Nicholas Goldowsky-Dill comments on Tracing the Thoughts of a Large Language Model