modes for seven different situations.

🔌

1. Break It Down

Split big problems into small parts

🌱

2. Start From Scratch

Build understanding from basics up

📚

3. Research Mode

Gather and compare evidence

🛠

4. Blueprint Mode

Plan how to build something

🔬

5. Prompt Coach

Improve the prompt itself

🎓

6. Expert Lens

Answer from a specific expertise

🥊

7. Debate Partner

Challenge your reasoning

When To Use Each

Break It Down: "My essay topic is too broad — help me split it into 3 focused sub-questions"
Start From Scratch: "Explain machine learning starting from the simplest possible concept, building up layer by layer"
Research Mode: "What are the 3 main positions on school uniforms? Give evidence for each side"
Blueprint Mode: "I want to build a study app. Design the architecture: what screens, what data, what features"
Prompt Coach: "Here's my prompt. What's weak about it? Rewrite it to get better results"
Expert Lens: "Answer as a constitutional lawyer would: is this policy legal?"
Debate Partner: "Challenge every assumption in my argument. Find the weakest points"

You're preparing for a debate and need to find weaknesses in your own argument. Which superpower?

Choose The Right Framework

Match your situation to the right thinking tool.

Decision Flowchart

Click the situation that matches yours:

😶
Problem too big
Overwhelmed, don't know where to start
🤔
Don't understand topic
Need to learn from basics
🔍
Need evidence
Gathering info on a topic
🔧
Want to build something
App, project, system
🔄
Prompt not working
Getting bad results
🎓
Need specialist view
Specific expertise needed
🥊
Need to stress-test
Check for weaknesses

Your science fair hypothesis might be wrong, but you can't see why. Which framework?

Prompt Builder Challenge

Put it all together: build a powerful prompt using what you've learned.

Build Your Prompt

📋 Prompt Structure Cheat Sheet

Keep this one-page reference card handy whenever you write prompts. It covers the 4-part structure, tips, and real examples adapted for students.

The 4-Part Prompt Structure

1. Context
Role, background, knowledge
2. Instruction
Steps, chain of thought, examples
3. Input
The text, data, or question
4. Output Format
Bullet points, JSON, table, essay

Quick Tips

  • Be specific: "Write an essay" is weak. "Write a 500-word argumentative essay on carbon taxes for my AP Environmental Science class" is strong.
  • Use delimiters: Wrap your input in quotes or dashes so the AI knows what's instruction vs. what's data.
  • Give examples: For style or formatting tasks, show 1-2 examples of what you want. The AI learns the pattern.
  • Break it down: For complex tasks, tell AI to work step by step instead of answering all at once.
  • Say what TO do: "Summarize in 3 bullet points" works better than "Don't write too much."
  • Iterate: Your first prompt is a draft. Refine based on what the AI gives back.

Student Examples

Extract key info from a syllabus:

"You are a study planner. From the syllabus text below, extract all assignment due dates, deliverables, and grading weights. Output as a table with columns: Assignment, Due Date, Weight. Syllabus: [paste text here]"

Rewrite for a different audience:

"Here are 2 examples of engaging TikTok-style scripts: [example 1] [example 2]. Now rewrite the following club announcement in the same style: [paste text here]"

Summarize into a title:

"Read the article below and generate a concise, descriptive title that captures the main theme. The title should be under 15 words. Article: [paste text here]"

🌟
Track 1 Complete! You now know 7 different thinking modes for AI. Use them in your schoolwork, projects, and creative work. Switch to Track 2 to learn how to make AI your personal tutor.

Every Project Can Teach You

Don't just use AI to finish work — use it to understand what you're doing.

🎓
The difference: Using AI to skip thinking = you learn nothing. Using AI to explain while it works = you learn everything.

Two Ways To Use AI

🚫 Just Finishing

"Write my essay about photosynthesis"

You get an essay but learn nothing. Can't explain it. Can't build on it.

✓ Learning While Doing

"Help me write about photosynthesis. After each paragraph, explain the concept in simple terms and give me an analogy I can remember."

You get an essay AND understanding. You can explain it to others.

The Learning Artifact

Every time you work with AI on a project, you can ask it to leave behind learning notes — a document that explains what was built, why decisions were made, and what you should remember.

Which approach helps you learn more?

