A great programmer will invent things an ordinary programmer would never even think of. This doesn't mean a great programmer is infinitely more valuable, because any invention has a finite market value. But it's easy to imagine cases where a great programmer might invent things worth 100x or even 1000x an average programmer's salary.
The job of someone trying to rebut this statement is to foreclose on the idea of imagining that a great programmer might invent something worth $10,000,000 (or that an "ordinary" programmer could do that while leaving their "ordinary" status intact). Graham didn't say this happens every day; he said "it's easy to imagine cases". And, obviously, it is.
Scale the hyperbolic 100x down to Fred Brooks' 10x. The rebuttal is still wrong. It compares ordinary programmers to Zuckerberg and Andreesen, on the premise that those are two programmers worth >100x. But nobody, including Graham, has said those two are worth more as programmers.
The telling comparison is to John Carmack. "Nobody would ever hire John Carmack instead of 100 programmers". OK. But they might hire Carmack instead of 10 developers. That's because there probably is such a thing as a 10x developer.
I do not like or for the most part agree with Paul Graham's immigration essay. I think regional labor stickiness is a much more reasonable problem to solve than immigration. Our industry should get better at sourcing talent from Tulsa before lobbying to let more talent in from Krakow. Sadly, I haven't yet read a lot of strong rebuttals to Graham.