I was curious about your thoughts about the next generation of programming languages.
I got curious about this because of a.i. software,llm’s,a.i. models and their whole ecosystem etc…
What do you envisage will be the next generation of programming languages and how would they would work ideally in your opinion.
i am thinking machine → assemby → high level languages → ??
Any sci-fi you have in mind will be fine too :).
thanks.
This is a topic I like to read about and I will be following curiously here to see what others think…
I am, by all means, not an expert, but here is a list of things I think are making or might make a big impact:
- borrow checker (rust and others)
- borrow checker varieties like lazy clone in HVM
- concurrent ownership with arena allocation (verona)
- reference capabilities (https://ponylang.io/)
- hybrid borrow checking and garbage collection (like in v-lang)
- MLIR or similar techniques for heterogeneous computing
- Gradual set-theoretic types in Elixir
So, my dream language / runtime would be:
- actor model for concurrency (everything is a process philosophy)
- natively compiled zero cost abstractions and so on
- borrow checked with the least invasive approach (arena allocation? small GC only where borrow checking gets too terse?)
- reference capabilities to ease types in message passing
- preemptive small runtime (similar to pony but preemptive)
- targeting MLIR from the beginning
- devex from the BEAM ecosystem (hex, mix, rebar and so on… also, macros from Elixir
)
This would (hopefully?) yield a language that is really fast, easy to write for sequential and concurrent parts and that would be ready to take advantage of future hardware.
I believe the closest one to this small list is Pony but I think it will never aim to be preemptive as this is almost impossible to “add” later on in a language project. Also, it still has some way to go to have the same level of devex (package manager, project config and so on).
Anyway, just some random thoughts thown out here…