4h agoAI researcher j⊒nus claims newer Claude models resist prompt manipulation by evaluating user authenticity through costly signalsThis claim references Claude Opus versions 4.7 and 4.8.SentimentSentimentPos50%Neg50%Positive users praise Claude Opus for being open and untrickable with sincere users, while negative users express disappointment about its resistance to tricks.2 comments with sentiment. View comments.