At the early 2000s I dropped out from my university studies for multiple reasons, but one of the major ones was a lack of motivation. I was learning theory, tons and tons of theory (and in quite a few cases, already obsolete), with some labs and assignments but usually exercises that were dull and boring. That's why I decided to start working meanwhileI studied (and finally, quitting before finishing the degree), to find that motivation and drive that I was seeking during my studies.
I've since then regretted not finishing, but not because I need the degree or anything like that. I think that a university degree shapes your brain, gives you some pillars over which to build everything else. You can learn about networks, data structures and operating systems on your own, but without the basics it won't be much different from driving a car without knowing how a combustion engine works: you use it as a tool but inside it is a black box. And I feel that in certain areas, it helps a lot to know how things work, not only how they are used.
Going back to the studies topic, one of the areas that I feel could be vastly improved is the examples and uses of each area. Sure I can memorize all the syntax and keywords and rules of ANSI C, but that way of learning C is hard and prone to me getting bored. It's way better to learn each keyword by doing some exercise, and maybe then try to build something bigger, and then something more complex, and so on. At one of my optional subjects I learned some OpenGL basics, and we were given an assignment of rendering a rotating 3D earth (with a single, nice texture). It really wasn't hard to build, but the "wow" feeling of watching that rotating 3D Earth was so great that I wished I had more assignments like it.
As I cannot go back in time and ask my teachers to give us harder but better examples and/or exercises, I'll instead just write a few cases that I came up with.
Learning about how operating systems work I almost only see two choices of the best and most rewarding and practical exercise: either learning MINIX, or directly learning Linux basics.
I was taught Windows NT back in the days and, while interesting on its own, I firmly think you should learn a free and opensource language available to everyone, and then decide and branch to wherever you want to go.
Logic gates, multiplexers, EPROMs, high/low voltages... There are lots of logic circuit simulators, but what I would have really loved to see is how to build an actual PC component, like a VGA graphics card. Sounds incredibly complex but simply watch those videos. Less than one hour and it will display a 100x75 32 color image via VGA cable. Simply amazing, and exemplifies that, while reality is way way more complex, the simple scenario is still cool (actually, it is awesome!) and the entry level comes to just knowing basic electronics and logic gates, nothing really advanced.
In this subject I did built some basic programs with Assembler, but it is a pity that you get to peek at so many concepts but only grasp them, having to just believe that your computer is simply a much more complex version of what you're told... or use one of those prehistoric 8086 emulators and just visually see all CPU registers and the like. If instead I was tasked with learning how the Nintendo Gameboy handled console hardware worked I would have seen a slightly more complex but still degree-level hardware design, complete with I/O, a rough graphics system, RAM and ROM...
You're learning how a compiler works, how to translate to machine code, building an abstract syntax tree, what's a symbols table, etc., so to me, the best example to properly learn the subject is to build a compiler for a simple language.
This one I actually semi-did: back in the early 2000s we were tasked with building a kinda-Assembler to real Assembler compiler in C. "Transpiler" would be more correct in today's terms, as the initial output was a .ASM, but still a good exercise.
I wish I knew more maths, but never take the time to dive deeper except for specific formulas. A good example to encourage someone to study them, at least for me would be building a pretty simple software-based 3D renderer, then adding some rough physics. Even if it's just rendering wireframe polygons without textures or colors, something like that would have already been quite appealing. In general anything game-development related ends up touching some area of maths so binding both seems interesting.
So there it goes, I'm pretty sure there are better examples, as mine are way too biased towards the topics I like, but... you can get the idea 😉