Dispatches from the present
In mid-October, I receive an invitation to the launch party for the Elephant, a new right-wing rationalist magazine. Its co-founders, Stevie Miller, Owen Yingling and Arjun Panickssery, are all rumored to be remarkably young: under 25, some of them practically teenage prodigies. The Elephant’s home page reads, simply, “Good things are good.” Its mission statement declares a belief “that institutions can be fixed, that empirical evidence should drive public policy, and that cultural confidence should rest upon liberal foundations.” Rationalists, the founders argue, “have proven better at challenging incumbent institutions, or shaping them from the inside, than conservatives. Yeah the New York Times sucks, but the New York Post is much worse.” What strikes me most, reading the full mission statement Miller shares with me ahead of the launch, is the weight they give to AI. “Right now, there is no institution that is both solidly right-wing and pro-safety,” they write. “We hope for The Elephant to become the most recognizable explicitly right-wing institution to also be explicitly rationalist-oriented on the threat of existential risk from AI.”
The Elephant launches at the Pierce School in Washington, D.C., a converted Capitol Hill schoolhouse whose penthouse belongs to the child-acto- turned-crypto-mogul Brock Pierce. I take the train from New York that evening and meet a group of writers I know from New York at PubKey, the “bitcoin bar,” for a drink. “One more drink and then we’re going to the Republican party,” one of the women is telling the group when I arrive. I drink a Monster energy drink that I procured at Union Station, and I change my shoes as one of the men’s Teslas self-drives us over to Northeast D.C. A party promoter I know from New York, Matt, who describes his mission as “making the right wing punk rock,” recognizes me when we arrive and pulls me past the coat-check gatekeepers, through a purple ballroom-esque living room lined with somewhat haunted-looking art and military paraphernalia, up a floating staircase, and toward a wall of fake bookshelves, which he pushes, finding give. “You’re with me,” Matt proclaims. “Leave your things in the secret room.”
Back downstairs, at the wine-beer-and-Pepsi-Zero-only open bar, the crowd is mostly men in suits. One of the only other women is a tech journalist I know from New York. “It’s so conservative here!” she exclaims. She rolls her eyes, laughs and runs off. I clutch my Diet Pepsi and head back up the floating staircase. I say hello to a lawyer from D.C., two of the co-founders of the magazine, the wife of a politician (the politician himself was disinvited “because of Antifa”), an infamous internet sex blogger and an abundance of young men who introduce themselves with acronyms, X usernames or the agency of the Trump administration in which they are employed. I ask a George Washington University freshman what he thinks of AI and he tells me he’s an accelerationist. “I’m an elitist,” he explains. I ask another man who tells me he works for the Small Business Administration the same question and he rolls his eyes. “Dude, whatever.” I ask people if they are rationalists, and some of them say yes. I ask a burly man who repeatedly tells me he works for “the president” about AI and he corrects me. “The president renamed it SI.” He tells me that in briefings with the labs he makes the briefers call it Superintelligence. “For three hours, they have to call it SI,” he gleefully exclaims. “They say ‘yes sir,’ and ‘no sir,’ and ‘certainly sir.’ Mostly, people seem to want to party, and so I stop asking much of anything at all.
Walking toward the mansion’s living room on the second floor, I meet a man named Andrew. Andrew has a background in tech and works in the Trump administration. “It’s not going to happen,” he says, after our conversation turns to various doomsday scenarios. There are too many variables, he tells me. The labs have too much financial incentive to sound the alarm for anyone to take the alarm at face value. The rationalists are too weird. The race with China is too tight.
“Listen,” he says. “It’s going to be crazy. Four really bad things are going to happen in the next two years.”
“Four really bad things…?” I ask.
He shrugs. An estimate. “But not human extinction,” he clarifies.
He explains that there are trillions of dollars on the table for the frontier labs ahead of their IPO or in a nationalization deal if the government gives in to their demands for pacing.
“Look,” Andrew says. “If we all die, at least we’ll die sticking it to the people demanding trillions of dollars.” He smiles, kidding, meaning to comfort me. I do not feel remotely comforted.
I leave the Elephant’s launch unsettled. The Elephant pitches itself as a home for the AI-safety right, but those celebrating its launch seem blasé, resistant to considering the very risks the magazine wants to take seriously.
