Sound like this could be used to train an open source LLM.