
Train your AI to reject your ideas, not validate every one
You'll have a documented set of pushback prompts you use before implementing AI suggestions, plus a log of ideas your AI killed that you're glad didn't ship.
What you'll make
AI Skepticism Toolkit
- Challenge Protocol: Your standard questions for AI suggestions
- Failure Mode Map: Where your AI consistently gets it wrong
- Rejection Log: Ideas AI pushed that you're glad you killed
- Implementation Rules: When to trust vs when to dig deeper
Write each prompt and log entry with the specific context where you used it
The 4 parts
Each part ends with a check you can see.
Test AI on one recent decision
Pick one decision you made this week and see what AI would have suggested instead
Done when
You have one decision mapped with AI's suggestion, your actual choice, and a judgment about which was better and why
Build your AI challenge questions
Create 3-4 standard questions you'll ask before implementing any AI suggestion
Done when
You have 3-4 specific questions that probe for AI's blind spots, tested on 2 different AI suggestions
Catch AI being confidently wrong
Use your challenge questions to reject 2-3 AI suggestions that initially sounded good
Done when
You have 2-3 documented cases where your challenge questions revealed AI suggestions you're glad you didn't implement
Document your oversight system
Combine your challenge protocol and rejection log into a reusable AI skepticism toolkit
Done when
You have a documented system with your challenge questions, failure patterns, and implementation rules that someone else could follow
Capstone: you put together your AI Skepticism Toolkit.
About this lesson
What you'll be able to do
- Run AI suggestions through a documented challenge protocol before implementing them
- Identify 3-5 specific failure modes where your AI gives plausible but wrong advice
- Have a working log of ideas AI convinced you to reject that you're glad didn't ship
Who it's for
- Anyone who finds themselves implementing AI suggestions without much pushback
- People who've noticed they're thinking less deeply since they started using AI heavily
- Beginners to AI who want to build good critical thinking habits from the start
What it won't cover
- Advanced prompt engineering techniques
- Setting up complex multi-agent AI systems
- Technical AI safety or alignment theory
- Replacing human judgment entirely
Skills you'll practise
- Structured AI output evaluation
- Cognitive dependency recognition
- Critical questioning framework design
- AI suggestion stress-testing