According to a report, Daniela Amodei, co-founder of Anthropic, used stuffed animals to navigate workplace challenges, assigning them names and distinct personalities for what she called “managerial role-playing.”
She and her husband, Holden Karnofsky, referred to these toys as their “personal advisory council,” as reported by The Atlantic on Wednesday. This unusual practice reportedly began before Anthropic was founded, during their time at OpenAI, as detailed in Kevin Roose’s upcoming book, “The AGI Chronicles: The Inside Story of the Race to Create an Artificial Superintelligence.”
These antics have come to light just as Anthropic was on the verge of a potential $2 trillion initial public offering, coinciding with heightened scrutiny from the Federal Trade Commission into Anthropic, OpenAI, and other leading research labs, as previously reported.
The article notes that the stuffed animals had their own “character profiles.” For instance, Jinji the cat was portrayed as “lazy” and encouraged Daniela to skip meetings and lighten her workload, while Maura, a pink bear, was characterized as “type-A” and detail-oriented.
Another stuffed animal, Beary Bonds—a panda—was imagined as the “patron bear of generosity and empathy,” which could reflect the couple’s connections to the effective-altruism movement, known for its controversial aspects.
For Amodei, blending humor with serious reflections about management challenges seemed to be the goal. She mentioned to Roose in 2023 that the stuffed animals helped her analyze different personality types in a meaningful way.
Interestingly, very few employees at OpenAI were aware of this unconventional advisory approach, and those who did were uncertain about how much weight Amodei placed on their insights, as reported.
This quirky practice took place during a period when both Amodei and her brother, who had joined OpenAI in 2018, expressed growing frustration with the company’s internal culture and leadership.
Despite her background in English literature, she was perceived by some at OpenAI as more emotionally aware than her brother. While Karnofsky served on OpenAI’s nonprofit board, Daniela’s role expanded to managing technical teams, leading to increasing disagreements with company leaders, including co-founder Greg Brockman and executive Mira Murati.
There was a time when Daniela and Brockman were close, as he had even toasted at her wedding to Karnofsky in 2017, but over time, their relationship became strained.
Meanwhile, Dario, concerned about CEO Sam Altman and Brockman, began to distrust OpenAI, contemplating a departure with like-minded researchers focused on safety. Eventually, the Amodei siblings formed what some referred to as “Dario’s clique,” setting up private Slack channels dedicated to AI safety discussions.
By 2020, Dario was actively seeking to recruit colleagues from OpenAI to start a new venture that emphasized creating advanced AI models with a stronger focus on safety.
Daniela was among the initial seven co-founders, alongside Jack Clark, Sam McCandlish, Tom Brown, Chris Olah, and Jared Kaplan. Ben Mann later joined as the eighth co-founder.
She also took the lead in negotiating their exit from OpenAI, which the founders internally labeled “the divorce.” They officially left in December 2020 to establish Anthropic, originally envisioned as a 50-person AI safety research lab.
Since then, the company has grown significantly, now employing thousands and emerging as one of OpenAI’s main competitors. Currently, Daniela oversees a range of areas at Anthropic, including research, engineering, governance, and commercial operations.
The Post has reached out to both OpenAI and Anthropic for comments.
