Give the LLM the ability to run the simulator and optimize the solution based on the outputs
Lots of amazing opportunities ahead