Autograding is a valuable mechanism for assessment which facilitates scaling a curriculum to a much broader audience. Perhaps the best example can be found in the case of massive open online courses through platforms like Coursera and edX. However, developing autograders for lively, blocks-based programming environments can pose a number of unique challenges. These include the lack of clearly defined start state, often visual program output, and required user-interaction. How can we automatically grade assignments in these sorts of environments? What types of assignments can be graded easily?