Read the full memo I submitted about Meta's giant AI Reorg

Meta has just done the biggest reorganization of artificial intelligence operations to date.

Business Insider has obtained the full email sent by 28-year-old Alexandr Wang, leader of Meta Superintelligence Labs (MSL) to all meta employees working at AI.

As the AI ​​arm race continues, Meta They are trying to build a talented All-Star team to poach researchers from rival AI labs Large offer When we race to build it.”Personal hidden collection. “With the newly formed team, tensions have already emerged between gorgeously compensated new recruits and existing researchers, some of which have threatened to quit.

In a recent email sent to employees, Wang wrote that “Superintelligence is coming,” and that the “take seriously” meta needs to make major changes.

This includes establishing four different teams focused on research, training, products and infrastructure.

The email also made one thing clear. Most of MSL's new divisional heads will lead the MSL product team and report directly to Wang, including investor and former Github CEO Nat Friedman, who was announced as MSL leader alongside Wang in June.

This is the latest in Meta's AI Reorgs series. Less than six months later, Meta has already disbanded two major AI teams.

Wang's email warned that organizational changes could be “disruptive,” but said that Meta needs to increase the speed at which emergency situations are achieved.

In response to a New York Times article on Reorg, MetaCommunication Director Andy Stone was posted on X on Tuesday, criticizing the media for “keeping chatter” about MSL's new structure.

Meta declined to comment further.

Meta's elite TBD lab explores a mysterious new model

Meta is centralizing the research of two units of TBD Lab, a small team focused on training large AI models, and Meta's longtime AI research organization, according to Wang's email.

Related Stories

Business Insider tells the innovative stories you want to know

Business Insider tells the innovative stories you want to know

The MSL research will be led by new chief scientist, Shengjia Zhao, co-creator of ChatGPT. This is the only leader who has not mentioned Wang's notes reporting directly to him.

TBD will also be responsible for researching “new directions” such as building “Omni” models. Wang's email does not provide details about “Omni”. Meta didn't respond when asked from a business insider what “omni” entailed.

The omnipotent model, which can understand everything, not just text, is consistent with the multimodal focus shown by the initial adoption of MSL, which includes multiple experts in audio, video and other media.

Separately, Meta has worked on “Project Omni” to train the chatbots to first messaging users and hyperengage them by remembering chats, Business Insider reported last month.

Fairs play a more active role

The email says that it will be carried out by sending research directly to the TBD Lab when it becomes an “innovation engine” for training MSL.

It's a more active role in fairs than before. The Fair is known for publishing high-level AI research and providing employees with a level of independence that is similar to that of academia.

In the past, there have been relatively few relocations between fair and meta's generative AI divisions, but that seems to be changing now. A fair employee who demanded anonymity for fear of retaliation told Business Insider.

The fair will continue to be run by Rob Fergus, and Jan Lekun will remain its lead scientist.

Bloomberg confirms that Lecun reports to Wang. This is confused amidst the confusion about the fact that MSL technically has two major scientists, Lecun and Zhao.

Nat Friedman reports to the King

Friedman, an investor and former Github CEO, will lead MSL's efforts to integrate AI into Meta's products, and Wang's email does not provide further details.

For years, Meta has been trying to create things like AI glasses and Quest Virtual Reality Headsets. These products have received good reviews, but have yet to make up for the meaningful portion of Meta's revenue.

When Meta internally announced the creation of MSL in June, Friedman said he will partner with Wang to lead MSL and lead the work on AI products and applied research. In an email, the king revealed that Friedman was reporting to him.

“Nat will continue to lead this work to me,” Wang wrote.

Meta's new infrastructure team has announced

Training and running powerful AI models requires a huge amount of specialized infrastructure, including clusters of thousands of Nvidia chips in a vast data center.

MSL's Infra team is led by longtime meta engineering VP Aparna Ramani, who says that LinkedIn Profile is leading all of the Meta AI infrastructure.

Infrastructure has been an early focus in MSL since MSL announced its first adoption.

Another AI organization has been abolished

As part of the reorganization, Meta is disbanding the AGI Foundation team, Wang's Memo said.

This is the second time Meta has closed its major AI units this year. The Agi Foundations are born from the Genai division and lead the Lama AI model of Meta, but themselves disbanded after Meta's latest AI model, the Llama 4, received a lukewarm reception.

Members of the AGI Foundations team are categorized as products, infrastructure and Fair, Wang's email state said. In particular, TBD is not mentioned as a destination for former AGI Foundation members.

The reorganization raises questions about whether yet another Reorg can bring the meta back to the cutting edge of AI.

Meta's constant reconstruction is in contrast to AI rivals such as Openai, Google and Anthropic, with fewer organizational changes.

Meta's reorganization plan was first reported last week by information.

Read the full notes from Alexandr Wang below:

Superintelligence is here. To take it seriously, you need to organize the key areas that are important to reach research, products, infrastructure, and more. We have built world-class organizations around these fields and have brought in some incredible leaders to move the work forward.

As previously announced, Shengjia Zhao will lead our research efforts as MSL's Chief Scientist, Nat Friedman will lead our product efforts, and Rob Fergus will continue to lead the FAIR. Today we are pleased to announce that Aparna Ramani will be moving to MSL to lead the infrastructure needed to support our ambitious research and product betting.

As part of this, they have disbanded the AGI Foundations Organization and moved talent from its team to the appropriate area. Our teams are naturally working and serving, and our products will move to NAT teams. Some researchers move fairly to double our long-term research, but the team working on Infra will move to Aparna's organization. Anyone who has changed teams will still receive updates from their manager or HRBP.

We have made three important changes to our organizational design that will help accelerate our efforts.

1. The core of TBD Labs and Fair, the centralization of basic research efforts.
2. Enhance product efforts with applied research that tackles product-focused models.
3. Establish a unified core infrastructure team to support your research bet.

The work is mapped to four teams.

The TBD Lab will become a small team focused on training and scaling large models to achieve close, pre-training, inference, and post-training, and explore new directions such as the OMNI model.

Fair is the innovation engine for MSL and aims to integrate and expand many research ideas and projects into the implementation of the larger models implemented by Fair by TBD Lab. Rob continues to lead the fair, and Yang continues to serve as the fair's chief scientist, both of which will report to me.

Product and applied research brings our product-centric research efforts closer to product development. This includes AI Tech assistants, voice, media, trust, embodied, and teams who have previously worked on the developer pillars. Nat continues to lead me this job.

The MSL Infra team integrates the elements of the Infra and MSL infrastructure teams into one. The team will focus on accelerating AI research and production by building sophisticated infrastructure, optimized GPU clusters, comprehensive environments, data infrastructure and developer tools to support cutting-edge research, products and AI development across the meta. Aparna is reporting this team to me.

Ahmad and Amir will continue to report to me, focusing on strategic MSL initiatives that will share more later.

I recognize that organizational changes can be disruptive, but I truly believe that taking the time to acquire this structure now will allow us to reach urgently at more speeds over the long term. We use updated rhythm and collaboration models across the team, including when we come together as a complete MSL Org.

I appreciate the flexibility as I adapt to this new construction. Every team at MSL plays a key role and I'm excited to work with you all.