Anthropic co-founder Daniela Amodei used stuffed animals to work through office problems, giving them names and personalities for “managerial role-playing,” according to a report.
The sister of Anthropic CEO Dario Amodei and her husband Holden Karnofsky used the toys as a “personal advisory council,” according to the piece published Wednesday in The Atlantic.
The bizarre practice predated Anthropic’s launch and took place while the power couple was still at OpenAI, according to the article, which was adapted from journo Kevin Roose’s forthcoming “The AGI Chronicles: The Inside Story of the Race to Create an Artificial Superintelligence.”
The off-color antics surfaced as Anthropic was headed to a possible $2 trillion IPO and as the Federal Trade Commission ramped up a sweeping probe of Anthropic, OpenAI and other frontier labs, as The Post reported.
Daniela Amodei and Karnofsky’s stuffed animals had “character profiles,” according to the article in The Atlantic.
Jinji was described as “a lazy cat who encouraged Daniela to cancel meetings and avoid taking on too much work.” Pink bear Maura was “a detail-oriented, type-A manager.”
A panda dubbed Beary Bonds served as the “patron bear of generosity and empathy” — a possible nod to the geeks’ ties to the controversial effective-altruism movement.
For Amodei, the exercise appeared to mix humor with a serious attempt to think through management problems.
The stuffed animals were “very useful in analyzing different personality types,” she told Roose in 2023.
Few OpenAI employees knew about the unconventional advisory council — and those who did were unsure how seriously Amodei took its recommendations, according to the report.
The stuffed animal sessions came during a time Amodei, who joined OpenAI in 2018, and her bro were growing frustrated with the company’s leadership and internal culture.
Daniela, who studied English literature before moving into tech, was seen by some OpenAI employees as the more emotionally attuned of the two siblings, according to The Atlantic.
While Karnofsky served on OpenAI’s nonprofit board, Daniel’s responsibilities expanded to managing technical teams. She increasingly clashed with company leaders including co-founder Greg Brockman and then-executive Mira Murati, according to the report.
Daniela and Brockman had once been close enough that he gave a toast at her 2017 wedding to Karnofsky, according to The Atlantic.
But years of working together strained their relationship.
Meanwhile, Dario had grown deeply distrustful of OpenAI CEO Sam Altman, and Brockman and was considering leaving the company with a group of safety-focused researchers, the report said.
The Amodeis eventually formed what some colleagues dubbed “Dario’s clique,” setting up private Slack channels and holding separate meetings centered heavily on AI safety.
By 2020, Dario was recruiting colleagues to leave OpenAI and launch a new company focused on building advanced AI models with a stronger emphasis on safety.
Daniela joined him as one of the original seven co-founders, alongside Jack Clark, Sam McCandlish, Tom Brown, Chris Olah and Jared Kaplan. Ben Mann was later added as an eighth co-founder, according to The Atlantic.
Daniela also led negotiations with OpenAI over the group’s departure, which the founders privately referred to as “the divorce,” the report said.
The siblings left in December 2020 and went on to form Anthropic, which initially envisioned itself as a roughly 50-person AI safety research lab.
It has since ballooned to thousands of employees and become one of OpenAI’s biggest rivals.
Daniela now serves as Anthropic’s president and board chair, overseeing broad areas of the company including research, engineering, product, governance and commercial operations. Karnofsky
The Post has sought comment from OpenAI and Anthropic.
