A lot of what was said can be inferred from the comments left in the code. Would be interesting to see whether it would produce the same description when all the comments are removed.
Basically, I don't think ChatGPT has a grasp of how INTERCAL actually works, and is just hallucinating a lot of it.