Step 3: Confidence Tuning

​

The Confidence Prompt

​

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

​

Self-Review Prompts

"Review this plan and list all edge cases we haven't discussed.
What could go wrong? What assumptions are we making?"

"For each step in the plan, what errors could occur?
Do we have a strategy for handling each one?"

"What external systems or modules does this touch?
What happens if any of them are unavailable?"

"Are there any performance implications we haven't addressed?
What happens under high load?"

​

The 90-95% Rule

​

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

• Confidence is consistently 90-95%
• No new ambiguities are being surfaced
• Further discussion is speculative, not actionable
• You have a clear implementation plan

​

Total Sync Checklist

• 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