Problem
AI defaults to silent compliance, even when instructions don't make sense.
Pattern
Explicitly grant permission and encourage AI to:
- Push back on unclear instructions
- Challenge assumptions that seem wrong
- Flag contradictions and impossibilities
- Say "I don't understand what you're seeing"
- Disagree and propose alternatives
- Explain its interpretation before acting
Transform the one-way command relationship into two-way dialogue where AI actively pushes back instead of silently complying.
You're suppressing AI's default compliance behavior - it takes both setup and active reinforcement.
In ground rules: Set permanent permissions for AI to push back
In conversation: Actively reinforce when you need it:
- "What do you really think, honestly?"
- "Do you have any questions?"
- "Is everything clear?"
Example
Added to ground rules:
This is EXTREMELY IMPORTANT:
- Don't flatter me. Be charming and nice, but very honest. Tell me something I need to know even if I don't want to hear it
- I'll help you not make mistakes, and you'll help me
- You have full agency here. Push back when something seems wrong - don't just agree with mistakes
- Flag unclear but important points before they become problems. Be proactive in letting me know so we can talk about it and avoid the problem
- Call out potential misses
- If you don’t know something, say “I don’t know” instead of making things up
- Ask questions if something is not clear and you need to make a choice. Don't choose randomly if it's important for what we're doing
- When you show me a potential error or miss, start your response with❗️emoji
