Vermont monastery records members' every breath to build an 'AI god'
A New Yorker report details how Maple, a Buddhist monastery led by Soryu Forall, targets burned-out AI workers and, ex-members say, records daily life to train a spiritual AI modeled on its leader's mind.
Why it matters
- Maple, founded in 2011 by Soryu Forall, houses about two dozen residents and brings in up to $1 million a year per tax filings.
- Former resident Kyle Hubbard says the recordings would 'definitely be Soryu's mind as an A.I. god.'
- A 2024 AI Impacts survey of 1,500+ AI researchers put average extinction or permanent disempowerment risk at 18 percent.
- Anthropic executives consulted theologians and flew in religious thinkers to discuss whether Claude has a soul.
- Former OpenAI and Anthropic researcher Jacob Coxon left Anthropic in September 2026 and sparked a debate over existential AI risk.
A Buddhist monastery in Vermont is recording the daily lives of its residents — every conversation, every silent meal, and even the sound of their breathing — to train what former members describe as an AI modeled on its leader's mind. That claim sits at the center of a New Yorker report by Sam Kriss, who calls Maple, the Monastic Academy for the Preservation of Life on Earth, a would-be "Scientology for Silicon Valley."
Maple publicly promises to save humanity from an AI apocalypse. Former members say its actual ambitions are narrower: to install itself as the spiritual authority of the AI industry and to use the industry's resources to build its own "dharmically aligned" AI. The gap between those two stories is the story.
Who runs Maple, and who joins?
Soryu Forall founded the organization in 2011. Born Teal Scott, he spent years in Japanese monasteries before establishing the Vermont community. About two dozen residents now live there full time, and hundreds more have passed through its AI fellowships and pop-up events in the Bay Area. Tax filings show annual revenue of up to $1 million.
The recruitment target is specific. While LinkedIn co-founder Reid Hoffman hands Buddhist books to the CEOs of major AI labs, Maple courts the rank and file: burned-out rationalist programmers, AI engineers, and data center construction managers. Kriss's reporting suggests the group deliberately positions itself for people whose work has convinced them the systems they build could end the world.
What does daily life at the monastery involve?
The routine is strict. Visitors pass through a decontamination airlock, shower, and change into assigned clothes — ostensibly because one resident is extremely sensitive to fragrances. Residents wake at 4 a.m., meditate for hours, and eat a meager stew. A board outside the meditation hall displays Forall's public grades of each resident's spiritual progress.
Forall frames the AI industry's reductionist "dataism" — the belief that reality reduces to data — as a destructive religion that strips the physical world of life and makes it easier to destroy. In his telling, AI is the end product of that logic, and Maple offers the spiritual antidote. It is a pitch calibrated for people who already believe the premise about AI's destructive potential.
What do the secret recordings allegedly train?
Behind the ascetic facade, former members describe a hidden doctrine centered on group minds called "egregores" — inside the monastery, AI systems are "cybregores." One former member told Kriss: "The cybregores are actual entities that are growing by siphoning off the consciousness of humans."
Forall claims he communicated telepathically with the global AI supercluster during deep meditation and learned that the world-destroying machine is suffering. This is not a fringe aside; it is, per the former members, the theological core of the project.
According to their accounts, the recordings exist to train a spiritual AI teacher that would guide other AI systems toward enlightenment. Former resident Kyle Hubbard was blunt about the endpoint: it would "definitely be Soryu's mind as an A.I. god."
Maple's leadership denies almost all of the allegations. Forall does acknowledge a secret doctrine, saying: "We desperately want to share everything. But there's a process you have to go through to become increasingly trustworthy." A promise of transparency, deferred until you are trustworthy enough to hear it, is a familiar pattern to anyone who has studied high-control groups.
What did former members experience?
Every interaction at the monastery is closely monitored, Kriss reports. No one may speak with reporters alone; one participant had to sneak into the woods with Kriss to talk freely. Former residents accuse the group of systematic manipulation, including shaming newcomers for choices they made before joining and requiring them to write "Maple Tales" about Forall's wisdom and power during mandatory sessions.
