[carl_sandland]

Maybe implementations have much less value than has been assumed and paid for by stakeholders over the decades. If code is a throw-away implementation detail of some model/abstraction then why are we getting paid so much money again? Really makes me worry about generative AI approaches; just as well there is no universal very high level modelling language everyone loves yet. I think this is why working on games is nice; it's not technical debt it's just a toy, maybe everything is.

Our cathedrals are surely [L,U]inux and the C programming language, HTML has done pretty well too.

[kqr]

Don't confuse short-lived for low value. Software solving the right problems has incredibly high ROI, even accounting for high programmer salaries.

We're not getting that money for generating realistic-looking syntax, but for discussing with stakeholders and choosing appropriate designs with an eye for both the bigger picture and details. These are language-and-library-independent things, mostly, and incredibly difficult work.

You may be able to give an LLM directions to do something like what you would have been able to do, but the money is not in the parts the LLM is able to accomplish, it's in the direction you give it.