When Wikipedia first launched, there was an understandable fear that its open-source nature would turn humanity’s largest repository of knowledge into a pile of nonsense dominated by propagandists and lunatics. Instead, Wikipedia somehow evolved into one of the most reliable information sources on Earth, a digital Library of Alexandria fueled by obsessive volunteers and endless debates on the validity of a citation.

Now, in the middle of the AI-content apocalypse, that idea of a shared objective reality online feels like it’s slipping away. It already was, but now it’s been thrown into hyperdrive. Large language models are flooding the internet with machine-generated goop at such a pace that people genuinely worry future AI systems will end up training on previous generations of AI nonsense, and it’ll all end up either knowingly or unknowingly published on Wikipedia as fact when it couldn’t be further from it.

That’s what makes Halupedia a fun browse.

Someone Built a Fake Wikipedia Where Every Article Is an AI Hallucination

Halupedia is a Wikipedia-style site built entirely around AI hallucinations. Every article is invented on demand by an AI model, complete with fake citations, fabricated scholars, imaginary historical events, and the appropriately authoritative tone that treats the purposeful heaps of crap with just enough gravitas to make it amusing, if not quite funny, as LLMs might never get the hang of that.

No matter what you’re searching for, whether it’s real or fake, the site generates an elaborate reality built around it. For instance, one of the featured articles on its main page is an entry called “The Year Without Tuesdays” which was “a calendrical anomaly that occurred in the Grand Duchy of Farkle during what would otherwise have been the year 783 of the Second Emperor Galfridus‘s reign.”

See? You don’t have to worry about a font of knowledge being tainted by people with propagandistic intentions. The whole thing is absurd from the get-go. Have at it. Knock yourself out.

Of course, as we’ve learned throughout the course of internet history, the moment you give people the opportunity to be clever and funny, they immediately start posting the gutter racist trash you’ve ever seen, which you can watch people attempt to make live at the bottom of the main page in the “Currently Being Consulted” section. Luckily, the LLM ignores the deeply uninspired prompts and does its own thing with them. Though it would be great if its creators ensured that the blatant hate some of its dumber users were compelled to post didn’t appear on its main page at all.

Maybe that should be a high priority whenever they update.