I am not following, a lot of things get turned into python calculations, so the LLM is not doing the precise math.