I'm on the downswing of AI. I no longer trust what AI tells me. Copilot gives one answer. ChatGPT gives another. Claude equivocates. Copilot flavors every new answer with the meat of previous conversations.

Reading AI generated code or blog posts you have to be extra dubious. AI will lie, and the human who prompted may or may not have vetted that information. The contents could be gold as easily as it could be sand.

For things that I need correct: I research on sources pre-2022. For things that don't really matter: I prompt AI and iterate.