Meta layoffs follow-up: Tian Yuandong was backstabbed, and Yao Shunyu and others collectively "poached talent"

The aftershocks of Meta's layoffs are gradually emerging.

Yesterday's news reported that Meta has cut about 600 positions in its artificial intelligence department, affecting FAIR, AI products, and infrastructure teams. Most shockingly, the team led by Tian Yuandong was dismissed with a wave of Alexandr Wang's hand. For more details, see the report: Meta AI massive layoffs, did they cut Tian Yuandong?

Tian Yuandong admitted on Twitter that "acknowledging being laid off is quite embarrassing," but he "might be the least scared person."

That said, it is clear that this round of layoffs has left people somewhat bewildered, and he seems to harbor some resentment.

There are various speculations regarding the reasons behind Tian Yuandong's dismissal, including the "prince competition" at TBD Lab, Alexandr Wang's "burning bridges," and some voices analyzing that the laid-off teams need to be accountable for the underperformance of the Llama 3 inference model and Llama 4.

"Meta has over 600 people who failed to provide an open-source inference model based on Llama 3. On the other hand, DeepSeek demonstrated how to perform inference based on Llama 3 in their R1 paper. It's truly embarrassing. That's why they all need to go." The original post has since been deleted by the author.

Tian Yuandong responded to this analysis with "????," bluntly stating that he was discarded after doing the dirty work.

This incident significantly highlights the chaos within Meta's internal research structure. Mobilization, reorganization, firefighting, and ultimately layoffs have not conveyed a sense of respect for talent in Meta's pursuit of "super intelligence."

According to several industry insiders, the Meta AI department is considered bloated, with research teams like FAIR and more product-oriented teams long engaged in a covert competition for computing resources. When the company grandly established the super intelligence lab, newcomers inherited an already overly inflated behemoth.

This round of layoffs is a move by Meta to continue "slimming down" while also reinforcing Alexandr Wang's core position in the company's AI strategy.

According to CNBC, after the layoffs, the staff size of Meta's super intelligence lab has been reduced to about 3,000 people. Some employees have been notified that their departure date is set for November 21. During this period, they are placed on a "non-working notice," meaning they no longer need to work for Meta, and their internal access has been revoked. The company email stated: "You can use this time to look for other positions within Meta." Meta will pay laid-off employees 16 weeks of severance and an additional two weeks of salary for each year of service (minus the notice period). Tian Yuandong himself revealed to The Paper that he received compensation equivalent to eight months' salary.

On the financial front, this seems to have been anticipated. Meta raised its total expenditure forecast for 2025 to between $114 billion and $118 billion during its earnings call in July, admitting that AI investments will cause spending growth in 2026 to exceed that of 2025. Meta is expected to announce its third-quarter results next week.

It now appears that "layoffs" may just be the prologue to this resource redistribution.

AI Companies' "Talent War"

Researcher Jiaxun Cui complained that Meta's actions are as crazy as "Squid Game" (brutal competition, elimination), with many PhD new hires having their accounts deactivated.

Tian Yuandong retweeted and apologized, stating that he had written a recommendation letter for Jiaxun to help him join Meta's RL team, but ultimately he did not engage in any RL-related research and instead fell into "endless departmental restructuring," being laid off just months after joining.

The ripple effect is significant; Meta's layoffs seem to present a great opportunity for tech companies and startups in Silicon Valley and even worldwide to attract talent.

Today, Tian Yuandong's tweet has garnered over three million views, with countless companies extending olive branches in the comments.

Other researchers "seeking jobs" are also in high demand under their tweets.

Notably, Yao Shunyu has also joined the "online recruitment" ranks. Speculation about his "100 million" salary to join Tencent has been rampant; it is unclear whether his recent post is for personal recruitment intentions or represents his new project seeking talent.

xAI's official account also tweeted: "Look at me, welcome to join~"

Ironically, the number of recruits may exceed the number of layoffs, proving that talent will always find a way out.

Xie Sainin lamented: "There is no banquet that does not end." This round of layoffs led by Meta has started with great fanfare, and those affected are the talents fighting on the front lines of AI.

FAIR was established by Meta (then Facebook) in 2013 and has been around for over a decade. During this time, the revolution in artificial intelligence has taken place. FAIR developed and open-sourced PyTorch in 2016, which can be said to be a stepping stone into the new world of deep learning and artificial intelligence.

However, by 2025, large models are evolving rapidly, and without sufficient innovation and competitiveness, they are destined to be eliminated. The "AI arms race" has become an inevitable reality. For Meta, one of the global leaders in AI, the strategic shift from open-source to commercial closed-source models and the sweeping structural reforms are indeed within expectations.

Of course, we still hope that Meta can allow the rare academic organizations within commercial companies to exist a little longer.

Risk Warning and Disclaimer

The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investing based on this is at your own risk