I did some research on this because I was curious too. So they downloaded the youtube video in 60 fps and used VLC to go frame-by-frame and analyze the movement with frames. They basically counted 10 frames in between movement. According to how you would break this down into readable terms, you would have to do 10/60 = .160 so that's the equivalent of 160ms of lag. For laymen, that would seem like a ping of 160 in a multiplayer game. I didn't make this method up, this was logic used by others to review videos of project xcloud and stadia.
I'm not disputing the video, BUT there are some factors not being taken into account.
Unfinished software and hardware (obvious point, but needs to be said again)
UI and other gameplay videos, like Astrosmash, don't seem to show this same lag (only in my limited testing)
Forward momentum - characters generally have some forward momentum, so you are not supposed to turn on a dime (vary by software title)
Overall I think the method is flawed because there are too many unknown variables, but I have never created a videogame either. From the video, it actually looks fine and I agree with others I am not worried. I hope that helps someone.