Skip to main content
Goal: Quantitatively verify that both you and the AI are ready for implementation.

​
The Confidence Prompt

Ask the AI for a quantitative confidence score (0-100%) and a list of remaining ambiguities.

​
Example Prompt

On a scale of 0-100%, how confident are you that:
1. You understand the requirements completely
2. The proposed plan will work
3. You've identified all edge cases
4. There are no missing dependencies

For each area below 95%, what specifically is unclear?

​
What You’re Looking For

  • ✅ 90-100%: Ready to proceed
  • ⚠️ 70-89%: Specific gaps identified, address them
  • ❌ Below 70%: Not ready, more exploration needed

​
Closing the Loop

Use the AI to review its own plan to surface edge cases or missing error handling that you might have overlooked.

​
Self-Review Prompts

Edge Case Discovery 
"Review this plan and list all edge cases we haven't discussed.
What could go wrong? What assumptions are we making?"
Error Handling Check 
"For each step in the plan, what errors could occur?
Do we have a strategy for handling each one?"
Integration Points 
"What external systems or modules does this touch?
What happens if any of them are unavailable?"
Performance Considerations 
"Are there any performance implications we haven't addressed?
What happens under high load?"

​
The 90-95% Rule

Guidelines for iterating on the plan until a high confidence score is reached.

​
Iteration Process

  1. Ask for confidence score → Note areas below 95%
  2. Address each gap → Discuss, decide, document
  3. Re-check confidence → Score should increase
  4. Repeat until 90-95% → Ready for implementation

​
When to Stop Iterating

You’ve reached the point of diminishing returns when:
  • Confidence is consistently 90-95%
  • No new ambiguities are being surfaced
  • Further discussion is speculative, not actionable
  • You have a clear implementation plan
Important: 100% confidence is a red flag—it means you or the AI is overconfident. 90-95% is the sweet spot.

​
Total Sync Checklist

Before moving to implementation, verify:
  • AI has read all relevant files
  • Architecture is clearly understood
  • Dependencies are identified
  • Edge cases are documented
  • Error handling strategy is defined
  • Testing approach is clear
  • Naming conventions are agreed upon
  • Confidence score is 90-95%
  • No major ambiguities remain
If any item is unchecked: Go back to Bootstrapping or Idea Shaping.
Key Principle: The Rehearsal is complete when both human and AI have 90-95% confidence. Not before. Not at 100%.