x
Towards reflection with relative optimal predictor schemes — LessWrong