Is fundamental CS about topics in which you can't make breakthroughs?

Fundamental CS gives you the context needed to make meaningful improvements, but also to understand how systems work and identify where they can be improved.

My PhD work is emblematic of this. I created an algorithm (PGE) that improved the SOTA in Symbolic Regression by 1000x+ by stitching together other algos (bfs, A*, nonlinear regression, CAS, memoization, NGA). Without the fundamentals, this would not have been possible.

I also find the fundamentals invaluable in understanding where bottlenecks are in production systems, which can appear anywhere in the stack from software to hardware

In terms of making progress, we think of the breakthroughs mainly, but daily, incremental progress is made. Look at almost any changelog in open source

Fundamental CS is only interesting because it has been broken-through so thoroughly that you cannot possibly optimize it any further. Transistor logic, floating-point standards, B-tree algos, bitwise operations and language syntax are well-explored technologies at this point. If better alternatives existed in-theory, we'd be teaching them instead.

It's mostly just the virtue of a public shared asset that the most demanded optimizations are done the earliest. If I wanted to push the state of the art forward, I'd have to move halfway across the country, get a new major, and then get paid peanuts to work in a TSMC factory. No thanks, I'll keep my Herman-Miller and Starbucks allowance working on boring old software.

There's still plenty of theoretical CS research going on (especially in cryptography and quantum) but it rarely has impact outside academia.