So very soon we'll have yet another reason not to trust the accuracy of LLMs, and this time there will not even be a theoretical hope of ever correcting it.