A Turing-complete language is just that, Turing-complete. Sorting used to be hard. It used to be novel. It used to be something you could study. Now it's something you're expected to just know. B-Trees, A* pathfinding, mapreduce, etc. were all, at one time, cutting edge computer science. Now they're the basis of it.
Knowledge breeds knowledge. And every piece of knowledge has the potential to open entirely new pathways of knowledge. Apple Pay depends on NFC depends on RFID depends on radio technology, etc.
It sounds like you're dismissing Turing complete languages as basically the same. I think that's kind of like saying all hardware and networking is just made out of electric circuits. It's true, but it's not illuminating.
> Sorting used to be hard. It used to be novel. It used to be something you could study
Quicksort was created in 1959 [0], Heapsort in 1964 [1]. Mergesort is from 1945 [2]! They were something undergrads were just "expected to know" shortly after that. There have been incremental improvements since then, but these things haven't been novel in a very long time.
I'm sure you won't change your mind, but this doesn't really support your argument above, "Compare programming in the 1970s and 1980s to now. We're way more sophisticated now than we were then."
> B-Trees, A* pathfinding, mapreduce,
Again, 60s and 70s ideas applied on modern hardware. I'll spare you the Wikipedia links for these.
[0] https://en.wikipedia.org/wiki/Quicksort
Before 1959, Quicksort was novel. After that, it was something you were expected to know.
Knowledge builds on knowledge. I don't have to invent heapsort. I can just study the implementation. It's now part of the common knowledge we're supposed to have.
And if you think programming is just as sophisticated today as it was in the 70s and 80s, then I don't think there's anything anyone can say to you to convince you otherwise.
And that's kind of the defining feature of being Turing-complete. No language is more or less adept at solving problems as any other as long as they can replicate a Turing machine.
I always thought so too, but this discussion makes me think otherwise.
I suspect this thread will fade without much more notice, but I'm half expecting someone to come along and set me straight that all this stuff was really invented in the 50s or 60s :-) (SNOBOL, Algol, Simula, or PL/I ???)
I think this is revisionist history at best. Your comments about baud rates and CPUs getting faster is true of course, but look at something like programming languages: I'd put ML from 1973 (later Standard ML, 1983) against Go or Rust, and I'd put something like Icon from 1977 against JavaScript or Python. If anything, I think we've regressed because ML and Icon could actually run on 70s era hardware, but Rust can barely compile its own compiler on 2020s era hardware, and Python wastes CPU and memory as though it was a design goal.
Seriously, look at Standard ML, Icon, or even Scheme and SmallTalk from the 70s, and ask how Python and Go improved on those.
https://en.wikipedia.org/wiki/ML_(programming_language)https://en.wikipedia.org/wiki/Standard_ML
https://en.wikipedia.org/wiki/Icon_(programming_language)