David Scott Krueger (formerly: capybaralet) comments on [AN #114]: Theory-inspired safety solutions for powerful Bayesian RL agents