Sudoku, really, are not terribly hard, once you figure out the methods to solving them. (I count two or three primary methods, depending on whether you include "go back and look, you missed something" as a method.) Really, the challenge to them is not making a mistake. Which is why I do them: more as a training for paying attention than any deep thinking. (And, they're a nice thing to settle my mind before sleep.)

What would be interesting would be to write a program that solves sudoku, but shows every step graphically. I bet it would make for interesting visuals. Also, I wonder if you would be able to visually see the logic: that is, the 'motion' of the logic. I also ponder if those visuals would be as good a teacher of the methods of solving as any text or person might be -- but, I think, that would depend greatly on the learner.

But what I find really interesting is that sudoku are purely deductive. At no point in the solving of the puzzle do you ever have to go "if X." As such, it is acurate to say that every sudoku is already solved at its very start: there needs only be the tumbling of the logical dominoes. (And, working a sudoku is really nothing more than locating each next domino. In fact, the only real question ever asked of a sudoku is asked during its making: "Is this array solvable, and to a unique solution." Once that question is answered in the affirmative, no further question need ever be asked.

Which, to me, is fascinating. Is there such a thing as sudoku in nature? The answer is no: it is pure abstraction (which thus permits the pure deductivity): the first number is two, and all numbers are abstractions.

What, then, is my pleasure in sudoku? I actually think that play to my OCD. But that's another issue altogether.