The point is, LLMs look smart but they are not, and an easily-verifiable data point is that they can't count.