Home Blog The Boy from the Atomic City Building Meta's Superintelligence: Meet Alexandr Wang

The Boy from the Atomic City Building Meta's Superintelligence: Meet Alexandr Wang

TahaSite Editor
Oct 23, 2026
203 Views
The Boy from the Atomic City Building Meta's Superintelligence: Meet Alexandr Wang

Alexandr Wang, founder of Scale AI and the youngest self-made billionaire, has taken charge of leading the superintelligence project and the personal agent, Muse, at Meta.

September 23, 2026: The lights in the Connect conference hall in Menlo Park dimmed, and a 29-year-old stepped onto the stage wearing a thrifted camouflage t-shirt, Crocs, and hair longer than the typical Silicon Valley executive. Alexandr Wang’s appearance looked more like a student coming to the event hall from a campus café than the Chief AI Officer of Meta.

Less than 15 months ago, Wang was the central figure in Meta’s $14.3 billion negotiations with Scale AI, and now he was on stage talking about the future of Muse—an agent that can read users' emails, perform tasks for them on the web, and will soon accompany its owner from inside smart glasses.

With this unconventional style, Wang is now responsible for developing Meta’s personal superintelligence. He is an executive and entrepreneur who became the world’s youngest self-made billionaire at 24. His first idea failed 10 years ago, and his story begins in a place few would guess: the city that built the atomic bomb.

A city in the middle of nowhere; the different path of a genius

Alexandr Wang was born in 1997 in Los Alamos, New Mexico, to a family of Chinese immigrants. Both his parents were physicists, and years later, in a TED talk, Wang described them as brilliant scientists. In another interview, he explained how his family’s scientific environment influenced his perspective on technology.

Los Alamos now enjoys global fame thanks to the Oppenheimer film, but Wang doesn't view his hometown as particularly romantic. At the 2026 Startup School, he laughed and said, "It was truly in the middle of nowhere." He spent his teenage years between math competitions and programming. He knew he wanted to do something big, but he hadn't yet found his path.

Alexander Wang gained experience working at major tech companies before college.

The spark for Wang's career path was ignited by a friend who went to work as an intern at Palantir after high school, offering him a glimpse of Silicon Valley, which led him to work at Quora for a year after high school as well.

In Wang's own words, the period between ages 17 and 19 was a time of constant change for him; he learned something new from the people around him every day, and his goals kept shifting. Years later, he advised teenagers to work at a real company for a while, because from the outside, you cannot understand what building a product and group decision-making really look like.

Dropping out of MIT for a dream whose time had not yet come

After Quora, Wang entered MIT to study machine learning and computer science. Google had just released TensorFlow, and he trained his first models using this tool; models that today, by his own admission, look more like toys.

Toward the end of his freshman year, several tempting paths lay before Wang. In December 2015, he was interviewing for engineering and machine learning opportunities at companies like Stripe, Snapchat Research, and DeepMind, and had received offers from at least some of them.

Eric Wu, CEO of Opendoor, tried to convince him to join his company, and during that conversation, he spoke about the advantages of taking risks while young—an argument that had an unexpected result. Wang began to wonder: if now is truly the time to take risks, why should he work for someone else's company?

Young Wang dropped out of MIT to build an intelligent medical agent.

When Y Combinator, the accelerator that had previously supported early-stage companies like Airbnb and Dropbox, accepted Wang's idea, other options fell by the wayside. However, joining YC meant dropping out of school, and for someone whose parents were physicists and who had two highly educated older brothers, this was no small decision.

Wang and his co-founder, Lucy Guo, wanted to build an agent that would help people access healthcare services and navigate the necessary steps on behalf of the user. They worked on the plan for a month or two until Jared Friedman, a partner at YC, honestly told them he doubted their project would go anywhere because the technology of the time was not yet capable of executing such an idea.

Wang abandoned the personal agent plan, but ten years later, he returned to a similar idea using Meta's resources.

From Scale AI to the Circle of AI Decision-Makers

گجت هوشمند متا (Meta) با کاراکتر سفید کوچک روی صفحه نمایش
گجت Muse Charm متاMeta

The failure of the first idea led Wang and Guo to a more fundamental obstacle. AI models needed data for training, but obtaining suitable data was not as easy as renting computing power or downloading code. In 2016, they launched the startup Scale AI to prepare training data—ranging from street images for self-driving cars to more complex datasets for the next generation of models.

