Meta’s Superintelligence Labs (MSL) launch came with astounding compensation. To staff MSL, Meta recruited heavily from rival AI labs and startups, with some packages reportedly reaching nine-figure sums tied to hyper-aggressive talent acquisition. According to Axios, Mark Zuckerberg offered compensation exceeding $100 million annually, or even $300 million over four years, to recruit top AI minds, as part of a multi-billion-dollar bid to catch up in the AGI race.
But the strategy is already facing turbulence. In the weeks after MSL was announced, at least eight researchers, engineers and product leaders associated with Meta’s AI operations departed. Three high-profile researchers: Avi Verma, Ethan Knight, and Rishabh Agarwal, have left within months, with some never formally starting. Their departures, along with that of longtime AI leader Chaya Nayak, have raised eyebrows.
Departing research scientist Chi-Hao Wu cited a “dynamic” work environment as his reason for leaving. Internally, Meta has restructured its AI division four times in six months, dissolving its frontier “Behemoth” model and shifting toward product delivery and infrastructure. This has been described as a move away from breakthrough research toward feature execution.
Some former staff moved to other leading AI organizations or startups: multiple people have gone to OpenAI, one to Anthropic, and one to a materials-science startup (Periodic Labs); others are either undisclosed or have announced independent projects.
Meta told reporters that some turnover was “normal” for an organization of its size and that hiring and internal reorganizations were ongoing. Meta has now reportedly frozen its hiring storm after the summer recruitment push.
Who Left?
1. Avi Verma – Researcher
- Tenure at Meta: Less than one month
- What’s next: Returned to OpenAI
Verma was previously a researcher at OpenAI. He joined Meta’s Superintelligence Labs in mid-2025 but left before his official start date.
2. Ethan Knight – Researcher
- Tenure at Meta: Less than one month
- What’s next: Returned to OpenAI
Knight joined Meta from Elon Musk’s xAI, where he had been a researcher, and had also worked at OpenAI earlier in his career. He left Meta’s AI team after only weeks to return to OpenAI.
3. Rishabh Agarwal – Researcher
“It was a tough decision not to continue with the new Superintelligence team, especially given the talent and compute density… but I felt the pull to take on a different kind of risk.” – Rishabh Agarwal
- Tenure at Meta: 5 months
- What’s next: Joining Periodic Labs
Agarwal joined Meta in April 2025 from Google DeepMind. He left after five months and will join Periodic Labs, a startup using AI to discover new physical materials.
4. Afroz Mohiuddin – Senior Staff Engineer
- Tenure at Meta: 1 year
- What’s next: Joined OpenAI technical staff
Mohiuddin worked at Google for more than 14 years before moving to Meta in 2024 as a senior staff engineer. He left in 2025 to join OpenAI.
5. Aram Markosyan – Research Scientist
- Tenure at Meta: 4 years
- What’s next: Undisclosed
Markosyan worked on safety and fairness in large AI models and contributed to Meta’s smart glasses projects. His next role has not been announced.
6. Chi-Hao Wu – Staff Research Scientist
“A lot of people in the AI team maybe feel things are too dynamic… my manager changed several times.” – Chi-Hao Wu
- Tenure at Meta: 5 years
- What’s next: Chief AI Officer at Memories.ai
Wu worked on machine learning and AI safety projects at Meta. He left after more than five years to join Memories.ai as Chief AI Officer.
7. Tony Liu – Engineering Manager
“From Facebook → Instagram → PyTorch, I’ve been lucky to learn from brilliant engineers and teammates.” – Tony Liu
- Tenure at Meta: 8 years
- What’s next: Launching an independent AI systems newsletter
Liu managed teams working on PyTorch GPU systems at Meta. He left after eight years and announced he would be publishing a newsletter on building and scaling AI systems.
8. Chaya Nayak – Director of Product Management
“Today, I’m joining OpenAI… exploring new opportunities at the frontier of AI. ” – Chaya Nayak
- Tenure at Meta: Nearly 9 years
- What’s next: Joining OpenAI to work on special initiatives
Nayak was Director of Product Management for Generative AI at Meta. She also led election transparency work, data-sharing initiatives, and was involved in developing the LLaMA family of models.
9. Bert Maher – Principal Software Engineer
“I’m excited to help make Claude even faster.” – Bert Maher
- Tenure at Meta: 12 years
- What’s next: Joined Anthropic
Maher co-created PyTorch and Triton at Meta, both widely used tools in AI research. After 12 years at the company, he left to join Anthropic.
Rivals were not shy about weighing in at the start of Meta’s hiring spree. On the one hand, Google DeepMind CEO Demis Hassabis called it “rational” from a competitive standpoint: “they’re behind and they need to do something,” but cautioned that many AI professionals place mission and ethical purpose above money.
On the other, OpenAI CEO Sam Altman lambasted Meta’s recruitment approach as culturally hollow. Describing the staggering $100 million offers as “crazy,” he stressed that innovation, rather than compensation, is what drives talent retention. He stated, “Missionaries will beat mercenaries,” and warned that such tactics could sow deep cultural problems down the road.
Now, Meta is regrouping, licensing AI tools from Midjourney, and reorganizing MSL into four teams: TBD Lab for Llamas, FAIR for long-term research, Products & Applied Research for consumer integration, and Infra for infrastructure.
But with top researchers leaving, and internal documents showing tensions over shifting focus from research to delivery, Meta’s effort to catch up in frontier AI is now colliding with the challenge of company culture and talent retention.
