x
Automated Alignment Research, Abductively — LessWrong