Very interesting project you have outlined.
I suspect the difficulty in creating a new programming language using AI is a bigger problem than the auhor imagines. I am far from expert in this but have been a user of languages for many years.
Language design involves a lexer, parser and compiler among other things.
Basically programming languages begin with some specific ideas or concepts. Variables, data structures like lists and for loops. All languages serve a general purpose in defining instructions for the computer to perform in a way that is easy for humans to understand, reason about, write and debug.
The more recent new languages like Go and Rust were originally designed to suit specific types of use. Rust has found use more or less as a system language and has grown into a more general purpose language. This is somewhat like Go in that it was designed with specific attributes in mind like simplicity and efficacy and then grown into a more general purpose type however in this case more suited to web applications.
Languages are characterized by several different features or usage patterns and similarities. C, C++, Smalltalk, Pascal and Erlang are usually considered in terms of syntax, i.e. C-like or Object Oriented.
Some languages were developed in what might be considered an academic or ivory tower type approach and were not successful when applied due to this lack of usage considerations. Other languages were developed in studiouly non-academic ways by developers just trying to get things done.
I think defining the goals of the new language and the use cases will turn out to be the most difficult parts. Should the language be simple to use, easy to learn, comprehensive, object-oriented, include a simple C-like syntax and useful in every kind of application area? Unlikely as some of these are mutually exclusive.
It would certainly be an interesting project and I would love to hear the results.