Also this "better than native" language needs to go die in a fire. Better than Native + crappy TAA maybe, but not better than native with an actual functional AA solution associated with it.
'Native' is still just a bag of tricks to mimic what you would see in reality, at an acceptable rendering cost. It's not actually what you would get if you'd film reality with a camera that has the same resolution as your screen. Anti-aliasing is just another trick, but one that is very expensive if you do it well.
Temporal upscaling actually does something similar to MSAA, in that both use multi-sampling of higher quality source material to produce better detail. However, MSAA is more like FSR 1 as it doesn't use previous frames. If you use DLSS/FSR 2 without any upscaling, it effectively becomes a relatively cheap anti-aliasing method.
Ultimately, you can of course do better if you give the PC 10 minutes to render each frame, rendering at 16k and then supersampling that to get top-tier anti-aliasing, but you don't get a playable game that way. Ultimately, it's all about doing as much as possible with a certain rendering time budget. My impression is that at least Nvidia thinks that people are going to prefer upscaled raytracing over having a rasterized 'native' image in the future, at the same FPS (and thus rendering budget).
Whether that is true remains to be seen. It's always hard to know what people prefer, especially since we don't necessarily prefer the most realistic image.