[pixl97]

GAN.. Just feed the output of your algorithms back into the LLM while learning. At the end of the day the problem is impossible, but we're not there yet.