On Monday, November 9th, Sentinel, a forecasting organization that tracks global catastrophic risks, puts its median forecast for artificial superintelligence at 2028, and the chance that humanity goes extinct within the next decade at 6 percent, with individual estimates running from 0.2 to 50 percent. On November 10th, Daniel Kokotajlo, formerly of OpenAI, and five co-authors, among them leading AI safety researchers, publish an analysis arguing that a new independent report on Anthropic “implies an AI-driven acceleration 5 months ahead of the fastest AI 2027 scenario.” AI 2027, a forecast cowritten by Kokotajlo, went viral in the spring of 2025. It imagined a race between the U.S. and China to build self-improving AI that comes to a head in 2027 and ends in one of two ways; a slowdown, or the end of humanity. The following winter, Kokotajlo pushed his timelines back. Now, he says, they’re running early.
That day, I meet Stevie Miller, the Elephant’s co-founder, for coffee at the Hungarian Pastry Shop in Morningside Heights. Over espresso, I tell him about the conversations I had at his launch, and about my concern that people in general, and in particular the people he most wants to reach (the right, his intended readers, the people currently determining policy on these issues), are not taking any of this seriously. Stevie nods. “This is the issue,” he says. “I want to show people that the stakes are very high, and there are real things we can do to lower p(doom).”
AI is meant to make up “25 or 30 percent of the content” of the Elephant, he tells me, “and then everything else is, like, clicks for that, basically. That’s what we care about. That’s the political project here.” Everything else, he suspects, “is going to be much less remembered by history.”
Miller identifies as a member of the effective altruism movement, which tries to do the most good per dollar, from malaria nets to mitigating AI risk. But rationalism, a movement devoted to the idea that everything—ethics included—is knowable through reason, “has much broader cultural ambitions,” he says. His working definition of rationalism is one he thinks should appeal to the right: “You want fewer things, on the margin, exempt from cost-benefit analysis.”
The right-wing rationalist tradition, as Miller tells it, lives mostly at George Mason University, in Fairfax. “Is this a George Mason project?” I ask. “Absolutely,” Miller answers immediately. “That’s half the board.” One of their board members, the economist Robin Hanson, he tells me, is “more on the accelerationist side,” but also “the most influential intellectual figure on us.” Of board member Chris Painter, the president of the AI research nonprofit METR, he says: “If you put Chris Painter in charge of an AI regulatory agency, I think it would go better than the nuclear regulatory agency.” (Miller considers the Nuclear Regulatory Commission “the worst thing in our civilization, man.”)
Miller tells me that his own p(doom) is “fairly high.” And yet: “We wouldn’t press the button to just have there be no AI,” he says. “We want to live forever. We want eternal fun. We want a much better society. There are a lot of opportunities from this technology. It’s going to be tremendous. But in order to do this, we need really good institutions. We need to internalize the risks of takeover.”
Miller hopes that the Elephant will be a home for a “more empirical, internet-influenced right,” one that “appeals to elites again.” A lot of effective altruists, Miller notes, are trying to win over the median voter on AI safety. “I definitely wish them luck,” he says. A lot of people are competing for the median vote, but the truth is that “the median voter is not stupendously influential on most kinds of policy … a lot of policies are just set by elites.” He thinks rationalism is more promising, since it has always been “this incredibly elite movement,” but one that appeals less to established elites than to “nineteen-year-olds who would go on to start companies and drop out, and who are now swallowing the rest of the economy.”
On a walk after our coffee, I describe the people I met at his launch: pro-industry and “based” and, when pressed, not certain that artificial intelligence couldn’t turn catastrophic. I tell him that I’m interested in his project because I’m curious how he’ll change these people’s minds. “I think you can get 30 percent of the way there by just having our institution,” he says, “that says, hey, look, we’re pretty based. … Also we like AI safety, and we’re right-wing, and we do parties.”
“If you don’t like safety regulations,” he adds, “I think you’d like our safety regulations better than other people’s.” The Elephant’s tagline is “Accelerate safely.”
Photo credit: Banksy via Dunk (Flickr / CC BY-NC 2.0)