Important snack bars from Mark Zuckerberg’s memo to Metas “Superintelligence” AI -Push and large adjustments

• Creation of Meta Superintelligence Labs (MSL):
  Meta consolidates all of his efforts that contain the structure of large AI models-in one thing of the teams that work on his LAMA models, product teams, its fundamental AI research team (fair) and a brand new unit that focuses on the development of the next generation of state-of-the-art AI concentrated-under a new department, MSL. The MSL is headed by ex-scale Cela and co-founder Alexandr Wang, who becomes the “Chief Ai Officer” by Meta and ex-github CEO and AI investor Nat Friedman.
• Strategic goal:
  The express goal is to build “personal superintelligence for everyone”, said Zuckerberg in the memo. Superintelligence would be AI, which can go beyond human level in most cognitive tasks.
• Aggressive talentaquisition:
  Meta Wildert Top -Ki talent of rivals such as Openaai, Googleand anthropic. For this purpose, it offers unprecedented signatory bonuses -up to 100 million US dollars, according to comments from the Openai -CEO Sam Altman. In his memo, Zuckerberg announced the setting of 11 AI researchers from these other AI laboratories. Meta also invested $ 14.3 billion as part of the contract to bring Wang to Meta, and reported reported billions in Friedman’s AI-focused risk capital company to secure his move to Meta.
• Failed acquisition attempts:
  The hiring stroll follows the attempts to acquire important AI startups, including the company Safe Superintelligence and Confusence of Openai, Ilya Sutskevers companies that acquire AI startups.
• Lama fights and competitive pressure:
  The latest Lama Ai Model family from Meta, Lama 4, has the expectations below average. The company has been criticized that it published misleading benchmark numbers for Lama 4, which make the models appear more competitive than they actually are. The publication of the Lama 4 giants, the largest – and, according to Meta, the most powerful – model it produced, has been repeatedly delayed. META has not yet debut any models with “argumentation” functions and lost the ground against rivals such as Openai, Anthropic, Google, Deepseek and Alibabas Qwen. This has led to an internal debate about the AI ​​direction of Meta and an increased urgency to revive its AI portfolio.
• Storage and reputation problems:
  Meta suffered from the loss of important llama researchers to competitors, which further increases the need to attract and keep first-class AI talents. The company is also on an ongoing legal and ethical examination of data practices.
• Solid capital commitment:
  Meta invests tens of billions in infrastructure, data centers and custom hardware to secure a leading role in the AI ​​era.

What follows is the full text as Zuckerberg’s memo, which was preserved, Assets reporter Sharon Goldman:

When the pace of the AI ​​progress accelerates, the development of superintelligence comes into view. I think this will be the beginning of a new era for humanity, and I am fully obliged to do what it takes to lead meta. Today I would like to tell some details about how we organize our AI efforts, to build ourselves for our vision: personal superintelligence for everyone.

We will call our overall organization Meta Superintelligence Labs (MSL). This includes all of our foundations, products and fair teams as well as a new laboratory that focuses on the development of the next generation of our models.

Alexandr Wang has joined Meta to act as our Chief Ai officer and lead MSL. Alex and I worked together for several years and I consider him the most impressive founder of his generation. He has a clear feeling for the historical importance of superintelligence and, as a co -founder and CEO Scaleai, has built into a rapidly growing company that is involved in the development of almost all leading models in the entire industry.

NAT Friedman has also joined Meta to work with Alex to lead the MSL, to lead our work on AI products and applied research. Nat will work with Connor to define his role in the future. He ran at Github MicrosoftAnd finally one of the leading AI investment companies led. Nat worked in our META consulting group last year, so he already has a good feeling for our roadmap and what we have to do.

We also have several strong new team members who join today or who have joined in the past few weeks that I have also excited:

• Trapit bansal -Pioneered RL in the chain of thoughts and fellow artists from O-Series models at Openaai.
• Shuchao BI -Mitschöpfer of GPT-4O language mode and O4-Mini. Previously run multimodal after training at Openaai.
• Huiwen Chang -Mitsmich deputies of the image generation of GPT-4O and previously invented maskgit and muse-text-to-image architectures at Google Research
• Ji Lin -HALF in the construction of O3/O4-Mini, GPT-4O, GPT-4.1, GPT-4.5, 4O images and operator argument.
• Joel Pobar – Inference at Anthropic. Previously 11 years at Meta on HHVM, Hack, Flow, Redex, Performance Tooling and machine learning.
• Jack Rae -Tech manager of the tech before training for Gemini and argumentation for Gemini 2.5. Gopher and Chinchilla brought early LLM efforts Deepmind.
• Hongyu ren -Mitschöpfer from GPT-4O, 4O-Mini, O1-Mini, O3-Mini, O3 and O4-Mini. Previously, a group led after training at Openaai.
• Johan Schalkwyk – Former Google Fellow, formerly involved sesame and technical manager for Maya.
• Pei sun -After training, coding and argumentation for Gemini on Google Deepmind. Previously, the last two generations of Waymo -perception models.
• Jiahui yu -Mitschöpfer from O3, O4-Mini, GPT-4.1 and GPT-4O. Previously, the perception team at Openaai and Co-Led Multimodal at Gemini.
• Shengjia Zhao -Mitschöpfer from Chatgpt, GPT-4, all mini models, 4.1 and O3. Previously led by synthetic data at Openaai.

I am happy about the progress that we have planned for Lama 4.1 and 4.2. These models Power Meta AI, which is used by more than 1 billion monthly active people in our apps and an increasing number of agents in META that help improve our products and technology. We have undertaken to continue building these models.

At the same time, we will start research on our next generation of models to reach the border next year. I have spent over the past few months to meet top people in Meta, other AI laboratories and promising startups in order to put together the founding group for this small talent density. We still form this group and we will ask several people in the KI -org to join this laboratory too.

Meta is uniquely positioned to give the world superintelligence. We have a strong business that supports the establishment of significantly more calculation than smaller laboratories. We have deeper experience in the areas of structure and growth of products that reach billions of people. We are groundbreaking and lead the CI glasses and wearables category, which grows very quickly. And our corporate structure enables us to move with much greater conviction and boldness. I am optimistic that this new influx of talent and parallel approach for model development is caused us to achieve the promise of personal superintelligence for everyone.

We have even more great people on all levels that follow these efforts in the coming weeks, so stay tuned. I look forward to diving and going to work.

***

Liability exclusion: The Fortune Generative AI was used for this story to help with a first draft. An editor checked the accuracy of the information before publication.