“Overly apologetic behavior in AI models like Claude is an undesirable trait that is hopefully decreasing over time.”