Okay, not all poker, but heads-up limit hold ‘em. Okay, not solved, but “weakly solved”, which means it is statistically extremely unlikely that a human would beat it (fair enough). And the strategy used, while cool and, classically, really AI, is an improvement on existing search-based techniques for game solving dating back to draughts.
It’s not that it isn’t a cool result. Drafts, Chess, Go, and so on are all perfect-information games, where players don’t have any information that their opponents don’t know. Poker isn’t, which makes play inherently probabilistic. The reporting of this is one was kind of atrocious, that’s all, with last-bastion-of-humanness-doomed headlines like the one above.
One thing I did find interesting, which has nothing to do with AI, is that several articles mentioned that it was strange that "inherently human" things, such as bluffing, are now being emulated by computers. Perhaps it’s just my background, but this is a really weird thing to write about. You see an opponent play a move which may be a bluff or may just be a bad move. Since you don’t know, you play as if it could be either. Ultimately the motive doesn’t matter. This has been the case in online poker for decades, where the only information available to you is the cards.