The people that actually did the work are Mr. Maxime Rischard, Dr. J. Isaac Miller, Dr. Mark Glickman, and Jeff Sonas. I don't recognize Rischard or Miller offhand, but Jeff Sonas and Mark Glickman are familiar names. Sonas did interesting work in the past on establishing good historical ratings, and Glickman created the Glicko rating systems, versions of which are in use by FICS (RIP), Chess.com, LiChess, and the USCF. The new system has a fine pedigree, and is funded by the Grand Chess Tour, the Kasparov Chess Foundation, and the Chess Club and Scholastic Center of Saint Louis.
One good thing is that they've adjusted the ratings so that they resemble the old ratings in terms of scale. Magnus is at 2852, so the scale and magnitude are familiar - at least they haven't created something that looks like the ECF monstrosity!
The new rating list does have some oddities worth remarking upon, however. First of all, Carlsen's lead has been extended over his rivals. Games at faster time controls are included in the URS, though given less weight, and Carlsen has been consistently dominant across all three domains. Caruana slips from a close second in the FIDE standard list to a distant fourth, as he's not comparatively that good at faster time controls. Kramnik and Nakamura are tied for second, and Nepomniachtchi rounds out the top five. You can see their January list here.
In any event, Jeff Sonas has written a (not at all brief) summation of what they've done and why. Check it out, if that's your thing. Personally I will say that the work looks good on the face of it, but I'm not sure the old system really needed that much improvement! One thing the old Elo system had going for it was that one could easily calculate one's ratings after an event, and the basic method was clear cut. Now? No so much. Here's an example:
Rather than inventing a specific formula that can be used to calculate ratings directly, like there is for the Elo system and for performance ratings, we have developed a probability model that analyzes a large domain of possible ratings for each player, with some ratings being more likely than others (based upon the overall population distribution of chess strength). Across those possible ratings, our system then determines how likely the actual results would have been to occur, and ultimately determines the most likely overall set of ratings, for all players at once, in order to best explain the actual results.*Still, this is work upon which Sinquefield (and others) wish to hang their hats, and he who pays the singer calls the tune. I also suspect that this may be part of a plan by Sinquefield, Pein, Kasparov and others to wrest ratings from FIDE's grasp in an attempt to weaken that organization, and perhaps start something new in its place. It's been tried before, though, and failed, so I wouldn't count on this replacing FIDE's ratings in most people's minds any time soon.
* Incidentally, such a system really would not have been practical in the 1950s when Arpad Elo's system went into first use. Such wonders as the URS are possible because computing power has made both the sorting and calculating tasks needed for such a system relatively easily doable.
