"AI" is good right now and feels inevitable. But the current models are trained on the extinct "pure" information state we had pre llm:s. Going forward we will have to start taking into account the current level of "ai slop" being added to the the information space. So I will have to trust my "detect ai generated information" LLM to correctly classify my main three llms responses as "hallucinating", "second level hallucinating", "fact based", "trustworthy aggregate" or "injection attack attempt". Probably should add another llm to check that response as well. Printed as a check list so that I can manually check it myself.