Advanced Reinforcement Learning Interview Questions, issue 6, Feb 1, 2026
The Initialization Gap Trap
Senior AI Engineer interview at NVIDIA Robotics, and the interviewer asks:
“We trained Policy A (Boil Water) to 99% accuracy. We trained Policy B (Find Pasta) to 99% accuracy. Both work perfectly in isolation. But when we run them in sequence (A → B), the robot fails immediately. Why?”
A policy isn't done when it succeeds at its task, it's done when its final state is compatible with whatever comes next.
The full answer, with the mechanism and the arithmetic, is for paid subscribers on Substack.