To be fair, they're probably more correct in this context. Even if you understand the basics of compilation (and by the way, if you're reading this comment and you don't have a basic understanding of how compilation works, you should go learn that. Like, right now, if you can) the chance of you actually understanding the entirety of GCC and/or Clang without taking three years to study them is pretty slim, AFAICT.
I could try and run a marathon but I'm very out of shape. One of my friends can try and run a marathon who isn't out of shape and while they won't win they can still do it and learn from the experiance. The goal isn't to win, it's to learn in this case.
That is, your friends' awe of the compiler is not only justified given their skill level, it's also likely justified at your skill level as well (although as I don't know you personally, I can't say for sure).
But this is different from having no conception of how compilers work and how one might implement one. As complex as modern compilers are, the foundational principles still apply to them, and an understanding of these principles makes them much less awe-inspiring. I think that having a clear view of the big picture, rather than having certain areas of the picture blocked off entirely as magical black boxes, allows for better engineering choices.
It's like when I go to my mechanic. I pop the hood, he looks at it, spouts off some Geordi LaForge-sounding nonsense, I pay him and I walk away feeling like I've made out like a bandit! "That moron! That stuffs so complciated this work was worth double that! Hahahaha!"
What's really happening is that my mechanic has touched under the hood and no matter the car he things "eh, I'll figure it out". Or "How hard could it be? I'f fixed thousands of cars and thousands of problems, I know what's wrong already!"
He gained that experiance from actually messing around under the hood. Not from driving the car at a high level like I have but from sticking his hands in and getting into the nitty gritty.
Going from the top down leaves people affraid to push further down into their machine. This is the fear of infinite complexity that they think is hiding under the surface. The truth is that all computer science material of of constant complexity when looked at on it's plane of theory. IF everything has been correctly abstracted and defined then nothing should be "complicated" but it should be "entierly understandable". Anyone who gets the basics should be able to go "eh, I'll figure it out" when referencing any part of the computer if it has been correctly implemented.