Why AI researcher Rishabh Agarwal quit Meta’s Superintelligence Lab after getting $1 million salary package

Rishabh Agarwal

Meta’s Superintelligence Lab, which had aggressively hired researchers from OpenAI, Google and Apple, has witnessed an early exit. Rishabh Agarwal, an AI scientist who reportedly joined on a million-dollar salary just five months ago, has announced his departure.

In a post on X, Agarwal said this would be his last week at Meta. “It was a tough decision not to continue with the new Superintelligence TBD lab, especially given the talent and compute density. But after 7.5 years across Google Brain, DeepMind, and Meta, I felt the pull to take on a different kind of risk,” he wrote.

Agarwal added that the original pitch from Mark Zuckerberg and Scale AI CEO Alexandr Wang to build in Meta’s Superintelligence team was compelling, but he chose to follow Zuckerberg’s own advice: “In a world that’s changing so fast, the biggest risk you can take is not taking any risk.”
Agarwal’s resignation is not the only one. Wired reports that at least three researchers have left the lab in recent weeks. Two of them have already returned to OpenAI. Among them is Avi Verma, who had previously worked at OpenAI before joining Meta, and Ethan Knight, who had earlier moved from OpenAI to Elon Musk’s xAI and then to Meta, before going back.

These exits are significant because they involve elite hires made as part of Meta’s ambitious plan to build “superintelligence,” a project Zuckerberg unveiled just two months ago. Reports suggest that Meta offered multi-million-dollar compensation packages to lure researchers from rivals including Google DeepMind, OpenAI and xAI.

Who is Rishabh Agarwal
Agarwal is an alumnus of IIT Bombay and holds a PhD in computer science from the Mila–Quebec Artificial Intelligence Institute. He started his career with internships at Saavn, Tower Research Capital and Waymo before joining Google Brain in 2018 as a Senior Research Scientist.

He later worked at Google DeepMind, focusing on advancing large language models with reinforcement learning and self-improvement techniques. His move to Meta in April 2025 was considered a big win for the company at the time