Seth Herd comments on Legible vs. Illegible AI Safety Problems