My early take on the LLM coding wave, maybe a year ago, was that it would be the death knell of language development. It may well be still (for all kinds of direct and indirect reasons).

However! What’s a good way to keep the slop machines from going off the rails in your complex project? Give them a little DSL with targeted, constrained capabilities. And what does an admirable job at building those? Yes, the slop machines (with engaged human guidance).

Let a thousand little
languages bloom.