Meta’s ambitious superintelligence initiative is experiencing significant turbulence as at least eight key employees have departed the company less than two months after CEO Mark Zuckerberg announced the creation of Meta Superintelligence Labs (MSL). The new division was designed to bring “personal superintelligence” to everyone, but the early exodus includes researchers, engineers, and a senior product leader—many of whom were veterans instrumental in building Meta’s core AI infrastructure.
Among the notable departures is Bert Maher, a 12-year Meta veteran who played a crucial role in developing PyTorch, the widely-used open-source framework for training and testing AI systems. Maher has joined Anthropic, where he announced plans to make Claude, the company’s AI chatbot, “even faster.” He also contributed to Triton, a programming language designed to optimize AI model efficiency.
Tony Liu, another long-tenured employee with over eight years at Meta, managed teams working on PyTorch GPU systems critical to training and running large AI models. Liu announced plans to launch a newsletter focused on building and scaling AI systems. Chi-Hao Wu, an AI and machine learning specialist with more than five years at Meta, left to become chief AI officer at Memories.ai, citing concerns about organizational instability due to frequent reorganizations.
The departures extend to Meta’s biggest rival, OpenAI. Chaya Nayak, a nearly nine-year Meta veteran who served as director of product management for generative AI and helped oversee the development of Meta’s Llama language models, joined OpenAI to work on special initiatives. Afroz Mohiuddin, a senior staff engineer who had joined Meta from Google just last year, also moved to OpenAI’s technical staff.
Even recent hires recruited specifically for the superintelligence push are leaving. Wired reported that two researchers, Avi Verma and Ethan Knight, left MSL after less than a month and returned to OpenAI. Rishabh Agarwal, who joined from Google DeepMind in April, departed after just five months to join Periodic Labs, a new AI startup.
The exodus comes amid reports of internal tensions sparked by Meta’s aggressive recruitment strategy, which has included compensation packages worth hundreds of millions of dollars to attract talent from OpenAI and Google DeepMind. These lavishly paid newcomers have reportedly created friction with veteran staff who joined before the superintelligence initiative. Meta has also repeatedly reorganized its AI division this year, most recently dissolving one team and creating four new ones, contributing to what some employees describe as an unstable work environment.
Key Quotes
Some attrition is normal for any organization of this size. Most of these employees had been with the company for years, and we wish them the best.
A Meta spokesperson downplayed the departures, attempting to frame the exodus as routine turnover rather than a sign of deeper organizational problems within the newly formed superintelligence division.
Speaking generally and not for myself, a lot of people in the AI team maybe feel things are too dynamic. There were a lot of organizational changes — actually, my manager changed several times.
Chi-Hao Wu, a former AI and machine learning specialist at Meta who left after more than five years, explained the organizational instability that has affected morale among AI workers, pointing to frequent reorganizations as a source of concern.
During an intense recruiting process, some people will decide to stay in their current job rather than starting a new one. That’s normal.
The Meta spokesperson addressed the departure of recent hires who left MSL after less than a month, including researchers who returned to OpenAI, attempting to normalize what appears to be an unusually high rate of early attrition.
Our Take
Meta’s superintelligence ambitions are colliding with organizational reality. The company’s strategy of throwing massive compensation packages at the problem reveals a fundamental misunderstanding: elite AI researchers are often motivated more by mission, stability, and the ability to do meaningful work than by money alone. The fact that employees are leaving for competitors—and in some cases returning to their previous employers after just weeks—suggests Meta’s culture and organizational chaos are significant deterrents. The loss of PyTorch veterans is especially concerning, as it represents a brain drain of the very infrastructure that underpins Meta’s AI capabilities. This exodus may indicate that Zuckerberg’s aggressive superintelligence push is creating more problems than it solves, potentially undermining Meta’s position in the AI race rather than strengthening it. The pattern suggests that building superintelligence requires more than capital—it demands organizational coherence and a compelling vision that resonates with top talent.
Why This Matters
This story reveals critical challenges in the high-stakes race to develop artificial general intelligence and superintelligence. Meta’s struggle to retain top AI talent despite massive financial investments highlights how organizational culture, stability, and mission clarity matter as much as compensation in attracting and keeping elite researchers. The departures to competitors like OpenAI and Anthropic suggest that Meta may be losing ground in the AI arms race, even as Zuckerberg pours billions into closing the gap.
The exodus of PyTorch developers is particularly significant, as this open-source framework has become foundational to the entire AI industry. Losing institutional knowledge around such critical infrastructure could hamper Meta’s ability to innovate quickly. The internal tensions between lavishly compensated new hires and veteran employees also point to broader challenges facing tech companies as they scramble to build superintelligence capabilities. This pattern of instability and attrition could signal deeper problems with Meta’s AI strategy, potentially affecting its ability to compete with more focused rivals like OpenAI and Anthropic in developing the next generation of AI systems.
Recommended Reading
For those interested in learning more about artificial intelligence, machine learning, and effective AI communication, here are some excellent resources:
Recommended Reading
Related Stories
- Sam Altman Addresses OpenAI Executive Exodus at Italian Tech Week
- OpenAI’s $157B Valuation: Can It Win the Brutal AI Race?
- Tech Workers Are the Real Winners in the AI Talent War, With Pay Set to Soar by 2024
- Google’s Gemini: A Potential Game-Changer in the AI Race
- Amazon Invests Additional $4 Billion in AI Startup Anthropic
Source: https://www.businessinsider.com/meta-superintelligence-team-researchers-exit-ai-push-2025-8