We use additional compute power to enable frameworks and abstractions that allow code to be written faster. Sure it's bloated but at least there is a reason.
Yeah, this is my experience too.
I view it more as beauracracy; in order to get my job done I now also have to understand a whole lot of arbitrary systems created by a group of programmers on the other side of the world.
If you break out of the desktop/mobile environment and into embedded, it's genuinely liberating.
one thing to consider is that the types of people writing code are very different. In the 80s it took geniuses to write a simple videogame within ram constraints. Now that is a feasible final project for an intro CS class.
Electrical engineers still write incredibly low-level code in embedded programming courses - we did some pretty interesting shit in MIPS assembler before we even touched C++ - and I doubt we were any smarter than the CompSci kids.
I think it's primarily a cultural difference - modern CompSci people love treating everything as an academic exercise and hand-waving away complexity, whereas the old-school "hackers" (before CompSci was really a huge thing) would love getting into the nitty-gritty of things and didn't really give a damn about whether or not they were following best practices.
If the average CompSci student (who, to be fair, is smart but not Genius-level) spent as much time reading the spec sheet for the Motorola 68000 as they did learning how and when a Turing Machine halts, they'd have no issue understanding 80's video game code.
Yes, 100%. Without question. I know it's trendy to hate modern languages and frameworks but they do reduce overall dev time even if they have more of an initial learning overhead.