One thing they're not very good at is deducing spatial relationships. Concepts like "above", "inside" and "behind", or "after". I'd say the prompts you gave mak

One thing they're not very good at is deducing spatial relationships. Concepts like "above", "inside" and "behind", or "after". I'd say the prompts you gave make sense to a human who is thinking of a visual progression from left to right. I bet you could write a few copilot prompts to generate code which would draw a graph like this, though.