In the early years, it seemed that "data" did not hold much appeal for Silicon Valley, and despite the startup's good revenue, investors doubted the sustainability of their business. Wang later remarked sarcastically that none of those investors had ever trained a model, yet now they were the ones writing articles about the vital importance of data.

Scale AI provided the data needed to train artificial intelligence.

With the explosion of large language models, Scale AI became the shovel seller of the AI gold rush, and companies like OpenAI and Google were added to the startup's client list. The growth of Scale AI made Wang wealthy and influential, and according to the company's 2022 valuation, he was named the world's youngest self-made billionaire at the age of 25.

But the gold rush had a dark side as well. To label massive volumes of data, the startup Scale employed about 240,000 outsourced workers through a subsidiary called Remotasks; these workers raised serious criticisms regarding their wages and working conditions. For this reason, the U.S. Department of Labor launched an investigation into Scale, closing the case in May 2025.

Scale's success brought Wang into the ranks of the wealthiest and most influential figures in technology at the age of 24.

At Scale, Wang gained experience in managing a large company and collaborated with major AI labs; however, he did not build an advanced model on par with OpenAI's products, and this gap later became one of the primary doubts regarding his selection to lead Meta's AI program, as noted by TahaSite.

Over these years, Wang did not neglect the importance of networking. He built extensive relationships in Washington, testified before congressional committees on artificial intelligence and national security, and his circle of friends in the AI industry grew to the point where Sam Altman, his former roommate, once jokingly asked him to stop networking so much.

A $14.3 Billion Deal to Build a Team

In April 2025, Meta released the Llama 4 model; a model whose disappointing performance drew criticism from experts and caused the company to rethink its AI plans. Zuckerberg consulted with several prominent industry figures to restructure Meta's AI.

Wang was in the middle of a meeting at Scale when he saw a WhatsApp message from Zuckerberg: "Do you have time for a call?" Their brief call led to weeks of discussions. Zuckerberg had known Wang for years and admired the clarity of his strategic vision. After speaking with more conventional candidates, he concluded that the Scale CEO could assemble a new team.

Meta ultimately paid $14.3 billion for a 49 percent stake in Scale, valuing the company at approximately $29 billion. It was decided that the startup Scale would remain independent, but Wang and a number of his colleagues moved to Meta. He took on the title of Chief AI Officer, and Meta's Superintelligence Lab (MSL) was introduced on June 30, 2025.

In a farewell message to Scale employees, Wang did not hide the bitterness of this separation: "Opportunities of this magnitude usually come with a price."

Mark Zuckerberg turned to Wang with a historic contract to lead Meta's AI team

His personal price was leaving the company he had built at the age of 19; Meta's price was entrusting its most ambitious project to a 28-year-old entrepreneur who was neither an AI researcher nor had experience in building large language models. Meta's researchers had to accept the leadership of someone who did not even have a university degree.

In a conversation with David Senra, Wang spoke candidly about his limitations; for instance, he stated that he is neither a researcher nor a product designer, and he views his role as finding true and efficient experts, setting the direction, and providing the necessary resources for their work.

At Meta, Wang launched a new research core called TBD Lab and stationed about 200 elite researchers, with nine-figure compensation packages, just a few steps away from Zuckerberg's office. He later said that his most important bet was "talent density"—something that grows like compound interest because talented people attract even more talented individuals.

Wang gathered a team of AI elites in the TBD lab

The lab's structure was also designed to be as flat as possible to minimize the costs of bureaucracy and hierarchy; however, the reorganization of Meta's AI division faced various challenges. For example, the income and bonuses of TBD team members angered other Meta employees and raised internal protests within the company.

In November 2025, Yann LeCun, Meta's chief scientist and one of the fathers of deep learning, announced that he would be leaving the company by the end of the year to launch his own startup. In March 2026, Zuckerberg also formed a separate applied engineering unit led by Maher Saba and reporting to Andrew Bosworth to work on model development alongside Wang's lab; a move some interpret as an insurance policy for a rainy day.

