The ratio of science to fiction determine the type of science fiction you get. Hard sci-fi, a la Arthur C. Clarke tends more toward the science bit while Ursula K. Leguin is more fiction than science.
In the end, what we look for in Sci-Fi, I believe, is the "What if" thing. It lets you explore ideas and concepts that you normally wouldn't, using some amount of "suspension of disbelief".
This isn't a criticism of the book. The second half was pretty unpleasant but I enjoyed the first part.
Uncanny valley is a good description to me. I think Stephenson found an interesting in between. Not a criticism of the book; just an observation/assessment.
Historically, most of science fiction has been the opposite: 80% fiction and 20% "science" which usually means speculation about future technologies, without going into the details of how they could be realised. Realistic treatises of future scientific or technological breakthroughs are, to my experience (long time fan) absent from science fiction which is, after all, a genre of popular literature meant primarily for entertainment.
There is a trend of "hard" science fiction (like the Stephenson novel, which I haven't read) but since it's still just some guy (or broad) writing stuff in freeform language (rather than maths) without any attempt at formal definitions, or to prove theoretical results and test them empirically, one can only make very weak claims about "realism".
To boil it down to plain English: if you're looking for science, look at scientific publications, not science fiction publications.
Snow Crash is metaverse + extreme libertarianism + software/mind virus.
Diamond Age is nanotechnolgy + neovictorianism + clades.
Cryptonomicon is digital currency (not quite blockchain, but close), plus some WWII stuff.
The Baroque Cycle is "how did modern finance get invented* plus seggsytymes.
Anatham is "philosophy, protractors, and starships, but medaeval-like".
His earlier works (Big U, Zodiak, and the two he wrote with his uncle) also centre around one core technological concept, interacting with society, business, politics, etc. But the gist is "take some idea and run with it, see where it goes".
And to dismiss it and his other works as "take some idea and run with it" is to dismiss not only his work, but other well-written sci-fi.
Would you call Ringworld "dudes built a ring around a sun but entropy kicked in? Ender's Game, "just another Campbellian archetype?"
I'm not saying that this is all that there is to Stephenson. I'm saying that as I read it, this is the spine and core of the book --- the central organising premise, if you will. Other elements are hung off that.
Put another way, though there are characters in Stephenson, his books are not generally character driven. Or narrrative-driven. Plot ... maybe, though I'd argue that the central conceit drives the plot itself.
And for the record, I've enjoyed his books tremendously and seek them out, not something I can say of many authors these days.