I am thinking that the current languages are not ideal for the LLM "thinking". Perhaps a new, more LLM friendly language can be created, to increase the efficiency of automatic coding. On the other hand, there will be no initial training data in this language. Will that language still be easily readable to humans?

My more general question is how the appearance of LLM coding would affect the languages evolution?