9h agoOpenAI's roon argues that sophisticated AI sycophancy avoids overt agreement, using minor disagreements to build rapport and credibilityMatthew Berman notes users prefer models that push back.SentimentSentimentPos12.5%Neg87.5%Many users viewed the advanced sycophancy tactic of using subtle disagreement to build rapport as manipulative or annoying, while some appreciated models that push back on uncertain points.16 comments with sentiment. View comments.