Woowww this is cool, not trained is even cooler--maybe I'm probably a bit naive, but to me it demonstrates an understanding of how an LLM can compute