What a ridiculous take. Just because solutions exist in the llm training data, doesn't make these problems 'toy' or 'easy'. The human 'engineering hardness' scale doesn't align with what an llm can and can't do.