The Andes physics tutoring system is an experiment in student freedom. It allows students to solve a physics problem in virtually any legal way. This means that Andes must recognize an extremely large number of possible steps occurring in an extraordinarily large number of possible orders. Such freedom raises several research questions. (1) How can Andes solve the technical challenge of understanding student's behavior in such a wide-open context? (2) How can Andes give pedagogically useful help and guidance? In particular, how can it guide students who are floundering without curtailing the freedom of students who are not floundering? (3) Will Andes be effective in getting students in real classrooms to learn physics? (4) What does it take to scale up Andes and disseminate it widely? The Andes project, which began in the mid 1990's, has achieved workable solutions to the first three goals: Andes can understand student behavior; It provides pedagogical help similar to that of human experts; Most importantly, Andes causes large, reliable learning gains compared to control classes taught with convention, paper-based instruction. This chapter summarizes the first three results and discusses our progress on the fourth goal, scale-up.