There are several technologies that can make transistors that are too expensive to pursue in earnest until 2nm reaches commodity pricing which means multiple companies making fabs near that performance point. That will be a while probably, but it will be inside a hundred years, unless we end ourselves. Some of those technologies may exceed silicon performance and allow our continued growth. After 2nm hits commodity pricing, but before graphene or diamond or whatever hit feature parity, I wouldn't be surprised to see more intrinsics popping up to support common usecases in devoted silicon, but that just makes more work for you as a compiler developer.

I do think in the nearer term, computing will get less abstracted and people will spend more time on optimization. Not because of moores law, but because the demand for new chips is far exceeding supply from bleeding edge fabs, and it will take a long time to catch up. This may get exacerbated by politics and shrinking manufacturing workforces. So we will have to do more with older tech. These exotic architectures the author posits have to get fabbed by someone, and they won't do that while they can use existing designs to print money.