Problem
Misalignment between your understanding and AI's only reveals itself after implementation, wasting time on wrong solutions.
Pattern
Before letting AI implement, make it show its understanding:
- "Tell me what you're going to do before you do it"
- "Show me your plan" or "Draw an architecture diagram"
- "What questions do you have?" (not "ask me 3 questions" - let it surface real confusion)
- "What are you trying to achieve?"
Force it to be very succinct - makes what it tells you scannable.
Why this works: When you see something, you immediately spot what needs adjusting. Catch it early, adjust before implementation starts - not after wasting time going the wrong direction.
Example
Before refactoring: "Show me how you understand the current architecture and what you plan to change, succinctly"
AI draws diagram, reveals it misunderstood service boundaries. Corrected before any code was written - saved an hour of going in wrong direction.
