Adam Jermyn comments on Scaling Laws for Reward Model Overoptimization