Ending 9 Months of Silence with the Birth of Muse Spark

مردی در حال ارائه مدل هوش مصنوعی Muse Spark 1.3 در کنار کاراکتر Muse
Jason Henry for WSJ

Alexandr Wang explains the difference between working at Scale and Meta with two images, saying that the data business at Scale was like building a skyscraper; an operational process that progressed by improving each stage. Model research at Meta is more like searching for diamonds. Most paths lead to nothing, but one good discovery can compensate for the cost of all the efforts.

At Scale, Wang had to try to have a positive impact on every part of the work; but at Meta, he mostly plays the role of a mentor to experts who need freedom to test their ideas.

On April 8, 2026, after 9 months of quiet work, the first product from Wang's lab arrived under the name Muse Spark, with the internal codename Avocado; the first model in a new family of Meta models that was not a continuation of Llama. Meta introduced Muse Spark as a small, fast model and the first step of a larger plan.

Muse Spark showed that Meta is testing new paths to stay in the race of artificial intelligence pioneers.

Muse Spark secured fourth place in the independent Artificial Analysis index, bringing Meta back into the competition of leading AI models. Unlike Llama, the Spark model was not released with open weights, and users required a Facebook or Instagram account to access it—a point that raised concerns among privacy advocates.

To describe the speed of AI progress, Wang recalled that models from ten years ago could identify cats in YouTube videos, while now "we are talking to a digital god." The exaggeration in this statement reflected his confidence; however, the lab still had to move beyond building a promising model to creating a product that people would be willing to entrust with a part of their lives.

Introducing the Muse Smart Assistant; The Return of a 10-Year-Old Dream

At 19, Wang had failed to build an agent to track healthcare services, but in early 2026, that same vision reappeared before him on a much larger scale.

At that time, most AI developers were focused on coding agents, and Wang and Nat Friedman, head of AI products at Meta, decided to develop an agent for everyone's personal tasks. The initial spark for the Muse agent occurred when they were testing the open-source tool OpenClaw.

Friedman allowed OpenClaw to access his daily life information, and then Wang followed suit. Apparently, by using emails, photos, and other data, the tool had engaged him in a highly personal conversation—an experience he likens to several years of intensive psychotherapy in a single moment.

Wang's personal agent idea, which had been halted ten years ago, returned with the Muse app.

The experiments clarified two points for Wang and Friedman: first, the potential and capability of a personal agent, and second, the level of trust required to access a user's most private information. This led them to write notes in February about the future of personal agents and the issue of trust.

Within a week or two, a small team built a prototype of Meta's agent. Many components of the final product, from the Jolly persona to the early version of the Muse Charm device, were present in that same prototype. Regarding the reason for choosing this path, Wang says:

Meta's products had an advantage that could greatly assist in the development of the company's personal agent. Many people tracked their relationships and interests on Instagram and Facebook, and Wang was aware of the importance of this history, while also believing that Muse should follow a single vision rather than being the sum of the tastes of dozens of teams; just like Apple and Steve Jobs.

On September 8, 2026, the Muse app was released, and the high level of user reception surprised even optimistic experts. According to reports, Muse was downloaded 2.8 million times within 12 days, and its iOS version download rate in the US and Canada reached 1.8 million, while ChatGPT had 1.3 million downloads in the same period. Muse reached the top of the US App Store, and by September 22, Meta's stock had also grown by more than 20 percent, as reported by TahaSite.

Muse could manage emails with user permission, handle online tasks, and execute duties in the background. Wang described it as a "second brain" for every human; a full-time assistant that understands what you want and takes care of it.

Intelligent assistants like Muse perform better the more knowledge and authority they gain. Conversely, the user must trust the provider company or the assistant itself enough to grant the necessary access—the same paradox Wang encountered during the first OpenClaw test.

A Barrage of Tweets; The Distinctive Style of a Gen Z CEO

الکساندر وانگ (Alexander Wang) با تیشرت طرح سگ در محیطی اداری
الکساندر وانگ در ویدیوی پروموشن Muse WonderMeta

The launch of Muse showed another side of Wang to the world. Leaders of major AI labs are often known for long papers, philosophical speeches, or an academic tone.

Sam Altman writes philosophical essays about the future. Dario Amodei publishes multi-thousand-word articles on risks and hopes. Demis Hassabis, the former CEO of DeepMind and Nobel laureate, speaks with the composure of a scientist; but Alexandr Wang has a genre of his own and does not stand quietly in any line.

