The Fall of the ‘Nostradamus of AI’

By Berber Jin, Ben Cohen and Anissa Gardizy

For the past few years, at happy hours and dinner parties in San Francisco, Leopold Aschenbrenner kept confusing people by sharing his favorite outlandish idea.

He wanted to buy galaxies. Some of his friends weren’t always sure what to make of Aschenbrenner’s intergalactic ambitions. He once left the room and they debated whether he was referring to physical galaxies or a type of private plane.

No, he assured them when he came back. He meant galaxies.

His idea was that advances in artificial intelligence would soon unlock resources on a cosmic scale, enabling humans to colonize faraway planets. He planned to save money now so he could spend on galaxies, he told them, and make his mark across the universe.

To almost everyone in this world, outside the AI bubble of Silicon Valley, the idea of scooping up property rights outside of Earth might sound improbable. But in Aschenbrenner’s world, that sort of fervent belief was among the reasons the 24-year-old investor was hailed as a visionary.

Since graduating from Columbia University in 2021 as the class valedictorian at age 19, he has worked his way into the upper echelons of technology. He spent time working out of now-convicted

crypto founder Sam Bankman-Fried’s compound in the Bahamas and then at OpenAI in San Francisco. Even as he drew attention as a troublemaker among some of his bosses, according to people who worked with him, he grew close to the industry’s top players.

Two years ago, he made bold predictions about the impending arrival of superintelligence in a 165-page manifesto called “Situational Awareness.” He quickly turned that interest into money for a hedge fund that eventually managed $45 billion in assets.

He was barely of drinking age and had no prior investing experience, but his cozy relationships inside the industry fed into his reputation as an oracle of AI. He mixed the personal and professional in a city where the two are rarely kept apart.

“There are perhaps a few hundred people, most of them in San Francisco and the AI labs, that have situational awareness,” he wrote in his viral essay. “Through whatever peculiar forces of fate, I have found myself amongst them.”

When Aschenbrenner’s hedge fund unraveled last month, veterans on Wall Street saw it as the latest cautionary tale of hubris—another inexperienced, overconfident investor taking on too much risk with far too much leverage. Securities regulators are also investigating the trading activity of his fund, Situational Awareness, though they haven’t made any formal allegations of wrongdoing.

“We are a highly-regulated business and will cooperate to the fullest extent with any regulatory request,” the company has said.

But in Silicon Valley, Aschenbrenner’s stature has only grown among the elite group of researchers and executives that share his zealotry in a future dominated by AI.

From his mansion in San Francisco’s rolling hills and an office downtown, he had entertained well-connected guests and grew particularly close to employees at the AI giant Anthropic. Many of them were at his wedding earlier this month, when Aschenbrenner married the chief of staff to Anthropic CEO Dario Amodei.

But by the time he walked down the aisle, Aschenbrenner’s fund was in tatters.

Two days earlier, Situational Awareness had blown up.

A ‘tall poppy’

Aschenbrenner grew up in Berlin, where he distinguished himself as one of the country’s brightest young scientific minds. But he felt out of place, attributing his desire to leave Germany to his country’s “tallpoppy” syndrome—the idea that successful people get criticized and cut down.

“They really, really, really hate anyone that deviates from the norm,” he said in a 2021 podcast.

He started college at Columbia when he was only 15 years old. Professors remember him as one of the most brilliant undergraduates they ever met. While he blogged about students wasting time watching Netflix, he found something else to binge that would end up being far more important to his career: effective altruism.

Aschenbrenner grew interested in the ideas behind the philosophical movement, which aims to apply data and reason to maximize good in the world. One focus is how to reduce longer- term risks that people typically ignore, including AI that could go rogue and declare war on humans.

His junior year, Aschenbrenner passed out fliers for the Columbia Effective Altruism club at the school activities fair with his friend Collin Burns. (Aschenbrenner says he is not an effective altruist.) They soon organized a club outing to watch Scottish philosopher Will MacAskill, whose ideas drove the EA movement, deliver a Saturday night lecture called “The Ethics of the Next Billion Years.”

When he graduated in 2021, he wasn’t sure what to do next. Aschenbrenner was accepted to Yale Law School and seriously considered the offer. Instead, he went to the Bay Area.

Within a few months of graduating, Aschenbrenner was in a group chat with Sam Bankman- Fried to discuss contacts in the Bahamian government, as the founder of the cryptocurrency exchange FTX was headed to a bitcoin mining farm in Kazakhstan.

Aschenbrenner had joined the FTX Future Fund, a philanthropy backed primarily by Bankman-Fried. The fund had a lofty goal of giving away up to $1 billion in 2022 alone. Among its top priorities was AI safety.

Future Fund was run by a small band of effective altruists that included MacAskill, the philosopher, and Avital Balwit, Aschenbrenner’s future wife. Their work put Aschenbrenner in proximity to the crypto tycoon himself.

Aschenbrenner stayed in the Bahamas with Balwit in the 600-acre luxury resort that FTX executives called home. Aschenbrenner and Bankman-Fried were also in 17 group chats together on the messaging app Signal, according to documents from Bankman-Fried’s trial, including conversations named “EA Discussion” and “Anthropic Chat.”

