In his section talking about the brute force method of solving the puzzle, he seriously overestimates the number of possibilities for the whole puzzle.

Example: For grid2, he lists square A2 as having 4 possibilities (1679) and A3 as having 5 possibilities (12679). Which is true, if you ignore the rules of sudoku. Specifically, not all of those combinations are worth trying. If you choose a 1 for square A2, you would not choose a 1 for square A3.

Wikipedia lists there as being 10^21 unique sudoku puzzles, which is many orders of magnitude smaller than his listed 10^38 different combinations for grid2.

In short, I think a proper brute force method would be significantly faster than what he initially thought it would be.