An akrasia case study — LessWrong