To your point, that we haven’t seen widespread examples of prompts that used to work but now don’t, seems telling about the accuracy of claims.