2025 analyses portrayed Wang as a policy-oriented executive who spoke mostly about AI regulation and geopolitical competition and was not much for discussion with users; but with the release of Muse, everything changed.

After releasing Muse, Wang posted over 300 times on X; more than 100 posts in just a five-day span. He created memes with Jolly, Muse’s cute little character, and poked fun at competitors—for instance, intentionally changing the name "Instinct" to "Insect" and writing, "Muse is absolutely crushing Insect." He emphasizes that he writes all the posts himself and calls this flood of content his "tweet artillery."

Unlike traditional executives, Wang uses the language of internet memes to introduce his products.

On the night of the Connect conference, when Meta released its internal test results showing Muse outperforming its competitors, Wang wrote next to the result: "No surprise." It was a short quip in the same promotional tone used to announce the results, something rarely found in the official announcements of major corporations.

Wang's behavioral style was entirely calculated; as he puts it, the internet rewards bold, unpolished content. He wanted Meta to become attractive to the younger generation again—a generation that views Facebook as their parents' platform—though he admits that it is not easy to accurately measure the precise impact of memes on product growth.

Following the initial success of Muse, The Wall Street Journal dubbed Wang the first Gen Z executive at Meta, and Mike Volpi, a board member at Scale, also made a blunt prediction about his style: "He is going to ruffle some feathers." Even when the conversation turned to exaggerated AI claims on LinkedIn, he advised ignoring LinkedIn, even though it is a good place to find customers.

Beyond the smartphone; an agent the size of a keychain

On September 23, 2026, two weeks after the app's release, Wang appeared alongside Zuckerberg at the annual Connect conference to demonstrate that the project he was leading was not intended to remain merely a chat assistant.

Wang wanted Meta's artificial intelligence to go beyond the phone, enter the workplace and daily life, and instead of just answering users' questions, handle tasks for people across various devices and services. On stage, he unveiled features that showed how this plan would be implemented.

By introducing Muse Charm, Meta transformed its smart assistant into an independent and more companion-like gadget.

Meta intended to assign a dedicated email address to every Muse assistant so that users could send emails to their assistant or include them in correspondence threads to delegate tasks. The company also introduced live video chat capabilities with the assistant's avatars; virtual characters whose voices and appearances users could customize.

The Mac version of Muse could also, with the user's permission, open applications, switch between them, and perform tasks that previously required direct human intervention. Wang explained its utility with a simple example: "You can ask it to help you run your small business or just do the work for you."

Meta announced the integration of Muse with retailers like Walmart and Best Buy, as well as services like Expedia, allowing the assistant to play a role in shopping and travel planning. It was decided that Muse would also make its way to Meta's smart glasses in the coming months, where it could identify what the user sees and engage in conversation about it.

In the vision outlined by Wang and Zuckerberg, instead of visiting multiple separate applications, the user would share their request with the assistant and leave the rest of the work to it.

At the end of the event, a small device called Muse Charm was unveiled; a gadget roughly the size of an AirPods case, featuring a two-inch touchscreen, a fingerprint sensor, and independent 5G connectivity. The virtual Muse character would appear on the device's screen, allowing the user to converse with it directly without taking out their phone.

Wang succeeded in drawing the younger generation's attention back to Meta's products.

The design of the Charm was reminiscent of the 1990s Tamagotchis—small digital toys that their owners always kept with them—but this time, instead of caring for a virtual pet, users carry an intelligent assistant in their pockets.

Alexander Wang and Meta still need to demonstrate how the promises of their AI assistant will be realized in real-world usage; however, it is clear that Muse is not intended to remain just a simple product alongside Facebook, Instagram, and WhatsApp, as TahaSite reports.

On October 12, Meta introduced the Muse Wonder and Wonder Full models, claiming they compete with the best state-of-the-art models in certain tasks. The active memory capability retrieves relevant information from previous conversations so that the agent can track a task over time.

On October 15, the weights for Muse Spark Open were released; a model that apparently aims to compete with Chinese open-source giants. However, the powerful Wonder model remains closed, and now Meta is pursuing two paths simultaneously.

