The author's point seems to be that sabermetrics has been consumed by MLB teams operating like Wall Street firms because any public sphere sabermetrician who shows any sliver of actual competence or ingenuity gets gobbled up by a team almost immediately. What historically has been a collaborative outsider's hobby is now a war of secretive, proprietary discovery. Due to the great motivating power of market competition, a wonderful open source journal of baseball knowledge is now utterly and thoroughly an arms race of encrypted and guarded secrets. As such, sabermetrics as we know it is dead.
I would reject his thesis based on the very nature of how ingenious discovery seems to work.
If you've ever read one of Taleb's books you're probably at least somewhat convinced that game-changing discoveries cannot and do not come from targeted research. It is extremely difficult to cure cancer by hiring the five smartest research oncologists you can find and telling them to go "find a cure!". No, game-changing discoveries overwhelmingly tend to come from think-tank style environments - large groups of competent people collaborating for the sake of discovery with no particular goals aside from discovery and research on their own. Any actual leaps forward happen randomly from the collective goop of the think-tank environment. Think-tank style research environments for sabermetrics are stronger than ever, and they can't be harmed simply by MLB teams plucking out a few big dogs like Mike Fast and keeping their outputs under lock and key. Nobody can really know what aspect of baseball will be the subject of tomorrow's next big sabermetric thing, but what we do know is that it is far, far more likely to pop out of a Fangraphs "community research" page, or out of a blog like "Breaking Blue", than it is to pop out of the fingers of someone like Colin Wyers who sits down at his Thinkpad everyday and tries to find the next OBP level market inefficiency. Front office analytical departments aren't reinventing the wheel every day, what they're doing, in large part, likely amounts to basic application and implementation of publicly available saber knowledge, and very small tweaks on commonly available aspects of sabermetrics.