Predictive coding & depression — LessWrong