I think that providing a large corpus of example programs for training will become practically mandatory for any new language or framework in the future. At least that way you can help “jump-start” LLMs before it gets adopted widely enough for organic training material to emerge.