I realize that you are asking a rhetorical question, as the answer is supposed to be a bit obvious and maybe even mind-blowing... but I actually struggle to come up with really good examples. Most of what I come up with are places where math struggled to keep up, notably in cryptography and consensus (and thereby, quite powerfully, in the intersection of these we have cryptocurrencies). Oh, here's one great such example: compression algorithms!!

However, the vast majority of the crazy things I feel like I'm able to do in software today aren't really because the software is--or even the protocols are--better, but that I'm capable of executing so many instructions in so little time on such vast quantities of stored data while communicating massive results over large distances reliably using computers that are so cheap and so small that virtually everyone in every financial class is surrounded by them.

Hell: even with large language models and the recent birth of working/useful AI, we are right on track for the timeline for progress along the road towards the singularity laid out in the late 1990s by Ray Kurtzweil, who was merely looking at computation per dollar, working off of the expectation that if the hardware can do it the software will figure it out. I could go back in time with the code for a modern LLM and I'd still have to wait for 2020 to deploy it.