Wouldn’t exactly take the comment as negative.

The output of current LLMs is hit or miss sometimes. And when it misses you might find yourself in a long chain of persuading a sassy robot into writing things as you might intend.