Kaj_Sotala comments on Tracing the Thoughts of a Large Language Model