Inspired by a blog post of Professor Alonso Castillo-Ramirez, I asked AI to look over an old survey paper I wrote in 2012. Long story short, ChatGPT-Sol (OpenAI) and Claude-Fable (Anthropic) found and corrected several errors. Embarrassing. (For the full mea culpa, see the abstract of the pdf of the corrected version.) AI even proved a conjecture I made in the paper and pointed out that another conjecture I made was a consequence of an Annals of Math paper published in 2020 . It’s embarrassing making a mistake, especially in a survey paper (apologies to the original authors!), but there is a rainbow at the end of the road.