I'm... slightly dubious, honestly, just because, historically, confident predictions of "this is the future of programming!" are almost always wrong. Throughout the 90s and into the early noughties, say, there was a very, very strong idea that things like VB and Delphi and Frontpage and Flash, where code was essentially intermingled with the UI, were The Future. And then all of that just died, to such an extent that there's really nothing like it at all today. Then there was the whole UML thing, and the "everything will run on XML" thing in the mid to late noughties...
Sounds a lot like "nothing ever happens"?
Well, arguably in programming nothing ever does happen; we're still using mostly imperative programming languages on mostly UNIX-y systems which are not dramatically different from those in the 1970s, with filesystems rather similar to those in the 70s...
Some stuff does happen, of course, but most prophesied things do not happen.
I can see it being useful for summarization, or creative writing. What makes you so sure that LLMs will be useful _for programming_ in the long run?
Because they are already useful for programming and unlikely to get worse!
We're not getting anywhere with your answer.
All tools I mentioned before were useful for programming. They didn't got worse. Still not enough to keep them relevant over time.
I chose those tools as an example precisely because, for a while, they achieve widespread success. People made awesome things with them. Until they stopped doing it.
What brought their demise was their inherent limitations. Code by humans on plain text was slower, harder, but didn't had those inherent limits. Code by humans on plain text got better, those tools didn't.
That said, even if the specific products like Cursor or ChatGPT are not here in 5 years, I am confident we are not going to collectively dismiss the utility of LLMs.