I guess my thought was that if they are moving beyond very limited use cases and into broader ones (e.g. communication, which almost certainly involves very bro

I guess my thought was that if they are moving beyond very limited use cases and into broader ones (e.g. communication, which almost certainly involves very broad allowable inputs) then the existence of provably correct programs for small problems is less relevant. But as I'm starting to understand, it may be that this is still a very limited case compared to the theoretical general case, even if it is significantly more general than e.g. proving a given program will halt.