This is actually a pretty interesting question. Here is the novella length version
In the 360 era, games got better much better looking over the course of the generation. There is still a legacy meme from that era, of developers "learning the hardware," but to a great degree that isn't true anymore. Yes, developers can still learn the ins and outs of a particular system. But during the 360 era, developers were learning to use a whole new class of hardware - programmable GPUs.
This was still new territory for home machines, and devs were digging up research papers from Pixar that had once been limited to supercomputer class machines and implementing them on home consoles. Devs were inventing whole new rendering techniques. That whole movement sort of peaked when devs started to do
PBR at the end of the 360/PS3 era and at the beginning of the PS4/Xbone era.
Since then, new
software technologies in rendering having been progressing at a much slower rate. There is still fantastic work happening, but the generational leap from technique alone on that hardware has stopped. But that's not the only hardware on the block.
When most folks (including us) say DLSS, they usually mean "DLSS Super Resolution", the upscaler. But DLSS is a
suite of tools all of which do what Nvidia calls
Neural Rendering - basically, using AI, and the tensor core hardware inside Nvidia GPUs, to render pixels in a totally new way. And this is an era where it really is possible to see huge leaps, and there have been two big ones already.
While we're not likely to see the ability to "download more performance" via firmware, DLSS upgrades do open up the possibility of big leaps over the course of the Switch 2's lifetime. And unlike some past leaps, DLSS is a library that Nintendo and Nvidia will provide integrated into the SDK. Meaning that updates to DLSS can make it into games much more quickly, meaning that I think we will see games get big performance patches for as long as Nvidia can make them.