I would want to go further and use domain specialized analyzers and type systems that are extensible with 'design time' information, ala F# type providers. It is indeed still a research level project that needs multi-national corporation level funding and I doubt that'll happen any time soon. I think it's more likely that a programming languages designed for use by LLMs will spur some investment there - type system ergonomics are often limited by latency and LLMs can be both be more patient as well as learn not to rely on it as often.

EDIT: Oh Hi! I didn't realize you're from the Luau team, cool, while Roblox isn't as big as Microsoft I wonder if it's a large enough multi-national to make a real dent in the research, I do hope you'll get your wish. I know I don't have the resources to take on such a project myself. I do type directed code generation which is about as close to the sun as I can safely fly.