In 2021, Bankman-Fried agreed to buy a roughly 8% stake in the AI startup for $500 million.

In November 2022, FTX crumbled into bankruptcy. Its $500 million stake in Anthropic was later sold for $1.3 billion to a group that included Jane Street, the quant trading firm that once employed Bankman-Fried and would later invest in Aschenbrenner’s hedge fund.

Aschenbrenner resigned with all of his Future Fund colleagues when FTX imploded. He said in the 2024 podcast that he and Bankman-Fried had already had a falling out by then due to a disagreement over biosecurity grants.

He didn’t have to wait long to find his next opportunity. That same month, OpenAI released ChatGPT, sparking the AI gold rush.

In 2023, Aschenbrenner took a job at OpenAI, and ended up joining a new team led by Ilya Sutskever, the company’s top scientist. It was responsible for “alignment,” ensuring AI models behaved as intended. Initially, OpenAI’s executives weren’t sure what to make of Aschenbrenner, according to people familiar with the matter.

He ruffled feathers when he began warning about potential Chinese spies that might be employed at the company. Aschenbrenner ended up sharing a memo he had prepared criticizing the lack of security protocols inside OpenAI with its board, resulting in a warning from HR. Aschenbrenner said in the 2024 podcast he was told it was racist to worry about espionage from the Chinese Communist Party.

In November 2023, when OpenAI CEO Sam Altman was suddenly fired, hundreds of employees signed a public letter pressuring the board of directors to reinstate him. Aschenbrenner didn’t.

After an interim CEO was appointed, OpenAI executives learned from multiple eyewitnesses that Aschenbrenner had appeared to welcome Altman’s replacement, according to the people familiar with the matter. They grew suspicious about who was loyal to whom, even after Altman was reinstated as CEO. Those who spent the following weeks trying to understand the crisis focused in part on Aschenbrenner’s role, concluding that he had been in contact with at least one director who voted to fire Altman.

OpenAI fired Aschenbrenner after he shared what executives felt was sensitive company information with Holden Karnofsky, a major figure in the EA world who is married to Anthropic’s president, Daniela Amodei. A few months later, Karnofsky joined his wife at Anthropic.

Aschenbrenner has said the information he shared wasn’t confidential.

When he was fired, Aschenbrenner rejected OpenAI’s equity offer worth almost $1 million, he said on the podcast, because he refused to sign a nondisclosure agreement that came with the money.

Aschenbrenner formulated the worldview behind his paper while he was still at OpenAI. He spent time quizzing colleagues for insights into OpenAI’s future strategy. The notes that he wrote up synthesizing his thoughts would form the basis of his viral manifesto.

He launched his fund with a nearly five-hour appearance on a hit podcast hosted by his friend Dwarkesh Patel. They grew close after Patel benefited from a small grant from Future Fund early in his career.

The fund loaded up on public companies across the AI supply chain—semiconductors, energy, software. It also invested in private companies. The crown jewel of the portfolio was a stake in Anthropic.

Many of the companies Situational Awareness backed also did business with Anthropic and OpenAI. They included CoreWeave and Cerebras, as well as smaller data-center companies such as Fluidstack.

Aschenbrenner’s social circle came to include some of Anthropic’s most influential figures. Even some top employees from OpenAI were invited to his social gatherings.

The unraveling

After years of success that inspired one podcaster to nickname him the “Nostradamus of AI,” Situational Awareness was undone by forces outside of his tightknit universe. He didn’t see them coming.

In July, the market was spooked by the release of cheap, powerful AI systems developed in China. With a heavily concentrated portfolio that was long on AI and highly leveraged, the brutal tech selloff left A s c h e n brenner’s fund exposed. Despera te for cash to meet margin calls from lenders, he considered unloading a $3.5 billion stake in Anthropic before deciding to dump most of his public portfolio instead. Citadel agreed to buy it at a discount. When the AI stocks in Aschenbrenner’s book rallied, they helped make July the best month in years for Citadel’s flagship fund.

By the time he signed the paperwork early on a Thursday, Aschenbrenner was already preparing for his wedding that weekend.

As the big day approached, Aschenbrenner’s friends, former colleagues and fellow members of the Silicon Valley intelligentsia filtered into the wedding hotel. The wedding was officiated by Ross Douthat, the former New York Times columnist who recently joined “60 Minutes.” He had reached out to Aschenbrenner after reading his manifesto, and they met last year when his family was on a California road trip, Douthat said.

When the couple asked him to officiate their wedding, he was surprised and sought the counsel of a priest friend, who told Douthat he should do it if he could be spiritually helpful.

“Hopefully I was in some small way,” he said, “even if the circumstances of the wedding ended up being very unusual.”

When it was time to toast the newlyweds, one speaker thanked Citadel CEO Ken Griffin for making it all possible.

Another guest gushed about something more romantic: Aschenbrenner’s promise to buy his wife a galaxy.

Leopold Aschenbrenner, above, attracted billions to his hedge fund before it unraveled last month. Below, a still from a 2024 Dwarkesh Patel podcast episode featuring Aschenbrenner.

JOSH EDELSON

Now-convicted crypto founder Sam Bankman-Fried in 2021.

ANTHONY KWAN FOR WSJ