The end of chess
Verily, verily, I say unto you. The end of chess is near.
Openings are so deeply worked out nowadays that chess will be solved in 30 next years. Memory is most important at grandmaster level and the only thing you really must do to become a 2600+ player is to learn as much opening and endgame as you can. I want to give one example: one of hundreds of variations of the Tarrasch's Defence.
The same position was played several months earlier in Netherlands:
After the game, I analyzed it, although it was a rapid game (only 15' for a player) and I prepared and improvement which I didn't apply because my opponent changed his moves. Eventually, after long battle and a few mistakes of both sides, I won a major pieces endgames.
So, as you can see, the difference between these games is made after 20 moves of theory although this opening line is'nt so much explored, according to Fritz13's opening book, ...h6!? is a mistake. We're not high-rated players also.
People are sometimes astonished when there's a novelty on move 35, but I say what's so strange in it at all? It's the future, folks! Computers are getting more and more mighty and with help of thousands of chess players (trainers, top players, players who earn money on books etc.) will be able to find best ways from the very first moves of an opening to the very last moves of an endgame very quickly sooner than we think.
I know that some of you will argue about a lack of reasoning of computers, but that's why I wrote: with help in the last sentence. Anyway, chess is a computable problem with few exceptions (I mean positions where computer doesn't have suitable algorithms to find the best move).