We have a short paper with preliminary results from our new project: "On the Emergence of 'Useless' Features in Next Token Predictors." It was presented as a spotlight paper at ICML 2025 Workshop on Assessing and Evaluating World Models!