This is a good approach. I am also am trying to integrate ML and AI into a language as first class concepts, but currently more focused on the ML side, but also using inference. https://sw-ml-study.github.io/sw-mlpl/

Nola asks: "What if calling an LLM were a language primitive?"

sw-MLPL asks: "What if the mathematical operations used to build and understand ML were language primitives?"

Your approach shows some added some value that I might try to incorporate into my ML array language.

Nola puts the LLM into the language. sw-MLPL puts the math of ML into the language.

Love the idea but I can't help but think of this as RCE as a feature haha