The Adulthood We Dreamed of in Childhood

نمای رابط کاربری اپلیکیشن هوش مصنوعی متا (Meta AI) با قابلیت‌های مدیریت وظایف و یادآوری
Muse پس از بررسی اطلاعات کاربر، مجموعه‌ای از کارهای شخصی‌سازی‌شده را پیشنهاد می‌دهد؛ از آماده‌کردن خبرنامه‌ی صبحگاهی تا تحقیق برای مقاله و هماهنگی مهمانان پادکستMeta

Wang considers his ultimate goal to be achieving "personal superintelligence"; a world in which billions of people each have an assistant tailored to their own lives. Such an assistant understands the user's life context and expands the range of tasks they are capable of performing. Wang believes that if intelligence and execution power become abundant, dreams and ambition will become the scarce resources of the future. He says:

However, the realization of Wang's goal depends on user trust; because an assistant that reads emails, schedules appointments, and performs the user's online tasks requires high-level access, and Meta is known as one of the most controversial technology companies regarding privacy.

Los Alamos, that small town in the middle of nowhere, once built the defining technology of the 20th century, and now a son of that same town has dreams of the superior technology of the 21st century, although this time the product is not meant to remain hidden in a remote laboratory; Wang wants to bring it into the lives of billions of people.

From Other Members of the Pen Family

The Magic of the Electronic Manual Gearbox in the Ferrari 12Cilindri Manuale

طرح گرافیکی موس اسپارک (Muse Spark) با خطوط منحنی و ذرات درخشان
Meta

Comparing the Dongfeng M-Hero 2 and BAW 212: The Best Imported Off-Road SUVs

From Bruce Lee to Abbas Kiarostami | Introducing Films Completed After the Director's Death - Filmzi

الکساندر وانگ (Alexander Wang) در حال گفت‌و‌گو با مارک زاکربرگ (Mark Zuckerberg)
مارک زاکربرگ در کنار الکساندر وانگ و شنگجیا ژائو، از مدیران برنامه‌ی تازه‌ی هوش مصنوعیzuck/Threads@

Off-Road Test and Review of the Electric Range Rover; How Is the Most Luxurious Electric SUV?

PlayStation 5 Security Barrier Breached Just 8 Hours After New Update

الکساندر وانگ (Alexander Wang) در حال ارائه سخنرانی در یک رویداد
الکساندر وانگ در TED Talks، سال ۲۰۲۳TED Talks

Can GTA 6 Break the Historic Record of Expedition 33 at The Game Awards?

5 Decorating Mistakes That Ruin the Beauty of Your Home

الکساندر وانگ (Alexander Wang) و همکارش در فضای یک دفتر کار
الکساندر وانگ و لوسی گوو بنیان‌گذاران Scale، وانگ در زمان تاسیس اسکیل ۱۹ ساله بودSCALE / Forbes

A New Plan for Iran's Internet; Is the Third "Protection Bill" Being Initiated in Parliament?

IRIB Files Complaint Against Aida Panahandeh and the Series "Badnam" in the Ethics Court

الکساندر وانگ (Alex Wang) با کلاه سفید و لباس سرمه‌ای
الکساندر وانگ سال ۲۰۲۴Getty Images/ observer

Is the Rumor of Kim Jong Un's Death Linked to Modern Warfare 4?

15 Years of Absence | Marjan Ashrafizadeh with "Goreh" at the Tehran Short Film Festival - Filmzi

ساختمان نگهبانی و تابلوی ورودی اصلی پروژه لوس آلاموس (Los Alamos Project)
دروازه‌ی اصلی سایت محرمانه‌ی لس‌آلاموس در سال ۱۹۴۳؛ شهری که والدین الکساندر وانگ دهه‌ها بعد در آزمایشگاه ملی آن به‌عنوان فیزیکدان کار کردندWikimedia Commons

5 Magical Destinations for Autumn Camping, from the Coast to the Heart of the Forest

Gas Station Fuel Cards to Be Phased Out in 5 More Provinces

Notorious Game Cracker Hunted; Court Authorizes Disclosure of voices38's Identity

From Dracula to Annabelle; 20 Figures Who Are the Nightmares of Horror Cinema

Tags:
Alexandr Wang Meta AI Scale AI Superintelligence Muse AI
TahaSite

TahaSite

Since childhood, I have always been fascinated by computers and technology. My passion for programming and web development has driven me to constantly learn and create. At TahaSite, my goal is to provide the latest tech insights and build cutting-edge software infrastructure.

Comments

No comments yet. Be the first!