mako yass comments on Is progress in ML-assisted theorem-proving beneficial?