Solving a problem with mindware — LessWrong