Build Your Learning Notes

Create a "FOR_YOU.md" — your personal learning notebook that AI maintains for you.

What Goes In Your Learning Notes?

📚

What We Built

Plain-English summary of the project

💡

Concepts Learned

Key ideas with analogies

🐛

Bugs & Lessons

What broke and what it taught you

🔬

Questions to Review

Things to quiz yourself on later

Template Builder

Choose your project type and fill in the template:

Learn By Explaining

The ultimate test of understanding: can you explain it back?

Three Power Moves

💬

Explain Back

"Now let me explain this in my own words — correct me if I'm wrong"

🧠

Quiz Me

"Generate 5 quiz questions on what we just covered"

💡

Why Does This Work?

"Don't just show me how — explain WHY this approach works"

The Explain-Back Prompt

You: Okay, let me try to explain what you just taught me. Correct anything I get wrong:
So photosynthesis is basically plants eating sunlight? They take in CO2 and water,
use the sun's energy to rearrange the atoms, and produce glucose and oxygen.
The glucose is their food and the oxygen is basically their exhaust.
Claude: That's a great explanation! A few refinements:
1. "Eating sunlight" is a good analogy — more precisely, they absorb light energy
2. "Exhaust" for oxygen works! It's a byproduct they release
3. Small correction: glucose isn't exactly "food" — it's stored energy they...
🌟
Track 2 Complete! You now know how to turn every AI interaction into a learning opportunity. Switch to Track 3 to learn how to catch AI mistakes using multiple models.

Why One AI Can Miss Its Own Mistakes

The same brain that wrote the work uses the same blind spots to review it.

👀
The blind spot problem: You don't proofread your own essay right after writing it — your brain fills in what it meant to write, not what's actually on the page. AI has the same problem.

Same Model = Same Blind Spots

When one AI writes something AND reviews it, the review uses the same reasoning that created the original mistakes. It literally cannot see its own errors because it would have avoided them in the first place.

👀 Self-Review

AI writes an answer → same AI says "looks good!"

Same assumptions, same gaps, same blind spots. Bugs hide.

👥 Cross-Review

AI #1 writes → AI #2 reviews → AI #1 fixes

Different training, different strengths, catches more errors.

Why is "passing tests" not enough to prove code is correct?

Claude Writes, Codex Reviews

A practical workflow where two AI models check each other's work.

The Cross-Review Workflow

Claude Writes
🔎
Codex Reviews
🔧
Claude Fixes
👤
Human Approves

Why Two Different Models?

Claude and Codex (by OpenAI) are trained differently, think differently, and catch different issues:

Claude

Fast, practical, great at building. Your daily driver.

📚

Codex

Rigorous, thorough, catches subtle issues. Your academic reviewer.

What The Reviewer Catches

In a real test with a REST API project, the cross-model review caught:

  • Security holes — database queries that could be hacked
  • Race conditions — two operations conflicting
  • Missing protections — no rate limiting, leaked error messages
  • Missing features — no pagination for large data sets
💭
School analogy: It's like peer review in English class. You write the essay, your partner reads it with fresh eyes, catches what you missed, and you make it better before turning it in.

Review Detective: Spot The Issues

Practice finding problems in AI-generated content. Click on the issues!

AI-Generated Study Guide — Find 4 Problems

This study guide was written by AI. Click on the highlighted sections that contain errors or problems:

The American Revolution began in 1774 when colonists decided to fight for independence from Britain.

The main cause was that British people were mean to Americans.

George Washington led the Continental Army to victory at the Battle of Yorktown in 1781. The Treaty of Paris was signed in 1783, officially ending the war.

After the war, the Constitution was written in 1789, establishing the federal government.

The Revolution gave freedom to all Americans and inspired democratic movements worldwide.

Found: 0 / 4 issues

Build Your Review Team

Design a multi-model review workflow for any project.

The 4-Agent Review Team

Builder

Creates the initial work

🔎

Critic

Finds errors and weak points

🔒

Safety Checker

Looks for risks and harmful content

🎓

Explainer

Makes sure output is clear and accurate

Design Your Review Workflow

🏆
All Tracks Complete! You've mastered three essential AI skills: thinking with frameworks, learning while building, and reviewing with multiple perspectives. These habits will make you dramatically more effective with any AI tool.