Anthropic Founders Used Stuffed Animals for Management

Before co-founding Anthropic, executives at OpenAI utilized toy personas to navigate complex leadership challenges.

Updated on Sept. 30, 2026 in Artificial Intelligence

A collection of plush toys arranged in a meeting circle on a clean wooden office desk. — Before forming Anthropic, former OpenAI executives Daniela Amodei and Holden Karnofsky used plush toy personas to analyze management challenges during their tenure at the startup. AI Illustration. Upload story photo >

Live Poll

Do you trust executives who use unconventional or whimsical methods to make business decisions?

Between 2018 and 2020, OpenAI employees Daniela Amodei and Holden Karnofsky employed a group of stuffed animals as a personal advisory council. They assigned specific personalities to the toys to analyze different management styles and workplace issues.

Why it matters

The practice highlights the unconventional internal dynamics that preceded the departure of several founders from OpenAI in 2020. This period of internal friction directly preceded the formation of Anthropic, now one of the most prominent AI research organizations.

The council included three primary personas: Jinji, a cat character tasked with canceling meetings; Maura, a bear serving as a detail-oriented manager; and Beary Bonds, a panda focused on empathy. These tools allowed for persona-based role-playing to deconstruct management problems.

The players

Daniela Amodei

A co-founder of Anthropic who previously worked at OpenAI during its early growth phase.

Holden Karnofsky

An early contributor to the AI ecosystem who participated in management role-playing exercises.

OpenAI

A prominent artificial intelligence research organization focused on the development of large-scale generative models.

Anthropic

An AI safety and research company formed by former OpenAI employees that builds the Claude series of large language models.

Dario Amodei

The CEO and co-founder of Anthropic who led the recruitment efforts for the company's formation.

The details

The founders maintained separate Slack channels and held private meetings to organize their exit from OpenAI. By assigning distinct character profiles to physical toys, Amodei and Karnofsky attempted to simulate different perspectives for complex organizational decisions. This method served as a structured, albeit atypical, way to test leadership strategies before the team officially departed to establish Anthropic.

2017: Daniela Amodei and Holden Karnofsky were married.
2018: Daniela Amodei joined the research lab OpenAI.
2020: Dario Amodei began recruiting colleagues to form a new company.
December 2020: The group of founders left OpenAI to establish Anthropic.
2023: Daniela Amodei discussed the advisory council in an interview.

The Tech Race

The formation of Anthropic marked a significant split in the AI research landscape, with the new entity focusing heavily on constitutional AI and safety. This history provides insight into the organizational culture that continues to shape the competitive trajectory between Anthropic and OpenAI.

This historical detail provides transparency into the leadership philosophies of executives now managing companies developing widely-used AI tools. Readers interested in organizational management may note how these early role-playing practices correlate with the current operational structure at Anthropic.

The takeaway

Management styles at high-stakes technology startups often rely on unconventional internal frameworks to resolve interpersonal conflicts. Observers should track upcoming leadership announcements and organizational shifts at Anthropic as it scales beyond its founding team.