Humans can learn from new experiences. LLMs have to be retrained (continuous learning isn't good enough yet), or you have to fit enough information into the context while still having enough for the task itself.