Former Anthropic employee describes burnout and internal tensions

Inside the early days

Ivan Vendrov, a former member of the technical staff at Anthropic, recently shared a detailed account of his time at the AI lab. He worked there from late 2022 to mid-2023, a period that saw the company racing to develop safe artificial general intelligence (AGI). His account offers a glimpse into the internal atmosphere of one of the world's most prominent AI firms.

According to the account shared with Business Insider, the workplace was characterized by a high-pressure pursuit of safety. The company faced the dual challenge of addressing fears of AI catastrophe while simultaneously trying to build tools for commercial use. This tension created a unique and intense environment for the staff involved in the project.

Vendrov noted that his perspective is limited to his experience as a junior employee during a specific window of time. He clarified that he is recalling the gist of dialogues rather than exact words. However, his description of the early days provides a look at the formative culture of the organization as it prepared to go public.

Perks versus purpose

The physical office environment and the perks provided to staff reflected a mix of luxury and austerity. Vendrov described a cramped and ugly office that was stocked with fresh croissants and featured a popular masseuse who visited twice a week. Despite these amenities, he noted that the masseuse was always fully booked, making it difficult for employees to actually schedule a session.

There was also a notable instance of luxury being met with internal resistance. The company once hosted a sushi chef in the office, but a cofounder reportedly expressed concern that it set the wrong tone for the team. The cofounder suggested that the company should not be focused on luxury, stating in a Slack message that they were not there to get rich and eat caviar.

This internal friction highlights the company's stated mission to prioritize safety over profit. The cofounder even suggested future dinners of rice and beans to align with the company's values. This contrast between high-end perks and a humble mission appears to be a defining characteristic of the early Anthropic culture.

The human cost of AI

The most striking part of the account involves the significant level of burnout experienced by the staff. Vendrov described sleepless nights and frequent walks with other colleagues who were also showing signs of extreme stress. He mentioned feeling unable to be honest about his own emotions because he did not want to create doubt in his coworkers.

He felt that his colleagues were smarter, more competent, and more solidly well-intentioned than he was. This sense of responsibility led him to keep his personal doubts to himself while he continued his work. The environment was one where the gravity of the task—building safe AI—seemed to weigh heavily on the individuals performing it.

The pressure of the work was compounded by external factors, such as a visit from the FBI. An agent warned that Anthropic was likely already a target of multiple foreign intelligence agencies. These external pressures, combined with the internal drive for safety, created a high-stakes atmosphere for the technical staff.

Equity and ethics

The company's approach to employee compensation also included a significant focus on charitable giving. During the hiring process, Vendrov was asked if he wanted to pledge two-thirds of his equity to charity. He noted that all of the co-founders had already made this pledge, and he agreed to the terms of the offer.

Anthropic's S-1 filing, which is a document filed with the SEC to register a security for public sale, describes an employee donation-matching program. This program allows workers to pledge a meaningful portion of their equity awards to various causes. The company reported that 50% of its employees are currently included in this program.

These financial structures suggest a corporate identity built on the idea of social responsibility. By encouraging employees to donate their stakes in the company, Anthropic aligns its commercial success with broader social goals. This remains a core part of the company's identity as it moves toward becoming a public entity.

Resignation and reflection

Despite the prestige of working at the lab, Vendrov eventually chose to leave the company after eight months. As he drafted his resignation letter, he considered describing his doubts to the leadership. However, he ultimately decided to praise the organization as an unbelievably good organization in his final correspondence.

His departure serves as a case study on the difficulties of working in high-intensity AI research. The account suggests that the path to building safe AGI is not just a technical challenge but a psychological one for the humans involved. It highlights the risk of burnout in fields where the stakes are perceived to be existential.

While Anthropic has declined to comment on the specific details of this account, the story remains relevant. It provides a narrative of the internal struggles that occur behind the scenes of major AI developments. For the reader, it underscores the human element in the rapid expansion of the artificial intelligence industry.