Jason Gross comments on Compact Proofs of Model Performance via Mechanistic Interpretability