On Charities and Linear Utility — LessWrong