MugaSofer comments on Patternist friendly AI risk