Rapid prototyping was always possible in PL design. It was very possible to go from idea to a working proof of concept language with a couple weeks' work. There are thousands of POC projects like this that popped up before LLMs existed.

What LLMs are doing now is allowing people to take prototypes and to publish them with an entire 200 page book no one (not even the author) has read, and a polished-looking website filled with marketing verbiage and a cute logo.

What would be interesting to me would be to see the process of rapidly refining the design, but I keep checking back on these "Here's my exciting new 400kLOC LLM language project I made in 3 weeks" and they all seem to die very shortly after the splashy announcements a few weeks later, as the author seemingly lost interest.

Which is not surprising because that's the way it always went with little languages -- writing a language has always been a marathon, not a sprint. It's just before, a 200 page book was an indicator of author dedication. Now, a 200 page book is just more bytes for digital kindling.