Fast food from Mark Zuckerberg Memorandum on “Supertillage” from Meta.

• Create Meta Superinteigence Labs (MSL):
  Meta unifies all its efforts that involve building large models of artificial intelligence-including its teams that work on Llama models, product teams, and basic artificial intelligence research team, as well as a new unit that focuses on developing the next generation of AI with an advanced edge-in a new section, MSL. MSL is shared by the former CEO widely, the founder Alexander Wang, who has become “the chief artificial intelligence officer” in Mita, former CEO of Jaytoub and artificial intelligence investor Nat Friedman.
• Strategic goal:
  Zuckerberg said in the memo that the explicit goal is to build “Sprz Sprz a personality for everyone.” Superintligence will be an artificial intelligence that can lead to the human level in most cognitive tasks.
• Gain of aggressive talents:
  Meta screams the talent of Amnesty International from my competitors such as Openai, Google and Noteropic. To do this, unprecedented signing bonuses – to $ 100 million, according to the comments made by the CEO of Openai Sam Altman. In his note, Zuckerberg announced the employment of 11 researchers of respected artificial intelligence from other artificial intelligence laboratories. Meta also invested $ 14.3 billion in Scale AI as part of the deal to bring Wang to Meta, and according to what was reported in billions of dollars in the investment capital company that focuses on AI for Frederman to secure his transfer to Meta.
• Failed acquisition:
  The recruitment area follows attempts to refuse to acquire the main startups of artificial intelligence, including former Openai partners, Elijah Superintelligence and Pinxity AI.
• Lama struggles and competitive pressure:
  The latest Llama Ai Model family, called Llama 4, expects to be a weak performance. The company faced criticism that it has published misleading numbers for Llama 4, designed to make the models look more competitive than they are already. The Llama 4 Beheemoth version – the largest – according to Meta, was delayed the strongest model – which it produced, over and over again. Meta has not yet appeared models with “thinking” capabilities, and the earth was lost in front of its competitors such as Openai, Noteropic, Google, Deepseek, and Alibaba’s Qwen. This has led to an internal debate about the direction of Meta AI and an urgent increase in the revitalization of the artificial intelligence portfolio.
• Check and reputation challenges:
  Meta has suffered from the loss of the main researchers in Llama to competitors, which increased the need to attract the talents of artificial intelligence on a global level. The company also faces continuous legal and moral audit on data practices,
• Huge capital commitment:
  Meta invests tens of billions in infrastructure, data centers and devices designated in an attempt to secure a leading role in the era of artificial intelligence.

The following is the full text as Zuckerberg note, which was obtained by luck Correspondent Sharon Goldman:

With the acceleration of the progress of artificial intelligence, the development of excellent cancellation on the horizon. I think this will be the beginning of a new era of humanity, and I am completely committed to doing what it takes to lead Meta. Today I would like to share some details on how to organize our efforts for the artificial intelligence of building towards our vision: SUPERINTLIGE for everyone.

We will call Superinteigence from the comprehensive institution (MSL). This includes all our foundations, products and fair teams, in addition to a new laboratory that focuses on developing the next generation of our models.

Alexandr Wang joined Meta to occupy the position of senior artificial intelligence officers and MSL leadership. I worked with Alex together for several years, and I consider him the founder of his most impressive generation. It has a clear sense of the historical importance of cancellation experts, and as a co -founder and executive director who built Scaleai in a fast -growing company that participates in developing almost all leading models throughout the industry.

Nat Friedman also joined Meta to partnership with Alex to lead MSL, where he presided over our work in AI products and applied research. Nat will work with Connor to determine his role to move forward. He operated GitHub in Microsoft, the last of which was managing one of the leading investment companies in artificial intelligence. NAT served in our Meta Consulting Group for the past year, so he already has a good feeling of our road map and what we need to do.

We also have many strong team members who have joined today or who have joined in the past few weeks that I am excited to participate as well:

• Trapit Bansal RL is a pioneer in the series of thought and the co -founder of the Series Models O in Openai.
• Shuchao Bi Participant historian GPT-4O and O4-MINI. Previously performed by multimedia LED after training in Openai.
• Huyun Zhang -Participated founder to generate GPT-4O photos, and was invented by MASKGIT and Muse Benate to a picture in Google Research
• Ji Lane -Help build O3/O4-MINI, GPT-4O, GPT-4.1, GPT-4.5, 4O-amgegen, and operator thinking.
• Joel Popar Inference in Antarbur. Previously in Meta for 11 years on HHVM, Hack, Flow, Ridex, Performance tools and Automation.
• Jack Ray -The technology before training and logic of Gemini 2.5 is applied. Gove and Early LLM led the efforts in DeepMind.
• Hongyu red Part-4O, 4O-MINI, O1-MINI, O3-MINI, O3 and O4-Mini. Previously, it leads a group for the training dimension in Openai.
• Johan Shawoyek Former Google Fellow, early sesame and technical leadership.
• Pei Sun -After training, coding, and logic of Geographic in Google DeepMind. Previously created the last two generations of models depicting Waymo.
• Jahwiu -Coinatistor O3, O4-MINI, GPT-4.1 and GPT-4O. Previously, he led the conception team in Openai, and participated in the multi -media in Gemini.
• Shangjia Zhao Cheatgpt, GPT-4, GPT-4, all mini models, 4.1 and O3. The hand of artificial data in Openai.

I am excited about the progress we planned for Llama 4.1 and 4.2. These models are Power Meta Ai, which are used by more than a billion monthly activities through our applications and an increasing number of agents through the definition that helps to improve our products and technology. We are committed to continuing to build these models.

In parallel, we will start searching in the next generation of models to reach the border next year or so. I have spent the past few months in interviewing senior people via Meta, other Amnesty International laboratories, and promising startups to assemble the founding group of this heavy effort. We still form this group and we will ask many people through Amnesty International to join this laboratory as well.

Meta is uniquely placed to connect the world to the world. We have a strong work that supports much more construction than smaller laboratories. We have a deeper experience in building products that reach billions of people. We are a pioneer and drive artificial intelligence glasses and wear wearable devices that grow very quickly. Our company’s structure allows us to move with greater condemnation and boldness. I am optimistic that this new flow of talent and the parallel approach to developing models will put us to provide an excellent cancellation promise for all.

We have more great people at all levels who join this effort in the coming weeks, so stay tuned. I am excited to dive and get to work.

***

Relationship: For this story, Fortune AI was used to help with a preliminary draft. Check an editor of the accuracy of the information before publishing.