Comment by minimaxir

Those are just simple prompt examples: obviously more prompt engineering would be necessary.

However, modern LLMs, even the cheaper ones, do handle the up to X constraint correctly without consistently giving X.