The Onion had a story up about Gorbachev when he died recently: “Mikhail Gorbachev, Soviet Leader Who Took Down Iron Curtain, Dead At 91”<p>Nothing satirical about it (as far as I can tell), and there were plenty of other stories about this shared on Hacker News, but this one still counts as satire in the game because it was on the Onion.
Satire seems to be from the Babylon Bee (<a href="https://babylonbee.com/" rel="nofollow">https://babylonbee.com/</a>) and The Onion (<a href="https://www.theonion.com/" rel="nofollow">https://www.theonion.com/</a>)
I did ~22 prompts. All of them I correctly identified as satire. I don't know how they are generated. Feeding titles into a language model? Very likely, because the resulting titles are very easily distinguishable as satire given the context.
With Hacker News containing mostly user submitted links, and those having target-provided descriptions, the web site is really a "real web site title or satire" quiz.<p>IOW, nothing particular about HackerNews, even if it uses HackerNews to aggregate the non-satire headings.
Uh, something is broken because I was given all the satire headlines before the HN headlines.<p>Also I didn't see any credit given to the Babylon Bee where all the headlines are from, although maybe it just wasn't prominent credit.
It seems that when one of the words starts lowercased it’s always real/hacker news. All satire strings seem to be created with uppercased words only.