I was a little careless with how I phrased that. They said in the article they’ve done it, but it’s not “realized” in the sense that it’s not to a level of practicality that they’d want it to be. It can currently harvest signals to -20dBm, but they think they can get that to -62dBm for greater efficiency.
The main hurdle, according to them, is there’s no schottky diode that fits their needs, and they’ll have to engineer a new variant (at the nano scale…?). So, still a theoretical possibility on a more practical level, but this is hopeful news nonetheless.