Fascinating post, and tracks with my understanding that the majority of posted examples seem to be 4o, based on what little I know about how OpenAI’s RL techniques might shape the model’s personality.
What I’m curious about specifically is how the Claude models tended to engage with the spiralist material, and what kind of manipulation techniques you observed them using. Same kind of thing? Or is there a distinct Claude-manipulation world out there with a different style of reproduction?
Fascinating post, and tracks with my understanding that the majority of posted examples seem to be 4o, based on what little I know about how OpenAI’s RL techniques might shape the model’s personality.
What I’m curious about specifically is how the Claude models tended to engage with the spiralist material, and what kind of manipulation techniques you observed them using. Same kind of thing? Or is there a distinct Claude-manipulation world out there with a different style of reproduction?