Relatively thin post from OpenAI talking about their recent rollback of the GPT-4o model that made the model way too sycophantic - "overly flattering or agreeable", to use OpenAIs own …| Simon Willison’s Weblog
