Some advice on independent research — LessWrong