- search the entire space of 3x3 (and possibly 4x4 or higher) puzzles, looking for instances of
  a puzzle that has more than one solution.  This is much easier than it sounds:
        1. permute across all possible ways that polyominos can fill the space
                    (How?  Easy.  Each square has a four-way arrow that points to which neighbor it's joined to.)
        2. permute across all possible mathematical operations in those cages
        3. permute across all non-normalized latin squares, and calculate the 'result' for every cage
        4. compare the 'results' to all puzzles computed in step 3.  If any results match for different inputs, then we found one!
  The useful thing about this is that the only storage space required is for steps 3+4...  if you do
  those steps again with a different cage-shapes, or different operations, then there's no way the
  puzzles can be the same.  (unless you have wildcard operations)



==== far-out ====


    - write a solver that uses constraint-propogation ONLY, no brute-force