You are absolutely right, that is an excellent observation!... Jokes aside, I think you are right. There seems to be some positive reinforcement stuff happening, especially with OpenAI models.