Really cool idea the sim game approach! I'm trying to find a good approach to learning in the age of agents, though more focused on the actual code I read and write in my day to day. I'm building tech-professor.com which is in beta, where my goal is to setup continuous learning for product teams and developers. It's a hard problem to solve, but I believe that llm-based learning will become better as models improve!