This document discusses using simulations to speed up learning for complex problems. It describes using Monte Carlo simulations to rapidly experiment with different policies for a Kanban system, such as work in progress limits and task size distributions. The simulations can explore a "fitness landscape" to help understand the system and identify the most promising policy changes to try in practice. Running many simulated experiments can help optimize outcomes much faster than relying solely on sequential trial-and-error learning from real-world experiments.