The talent yo-yo: Meta’s AI hiring spree backfires as top recruits quit within weeks

Mark Zuckerberg’s aggressive AI hiring spree is backfiring, with several high-profile recruits threatening to quit or actually leaving Meta within weeks of joining, including ChatGPT co-creator Shengjia Zhao who nearly returned to OpenAI before being named chief AI scientist. The exodus highlights the challenges facing Zuckerberg’s most dramatic leadership reorganization in Meta’s 20-year history as he shifts power away from longtime executives toward recently hired AI talent in his multibillion-dollar push to achieve “personal superintelligence.”

The big picture: Meta is undergoing its fourth AI organizational restructuring in six months, with the newly renamed Meta Superintelligence Lab (MSL) divided into four distinct teams under the leadership of 28-year-old former Scale AI CEO Alexandr Wang.

Key departures: Multiple AI researchers have already left or threatened to leave despite lucrative nine-figure sign-on bonuses and promises of vast computing resources.

• Ethan Knight, a machine-learning scientist, departed after joining just weeks ago.
• Avi Verma, a former OpenAI researcher, completed Meta’s onboarding process but never showed up for his first day.
• Rishabh Agarwal, who started in April, announced his departure Wednesday, saying he “felt the pull to take on a different kind of risk.”
• Veteran employees Chaya Nayak and Loredana Crisan, with 9 and 10 years at Meta respectively, are among more than half a dozen longtime staffers announcing exits.

Leadership tensions: Wang’s management style and Zuckerberg’s hands-on approach are creating friction within the organization.

• Wang and Zuckerberg have struggled to align on timelines for achieving superintelligence, with the CEO pushing for faster progress.
• Some describe Zuckerberg as “micromanaging” the secretive “TBD” (to be determined) department filled with marquee hires.
• Former Scale AI staffers are adjusting to Meta’s bureaucratic structure and lack of revenue goals they experienced at their startup.

Strategic shifts: Meta has quietly abandoned plans to release its flagship Llama Behemoth model to the public after disappointing performance.

• The TBD team is now focused on building newer cutting-edge models instead.
• Meta is “temporarily pausing hiring across all [Meta Superintelligence Labs] teams, with the exception of business critical roles.”
• The company is weighing potential cuts to the AI team going forward.

Power realignment: The reorganization has sidelined several longtime Meta executives in favor of new AI leadership.

• Yann LeCun, Meta’s chief AI scientist since 2013, now reports to Wang despite retaining his title.
• Ahmad Al-Dahle, who previously led Meta’s Llama and generative AI efforts, has not been named head of any teams.
• Chief Product Officer Chris Cox remains in his role but Wang now reports directly to Zuckerberg, cutting Cox out of generative AI oversight.

What they’re saying: Meta dismissed concerns about the leadership turbulence and departures.

• “We appreciate that there’s outsized interest in seemingly every minute detail of our AI efforts, no matter how inconsequential or mundane, but we’re just focused on doing the work to deliver personal superintelligence,” a Meta spokesperson said.
• The company characterized alignment issues between Wang and Zuckerberg as “manufactured tension without basis in fact that’s clearly being pushed by dramatic, navel-gazing busybodies.”
• One investor close to Meta’s new AI leaders observed: “There’s a lot of big men on campus.”

Why this matters: The internal upheaval underscores the intense pressure and high stakes in the AI arms race, where even Meta’s massive resources and generous compensation packages aren’t enough to guarantee talent retention as the company races to catch up with rivals like OpenAI.