But yes, Egan's philosophy is that sci-fi can be about the human story OR about the science. Both are valid.
From this long interview:
> I’m interested in science as a subject in its own right, just as much as I’m interested in the effects of technology on the human condition. In many things I write the two will be combined, but even then it’s important to try to describe the science accurately. In a novel such as Incandescence, though, the entire point is understanding the science, and it really doesn’t bother me in the least that it’s not an exploration of the human condition.
> There are times when it’s worth putting aside the endless myopic navel-gazing that occupies so much literature, in order to look out at the universe itself and value it for what it is.
Of his recent fiction, I enjoyed "Zeitgeber" [1] and "The Nearest" [2]. I found the latter to be an interesting exploration of what happens when you suddenly discover that you don't recognize your loved ones as human.
I think he achieved exactly what he set out to do with his writing.
But the characters still aren't why anyone reads them. Nowadays they're serviceable; years back they were cardboard cutouts. Nobody who enjoys Egan is bothered by either.
You can get all sorts of combinations of various attributes, but you just can't get them all. Certainly Egan is in the Silver Age Asimov/Clarke tradition of sacrificing characters to fit the rest in, but what he does is so good I can't complain. For all the crazy science fiction I've read, Egan has, just to name one example, the only story set in a serious attempt at an alternative space-time geometry in which there are two space and two time dimensions, and not in some vague what-if "parallel dimension" sense, but by actually working out the consequences of what two spatial and two temporal dimensions looks like when you feed that into the equations of general relativity: https://www.gregegan.net/DICHRONAUTS/00/DPDM.html
(I find this one particularly interesting because many of us have imagined "what if there were two time dimensions"; Egan shows that what came to most of our imaginations was so laughably wrong as to not even be in the ballpark, and that in fact you can perfectly sensible discuss the topic grounded in real physics.)
Even in the field of science fiction writers there's only a certain percentage of writers who could even properly follow that page, let alone generate it themselves.
But as counter-examples I submit J. G. Ballard, Samuel Delaney, Ursula LeGuin, John Crowley, Gene Wolfe, William Gibson, Harlan Ellison, Ray Bradbury, and Theodore Sturgeon — and there are surely others I'm forgetting. All great prose stylists, all masters of characterization, and arguably among the greats of the genre.
'Nuff said.
Joe Haldeman's non-SF novels were all very good in their way. I wonder if he would have been happy to make the switch to full time non-genre writing if 1968 (his Vietnam novel - although he'd joke that all his novels are Vietnam novels) hadn't fizzled. I'm not the right person to criticize his poetry, but I enjoyed some of it.
As some others mentioned, Iain Banks, Ursula LeGuin, and William Gibson wrote some great non-genre stuff. In a different world William Gibson could have been a great essayist in the vein of John McPhee.
It is amusing to think that a lot of people who are defending Egan on this point would probably be uncompromising on their very low opinion on a book containing badly put together science. They will probably not say "it's not for me, but why not, the author was interested in the characters, the science is just an excuse", but rather "no, they use science, so they need to do the science correctly". But they will ignore this argument ("they use characters, so they need to do characters correctly") is the case of Egan.
(I had thought Wells' relationship between Morlocks and Eloi might've been a reply to Kipling's Sons of Martha https://www.kiplingsociety.co.uk/poem/poems_martha.htm , but the former is from 1895 and the latter was not until 1907)
* although hard sci-fi has its own interwoven network of influences and references, so like capital-L literature and hip-hop, it really does count as a literature.
The Best of Greg Egan is a perfectly respectable place to start.