I wonder if Github Co-pilot or DALL-E type work could be expanded to generate ASCII diagrams (or art) based on descriptions provided.