The fact that in 2020 we still don't really have CPUs that are significantly faster than those released in the middle of the decade in terms of single threaded

The fact that in 2020 we still don't really have CPUs that are significantly faster than those released in the middle of the decade in terms of single threaded perf (which is still the paradigm that most programming languages today operate under for the most part) kind of put a nail in the coffin of that prediction. I can totally picture new languages as described doing well today if we had exponentially more single threaded processing power/watt at our disposal than we actually do. I think there is still opportunity for languages that offer huge leaps in expressiveness w.r.t handling concurrency even at the cost of raw single threaded performance. Though I would not label a language like that as "slower" as it'd allow us to actually make much better use of our computing resources than we reasonably could today without blowing through most of our complexity budget, resulting in faster programs in practice.