x
Goodfire and Training on Interpretability — LessWrong