I guess what I’m hoping is that the decline in processor speed growth will fuel research money into optimizing these things in addition to a general movement of developers back to languages that are more performant. I can also see a movement to whatever languages/paradigms that research winds up being successful for.

And you are 100% right that the biggest challenge in all of this is how you deal with memory. I think that requires whole new areas of research for programming languages and program analysis. We need tools that can intelligently break up memory. There will inevitably still be a burden on the developers to write code in a way that facilitates that break-up.

A fun question is how much of this can be done in static analysis and how much will have to be in runtime. The answer to that question actually has an impact on which way languages go. If most of this can be done with static analysis then a Rust-like model could be dominant. However, if this requires runtime analysis of where the data is flowing then that might boost VMs as they can monitor those things and optimize on the fly. I’m not even certain what I hope for in that regard.