prompt is to detailed (llm written?) better test of model creativity is to not constrain prompt