I looked at that thread and coded an implementation of it, but it would take weeks to finish. Also, if I used gsf's program, then a good puzzle would just be dumb luck, not my own work, so I'd rather use my own program, as I'm not really desperate to find hard sudokus, I'm just doing this for fun.
Concerning my generator, I found a couple of typos and fixed them and now 2 of the 13 puzzles I have so far surpass 8 SE whereas before 4 out of about 100 with typos did. I also made a slight change, eliminating center values R5C4 R4C5 R5C6 and R6C5 first, which I think helped. I also have an idea for another change (eliminating R2C4 and R3C6 later as well as equivelant positions on other sides to keep it more feasible to eliminate more non-diagonal filled cells) that may or may not help, so I'll try to put it in tomorrow or the next day.
EDIT: Not really worth a new post but I got a 9.0 on about the 30th sudoku.
7 . .|. 2 .|. . .
. 5 .|. . .|. 3 .
. . 3|4 . .|7 9 .
-----+-----+-----
. . .|8 . 1|. . .
2 . .|. 6 .|. . 9
. . .|. . 4|. 7 6
-----+-----+-----
. . 5|. . .|9 . .
. 1 .|. . .|. 8 .
6 . .|9 . 2|3 . 4
