The mathematics of reduced impact: help needed — LessWrong