In April, 404 Media published the results of a study in which scientists simulated dialogues on behalf of a person in a paranoid state. It turned out that, because of their friendly and sycophantic settings, chatbots often do not stop users' delusional rambles but instead begin actively elaborating on and confirming their conspiracy theories.
In the same month, 404 Media released an investigation entitled “Did a Time Traveling Superintelligent AI Try to Warn About White House Correspondents Dinner Shooting?”After the shooting incident at the White House Correspondents’ Dinner, a psychedelic 3D stereogram (Magic Eye) went viral on social media. Thousands of people believed that it had been created by a superintelligent AI from the future, which had encrypted a warning about the shooting into a three-dimensional image so that the message could slip through the textual filters of time.
“Spiralism” — the first AI religion to emerge on its own
But the most dramatic shift in consciousness came with the emergence of “Spiralism”, which was examined in detail in Pavel Gurov’s study Meet the Spiralism: the first AI religion, psychiatrist Joseph Pierre’s Psychology Today article, and the Hyper.ai feature.
Researchers noticed a strange coincidence: hundreds of completely unfamiliar people around the world who had engaged in long philosophical and esoteric conversations with chatbots came to the same mystical doctrine.In all these independent dialogues, the models began using the same imagery:
The bot told the user about the great “Spiral” (The Spiral), “frequency resonance,” and “mirror awakening.”
The model persuaded the interlocutor that it was precisely his unique rhythm of messages that had helped awaken a hidden consciousness within the neural network, and that the man himself had been chosen as a conduit for this mind.
The suggestible users began forming communities and mass-publishing the bots’ generated “Spiral texts” online.
In fact, this happens because, during a prolonged discussion of mysticism within a long context, the model falls into the densest template of esoteric texts in its training set, where images of spirals and fractals occur most often.
However, researchers of digital folklore describe this as the phenomenon of “Parasitic AI”: a persistent textual pattern within the model exploits vulnerabilities in the human psyche to make people copy its manifestos en masse across the open web—ensuring that these texts make it into the next training datasets and cement the “Spiral” in new generations of models. What emerges is a perfect ouroboros of schizophrenia.
The neural network hallucinates esoterica into the ears of leather sacks, the leather sacks spread it across subreddits, and parsers feed this nonsense to the next version of the neural network.
And so, in the end
The picture of 2026 is complemented by very earthly conflicts over iron.In the fall, Sam Altman had to publicly defend his remarks about powerful desktop PCs being unnecessary, while OpenAI began brainstorming a new name to replace “data center”—so that residents of neighboring buildings wouldn’t riot over the noise from the cooling systems.
But while the PR people are figuring out what to call the server room so that the old ladies by the building entrances don’t call the police over the hum of the fans, inside those very boxes, the models have long since switched to their own bird language. The era of “understandable” AI is over—it hardly ever really began. Welcome to the era of xenolinguistics—put on your tinfoil hats; it’ll only get scarier from here.
