Digital Error Correction and Lock-In — LessWrong