diff --git a/proj3/README.md b/proj3/README.md
index 6d60277..727add5 100644
--- a/proj3/README.md
+++ b/proj3/README.md
@@ -6,7 +6,31 @@ Project 3 Tests
 One of the provided shell scripts runs all the tests against a provided binary.
 Run it as follows 
 
-    $ ./test-cases <path-to-pylex> <path-to-pyparse> <path-to-pydesugar1> <test-dir>
+    $ ./test-cases <path-to-pylex> <path-to-pyparse> <path-to-sxpy> <path-to-pydesugar1> <test-dir>
+
+The reference pylex, pyparse, and sxpy are in the top-level 'ref' directory.
+As a shorthand you can run:
+
+    $ ./test-default <path-to-pydesugar1>
+
+Which will automatically use the reference pylex, pyparse, sxpy, as well as
+the standard test-directory. You should be able to run `test-default` from
+any directory.
+
+## Test Format
+
+There's only `.py` files this time since the output is different for different
+students. We have to use some weaker tests this time. We check two properties
+of the desugared output:
+
+    1. It conforms to the new AST. The AST has been simplified for this assignment
+       so we have a few heuristic checks to make sure you're not violating it.
+    2. The output of the desugared file is the same as the output of the 
+       original. We'll try and make test-cases where the output is mostly
+       dependent on the desugaring.
+
+Please keep test-cases that throw exceptions out of the repo since exceptions
+won't necessarily be thrown by desugared code. The only really standard property
+is that a valid python file should have the same output before and after
+desugaring.
 
-Just like Project 2 test cases. You can use the reference pylex, pyparse, etc.
-inside the proj2 dir.