Anthropic co-founder Daniela Amodei and her husband once worked through office conflicts by consulting stuffed animals with names and personalities, a practice she later called useful for reading people.
The New York Post reported that Amodei, now Anthropic’s president and board chair, described the habit to journalist Kevin Roose in 2023. The account comes from material adapted from Roose’s forthcoming book and carried in The Atlantic.
She and her husband, Holden Karnofsky, treated the toys as a “personal advisory council” while both were still tied to OpenAI. Karnofsky had served on the company’s nonprofit board. They gave the plush figures names, personalities, and character profiles and used them for managerial role-playing before the Amodeis left to help launch Anthropic.
Amodei told Roose the stuffed animals were “very useful in analyzing different personality types.” People familiar with the practice were unsure how seriously she took the toys’ guidance. The exercise mixed humor with a real attempt to work through management problems.
One toy, Jinji, was cast as a lazy cat who encouraged Daniela to cancel meetings and avoid taking on too much work. A pink bear named Maura played a detail-oriented, type-A manager. A panda called Beary Bonds was framed as the patron bear of generosity and empathy.
That last profile sat near the piece’s note on ties to the effective-altruism movement, which the reporting labeled controversial. The animals were not a public management system. They were a private tool the couple used on office and people problems inside a company racing to build advanced AI.
Amodei studied English literature before moving into tech. She joined OpenAI in 2018. Her responsibilities grew to managing technical teams. She increasingly clashed with OpenAI co-founder Greg Brockman and then-executive Mira Murati.
Brockman had once been close enough to give a toast at her 2017 wedding to Karnofsky. Years of working together strained that relationship. Her brother Dario Amodei, now Anthropic’s CEO, grew deeply distrustful of Brockman and of OpenAI CEO Sam Altman.
Dario and Daniela formed a group some colleagues dubbed “Dario’s clique.” They used private Slack channels and separate meetings centered on AI safety. By 2020, Dario was recruiting colleagues to leave OpenAI and start a new company focused on advanced models with a stronger safety emphasis.
In December 2020 the siblings left. Daniela led the departure negotiations. Founders privately referred to those talks as “the divorce.” She became one of Anthropic’s original seven co-founders, alongside Jack Clark, Sam McCandlish, Tom Brown, Chris Olah, and Jared Kaplan. Ben Mann was later added as an eighth.
Anthropic was first envisioned as a roughly 50-person AI safety research lab. It ballooned to thousands of employees and became one of OpenAI’s biggest rivals. Daniela oversees research, engineering, product, governance, and commercial operations and serves as board chair.
The same reporting placed the company on a path toward a possible $2 trillion IPO while noting that the Federal Trade Commission had ramped up a sweeping probe of Anthropic, OpenAI, and other frontier labs. The Post sought comment from both companies; no responses were included in the account.
The stuffed-animal sessions happened before Anthropic existed, while Amodei was still inside OpenAI and wrestling with real management friction. She later framed the toys as a practical aid for sorting personality types. The scale of what followed is hard to miss: a safety-focused spinout that grew into a major commercial AI player now under federal scrutiny.
Readers do not need invented drama to see the gap. Leaders who talk about existential risk and careful governance once ran workplace disputes past a lazy cat, a type-A bear, and an empathy panda. Ordinary workers and customers live with the products and policy choices these labs ship. They are entitled to expect adult judgment when the stakes leave the playroom.
Silicon Valley sells the future. It still has to answer for how its top people handled the present when the conflict was sitting across the table.