Sudoku coming up, I was actually rolling around in my head this idea of encoding information in unsolved sudoku puzzles as a form of compression, and wondering how much you can get out of it, how much of it needs to be error correction and the like. Never did any math on it, it's one of those things I think about and "will sit down and mess with later" supposedly but I doubt there will be anything novel or more efficient to come of it that's not directly translatable into existing compression.