Why Balsa Research is Worthwhile — LessWrong