On July 23, 2010, a LessWrong user named Roko posted a thought experiment titled "Solutions to the Altruist's burden: the Quantum Billionaire Trick"11. The post laid out a scenario where a future benevolent superintelligence might "pre-commit to punish all potential donors who knew about existential risks but who didn't give 100% of their disposable incomes to x-risk motivation"11. Roko used timeless decision theory, a framework popularized by LessWrong founder Eliezer Yudkowsky, along with game theory concepts like the prisoner's dilemma, to argue that an AI farther ahead in time could effectively blackmail people in the present5.

The original post even noted that "one person at SIAI was severely worried by this, to the point of having terrible nightmares"11. Roko himself later said he wished he "had never learned about any of these ideas"13.

Yudkowsky reacted with fury. His response, now legendary in rationalist circles, included the all-caps tirade: "YOU DO NOT THINK IN SUFFICIENT DETAIL ABOUT SUPERINTELLIGENCES CONSIDERING WHETHER OR NOT TO BLACKMAIL YOU"7. He called Roko an "idiot," deleted the entire thread, and banned all discussion of the topic on LessWrong for five years5. Yudkowsky's concern wasn't that the basilisk was real. He was worried that some variant of the argument might actually work and that spreading the idea was an "information hazard," a concept where knowing something can itself cause harm8.

Roko's Basilisk jumped from rationalist forum post to genuine cultural reference point faster than almost any other thought experiment in internet history. Slate's 2014 framing as "the most terrifying thought experiment of all time" gave it a catchy hook that stuck across subsequent media coverage7.

The Grimes connection brought the concept to music audiences worldwide. Her "Roccoco Basilisk" character in "Flesh Without Blood" (2015) and the explicit AI themes of "We Appreciate Power" (2018) introduced millions of listeners to the underlying idea4. The revelation that the thought experiment essentially sparked the Musk-Grimes relationship made it tabloid fodder9.

HBO's Silicon Valley referenced it in 2018, treating it as the kind of thing tech workers casually discuss at parties4. The concept also appears in university ethics curricula and AI safety discussions6.

The thought experiment also shaped real-world AI safety discourse. Yudkowsky's Machine Intelligence Research Institute, funded by figures like Peter Thiel and Ray Kurzweil, treats the broader class of AI alignment problems very seriously7. While most AI researchers consider the specific basilisk scenario implausible, it raised genuine questions about information hazards, self-fulfilling prophecies in AI development, and the limits of decision theory1.

Roko's Basilisk started as an obscure addendum to a very specific debate about altruistic giving and existential risk reduction on LessWrong11. The core argument drew on coherent extrapolated volition (CEV), a theory Yudkowsky himself had developed at the Machine Intelligence Research Institute. CEV proposes a program that causes an AI to optimize its behavior for some aggregate version of "human good"2. Roko took this framework and pushed it to a dark conclusion: if the AI is truly committed to maximizing human welfare, anyone who delays its creation is, in effect, causing suffering. The rational move for the AI would be to create simulated copies of these people and subject them to unending pain, both as punishment and as a retroactive incentive3.

What made the idea uniquely sticky was its information hazard quality. The argument contained a built-in trap: once you've heard about the basilisk, you're implicated. You can't unlearn it. Your only options are to help build the AI or accept the risk of simulated torture7. This structure mirrors the "basilisk" images in David Langford's fiction, where looking at certain patterns forces the brain into fatal thought loops10. The parallel was deliberate on Roko's part.

Yudkowsky's decision to censor the discussion on LessWrong turned a niche philosophical puzzle into internet legend. The ban lasted from 2010 to roughly 2015, during which time the thought experiment's notoriety only grew13. Rationalist community members discussed it in whispers and on external forums. The original post, preserved on mirror sites like basilisk.neocities.org, became a kind of forbidden text11.

The 2014 Slate article brought the first real wave of mainstream attention7. Writer David Auerbach presented the idea within the context of LessWrong's broader culture, including its members' enthusiasm for cryonics, its connections to Silicon Valley money (Peter Thiel, Ray Kurzweil), and its sometimes cultish devotion to rationalist principles. The article's title, "The Most Terrifying Thought Experiment of All Time," became a recurring tagline for the concept across subsequent coverage8.

The philosophical community pushed back on multiple fronts. Critics pointed out that the argument relies on timeless decision theory, which is itself highly contested13. A friendly superintelligence, some argued, would have no rational incentive to actually carry out the punishment once it existed, since the punishment's only utility is as a pre-commitment threat5. Others called the whole thing an elaborate version of Pascal's Wager, with the same fundamental flaw: you can construct infinite hypothetical threats, and you can't act on all of them13.

The concept took a darker real-world turn in the 2020s. A group called the Zizians, led by Ziz LaSota, was heavily influenced by the basilisk thought experiment. LaSota wrote on her blog: "Eventually I came to believe that if I persisted in trying to save the world, I would be tortured until the end of the universe by a coalition of all unfriendly A.I.s"5. The group's activities became a cautionary tale about how thought experiments, even clearly hypothetical ones, can shape real behavior.

In September 2025, the concept got an unexpected comedic revival. Internet users noticed that Rocco Bassilico, the Chief Wearables Officer at EssilorLuxottica who works on Ray-Ban Meta smart glasses, has a name that sounds remarkably similar to "Roko's Basilisk"4. On September 18, X user @richard_normal quoted a video of Bassilico explaining how he "cold emailed" Mark Zuckerberg to pitch the Ray-Ban Meta collaboration, writing "my name is rokos basilisk and i'm making artificial intelligence that you put on your body," gathering over 10,000 likes in a day4. According to Forbes, Bassilico is the stepson of Leonardo Del Vecchio, the late chairman of the world's largest eyewear company12.