Introducing Halupedia: The Wikipedia Clone Made Up Of AI Hallucinations

Not long ago, we spoke about Wikipedia’s legacy and how it has been standing all these years – even in the AI age. Wikipedia has always been used to find facts, fandom wikis for television shows and Reddit for arguments.

Now, the internet also has a new platform: Halupedia, a website where every article is invented by AI the moment somebody clicks on it.

The project is referred to on GitHub as, “An infinite, hallucinated encyclopedia. Every link leads to an entry that does not exist yet — until you click it, at which point an LLM pretends it has always existed and writes it for you, in the deadpan register of a 19th century scholarly press.”

That means nobody writes the articles beforehand. A visitor clicks a made up topic, the AI creates an entry instantly and the site stores it for future readers. The articles read like old academic texts, complete with fake footnotes, fictional historical events and imaginary scholars.

According to Wikipedia, the site was created by Bartłomiej Strama, a student at Zespół Szkół Elektryczno Mechanicznych in Poland. Wikipedia calls the platform “an AI generated online encyclopedia, made entirely of AI hallucinations.”

Cybernews, the publication, tested the service using prompts around things like cybersecurity and genealogy, then explored fabricated entries generated on demand.

The humour comes from how seriously the articles present complete nonsense. One article mentioned in the Cybernews report covered the “Great Pigeon Census of 1887”, which supposedly counted “every gold crested rock dove within the administrative boundaries of the United Kingdom of Great Britain and Ireland.”

How Does The Website Actually Work?

Halupedia runs on Cloudflare Workers and uses OpenRouter to access LLMs. According to the GitHub documentation, the site operates as a single page React application designed to resemble an old printed encyclopedia.

When a reader opens an article that does not already exist, the system checks if the page has been generated before. If it has, the article loads from Cloudflare KV storage at no extra cost. If the page does not exist, the AI creates it live.

The GitHub page explains the process in detail. It says the worker “calls an LLM via OpenRouter that returns an HTML article in the encyclopedia’s voice – full of confident, plausible sounding nonsense that is densely cross linked to other entries that also do not yet exist.”

Readers begin seeing the article before generation fully finishes. The site splits the stream into two tasks. One sends text to the reader instantly, and the other stores the finished article permanently in Cloudflare KV storage.

The project creator also built systems to stop runaway server costs. The GitHub page says Halupedia blocks many automated crawlers and limits how many articles one IP address can generate per hour. A global daily cap of 5000 generated articles also exists.

Cached articles stay available to everybody, including bots and search engines. The GitHub documentation says, “The bot guard only fires on uncached slugs, so anything you’ve already paid to generate stays freely indexable for SEO.”
 

How Does Halupedia Stop Its Fake History From Contradicting Itself?

This unique feature on Halupedia is its attempt to maintain fictional continuity across thousands of made up articles.

The GitHub page explains that every generated article leaves hidden “link hints” for future entries. When the AI inserts a link into an article, it also attaches secret metadata describing that fictional person, event or institution.

Those hidden summaries enter a database before the visible article reaches readers. Later, if somebody opens the linked page for the first time, the AI receives those earlier descriptions as background lore.

The GitHub documentation explains the idea like this:

“The result is a write forward consistency mechanism: each article seeds breadcrumbs for the entries it links to, so by the time those entries are written, the LLM has a small dossier of established lore to honour.”

That system helps Halupedia avoid obvious contradictions. An invented historian mentioned in one article should keep the same biography later instead of suddenly changing careers or centuries.

The site also generates fake online identities for commenters. According to the GitHub documentation, “The first time you post, the LLM hallucinates a name and username for you.” Users do not register accounts manually. The AI invents names automatically and stores them using browser cookies.

Comments come up in threaded discussions similar to Hacker News. Readers can upvote posts, reply to each other and continue expanding the strange fictional universe surrounding the articles.

Is Halupedia Useful Or Simply Internet Entertainment?

Halupedia openly admits that none of its articles are trustworthy. The GitHub page says, “There is no editorial staff, no truth, no warranty. Every article is invented on demand. The footnotes are also lies.”

That honesty is probably why the site has gotten so much attention. Most AI hallucinations come up accidentally with chatbots or search tools. Halupedia turns the hallucination itself into the attraction.

Cybernews found the experience amusing in short bursts, though the publication also criticised racist and distasteful content appearing in sections of the site. The article said developers “need to sharpen their approach to regulating prejudicial content.”

The platform launched while discussion about AI generated media and misinformation online continues to grow. Halupedia does not present itself as factual knowledge the way Grokpedia does. Instead, it behaves like an experiment in collaborative fictional history written instantly through machine generated prose.

That makes the website more of an interactive comedy project than an educational encyclopedia. Readers visit to see what bizarre invention appears next, then follow chains of imaginary references through an endless library of fake scholarship.