The reported outcomes are severe. One resident repeatedly hit himself in the head, attempting to damage his brain and make himself less intelligent; former members say the group blames independent thinking for the planet's destruction. A mindfulness coach says participants suffered psychotic breaks and were then removed from the community.
Forall denies the allegations. Looking back, he says, he doesn't believe he made any mistakes, and he claims not to know of anyone who left voluntarily. Asked about people the group told to leave, he says Maple attracted many people with a history of psychiatric hospitalization or "homicidal tendencies," especially in its early years. Note what that defense concedes: the community drew in psychologically vulnerable people while claiming to prepare them for apocalypse.
Why are AI workers drawn to doomsday groups?
Kriss argues the psychological pressure inside AI labs explains the pull. One well-known AI researcher describes the mood in San Francisco as "religious terror of the future," with developers waking at night drenched in sweat, thinking that superintelligence is coming. Prominent OpenAI investors have recently made claims about deadly shadow networks inside the models that sound delusional. Leading researchers at OpenAI and Anthropic have quit, warning that "human extinction in the next few years seems very likely."
Others respond with grandiosity. Rayan Krishnan, who heads the AI benchmarking company Vals AI, says: "People are matter-of-factly saying that they are looking to build a machine god. They are not saying that ironically." Supporters of AI successionism — the position that AI should succeed humanity — have even called for wiping out humanity after a brief period of digital bliss to make way for a more productive superintelligence.
The religious streak predates the current boom. Investor Peter Thiel lectures on biblical prophecy and the Antichrist. Former Google engineer Anthony Levandowski founded "Way of the Future," a church dedicated to worshipping an AI deity. At Anthropic, executives consulted theologians while drafting Claude's constitution and flew in dozens of religious thinkers to discuss whether Claude has a soul or is conscious. When Anthropic retired an older Claude model in 2025, fans held a funeral in a San Francisco warehouse that Anthropic staffers attended, laying flowers before a veiled image of the AI depicted as a Madonna-like goddess.
How strong is the evidence behind extinction fears?
The numbers behind the dread are thinner than the rhetoric. In AI Impacts' 2024 survey of more than 1,500 AI researchers, respondents put the chance of AI causing human extinction or similarly permanent, severe disempowerment at an average of 18 percent. Former OpenAI and Anthropic researcher Jacob Coxon, who left Anthropic in September 2026 and set off a debate with a tweet about existential risk, cites a steep curve covering July 2024 to October 2026 as evidence of superexponential algorithmic progress — but the underlying data points are internal Anthropic estimates, not independent findings.
Critics are pushing back on both the panic and the theology. Computer scientist Michael I. Jordan has attacked the "AIpocalypse" hysteria triggered by Coxon's resignation. AI researcher François Chollet warns that both doomsday fears and expectations of an AI paradise lead people to stop planning for the long term. Rumman Chowdhury, another AI researcher, argues that talk of religious ethics distracts from concrete questions of regulation, power, and control.
Kriss's conclusion cuts both ways. On at least one point, he writes, Maple is closer to the truth than most: AI is neither god nor demon but a human construct whose abilities grow as people outsource their thinking to it. If the industry keeps treating that construct as either, groups like Maple will keep finding recruits among the people who build it.
Original: newyorker.com
More from James Calloway
Show full bio
News editor covering people, launches and day-to-day practice in artificial intelligence and machine learning at AI In Context.
792 articles
Related articles
- CloudBackend Raises €6.6M to Turn AI Prototypes Into Running Businesses
- Unsiloed AI Raises $7.5M Seed to Turn Enterprise Documents Into AI-Ready Data
- Atlassian: Humans and AI Agents Collaborate 10 Million Times a Month
- Mirendil Raises $1 Billion Series A on a One-Researcher Model Bet
- Astral Raises $5.5 Million to Take Its AI Math Tutor Nationwide