Reviewing is on average about reading bad papers that won’t get accepted in great detail to try to figure out what’s actually going on.
At best, it tends to be reading solid work adjacent to your subfield which you can respect but aren’t really that into.
It’s pretty rare for it to be as useful to me as actually choosing something to read.
The new LLMs are smaller and more likely to run on hardware available in your country. After all they were developed despite the US chip embargo.
This also means the US might slow down on the new power stations and resulting climate change, but probably not.
Not sure the new tech will actually be useful though.