Hacker News Reader

This would be such a great use case for an LLM.

Maybe for one shots, but not for continuous integration testing. Don't LLMs generate different responses for each request?

Why?