Analogies between Software Reverse Engineering and Mechanistic Interpretability — LessWrong