After making two contrarian bets, can Wang Tao really take over Mark Zuckerberg's position?
The AI industry sees wave after wave of emerging trends, with crowds flocking to each new track, but gold mines are never guaranteed in crowded fields. Over the past decade of Tao Wang's career, he made two choices that ran counter to the mainstream: dropping out of school at 19 to do data labeling, and leaving the 9-year-old company he founded at 29 to work at Meta, both of which brought him success in the secular sense.
After that, the public has reached a highly consistent label for Tao Wang: MIT dropout, founder of Scale AI, top AI product leader. These labels piece together a seemingly smooth and brilliant scientific innovation resume. But putting aside the aura given by the outside world, Tao Wang's 10-year trajectory is not a legendary counterattack story.
Or to say, the two critical choices over the decade do not form a complete story on their own. What is really worth recording is what happened after he made those choices, including the events that were out of his control.
Los Alamos and a Disney Ticket
Tao Wang's parents are physicists at Los Alamos National Laboratory. He grew up surrounded by scientists: the topics at family dinners were atomic theory and numerical simulation, and Halloween activities were scientific experiments. His parents never assigned him piles of homework, but hired a programming tutor once they found out he was interested in computers. They never banned him from playing games, but Tao Wang never had time to get addicted to them.
In the sixth grade, there was an opportunity to get a free trip to Disneyland by winning MATHCOUNTS, so Tao Wang got in touch with math competitions for the first time. He later admitted to Forbes that it was that free ticket that sparked his interest in competitive mathematics.
After that, Tao Wang competed all the way to the USA Mathematical Olympiad program, the U.S. Physics Team, and even advanced to the USACO finals. By high school, he had already worked as a full-time engineer in Silicon Valley, first at Addepar, then at Quora, where he met Lucy Guo, the later co-founder of Scale AI. At 17, he was managing a system that handled hundreds of millions of requests every day.
He enrolled in MIT at 18, and took graduate-level computer science courses in his first semester. He dropped out one year later, telling his parents that "I'm going to do one thing this summer", and never returned to campus.
In 2016, Tao Wang and Lucy Guo founded Scale AI in San Francisco, focusing on data labeling. How non-mainstream was this choice at that time? The entire industry was competing to iterate algorithms, and data labeling was regarded as repetitive manual labor with no technical content and no room for imagination.
But Tao Wang followed a different logic. He believed that algorithms define the upper limit of a model, while data determines whether the model can be implemented. Betting directly on the "shovels" of the entire AI industry is more meaningful than competing to iterate algorithms.
Tao Wang mentioned his own experience in the interview
Scale started with autonomous driving, signed contracts with Cruise and Tesla, signed its first contract with Facebook in 2019, and later expanded its client base to OpenAI, Microsoft, and the U.S. Department of Defense. In 2021, 24-year-old Tao Wang had a net worth exceeding 1 billion U.S. dollars, becoming the youngest self-made billionaire at that time.
Behind the success narrative lies the situation of "AI migrant workers". According to reports, Scale's Remotasks platform has about 240,000 registered independent contractors around the world, mostly based in Kenya, the Philippines, Venezuela and other countries. Surveys show that these workers are paid only 1 to 3 U.S. dollars per hour, and generally lack pensions, medical insurance and long-term contracts.
According to records from Amnesty International, a Filipino worker said he worked ten to twelve hours a day, seven days a week on Remotasks, with a daily income of about 8 U.S. dollars. A 2022 study by the University of Oxford pointed out that the platform only meets two of the ten fair work standards.
It is worth mentioning that the relationship between Tao Wang and co-founder Lucy Guo was not all smooth sailing. When managing the operation team, Tao Wang and Lucy Guo had differences in philosophy: Lucy Guo believed that Tao Wang failed to properly solve the payment problem for contract workers, while Tao Wang stuck to his own opinions. Eventually Tao Wang fired Lucy Guo, and the two went their separate ways.
Joining Meta and Building the Viral Muse
After leading Scale AI to become a top player in its niche field, Tao Wang gradually clearly saw the limitations of the business. Data services are supporting infrastructure for the AI industry, always serving algorithms and end products, unable to directly reach C-end users, and difficult to develop AI product forms that are closer to the public's daily life. The ceiling of this track made him start to look for new development directions.
When Tao Wang was planning for transformation and breakthrough, Meta AI was facing a development bottleneck. Mark Zuckerberg called Tao Wang personally, opening the conversation with just one sentence: "Do you have time for a call?" The two first got to know each other through a mutual friend, and a full year passed between the introduction and their first meeting. It was not until the release of Llama 4 that Zuckerberg officially reached out to him. It only took five or six weeks from the first contact to the announcement of the deal.
The final plan was that Meta acquired 49% of Scale AI's shares for 14.3 billion U.S. dollars. Tao Wang stepped down as CEO, served as Chief AI Officer, and took charge of the Superintelligence Lab. Describing leaving the company he had run for nine years, Tao Wang said it was "like giving up your own child".
Regarding Tao Wang's joining, Yann LeCun, Meta's former chief AI scientist, once gave a very direct and sharply worded comment: "Tao Wang has no scientific research experience, does not know how to carry out scientific research activities, and even less understands the preferences of researchers."
Mark Zuckerberg and Tao Wang
That statement was true. Tao Wang knew it himself, so he did not try to act as a technical leader. He gave the position of chief scientist to Shengjia Zhao, and was responsible for recruitment, direction setting and progress control. He mentioned in a podcast that he was not an AI researcher, and his most important task was to recruit outstanding talents and create a good environment for the research team. He managed more than 200 direct reports, but emphasized that "he is not the best manager for more than 200 people".
But on the other side of listening, there was also a reshuffle. In October 2025, Meta laid off about 600 employees, focusing on the FAIR basic research laboratory founded by Yann LeCun, while the TBD Lab directly led by Tao Wang was not affected. Before the layoffs, researchers at FAIR had to get approval from TBD Lab before publishing papers externally, and papers with high value would be "held back". A former Meta executive said: "There were many deep-rooted vested interests at that time, but everyone knew they had to make way."
A person who has worked with both of them said that Tao Wang reminds him of the young Mark Zuckerberg. Zuckerberg did give Tao Wang the autonomy that few people in the company can enjoy, so some people believe that Tao Wang has been regarded as a potential successor to Zuckerberg.
By the beginning of 2026, the entire industry was flocking to coding agents, and Claude was far ahead in the programming field. Tao Wang recalled that many people at that time thought Coding Agent was the only direction in AI worth fighting for, but he said the opposite to Zuckerberg: "Meta is best at consumer-grade products, and its apps cover about 3.5 billion users every day. AI should help people finish the work they have to do, so as to free up time for the things they really want to do."
So Tao Wang led the team to start "playing". Nat Friedman, the product leader, was the first person in Meta's Superintelligence Lab to deeply use the open source tool OpenClaw. He let the agent access the camera at his home to supervise himself to drink water, and he would be praised with a "good job" after drinking.
JollyBot
Tao Wang let the agent read his emails and photos, and did several rounds of in-depth questioning. He described the feeling as "compressing three years of psychotherapy into one moment".
In February 2026, the two each wrote memos. Friedman wrote "The Lobster is the Law" and "Trust is Indispensable", while Tao Wang wrote that "personal superintelligent agents may be the ultimate form of consumer products". In less than two weeks, the Muse prototype was born, and the mascot JollyBot and the initial hardware prototype were also completed at that time.
But polishing Muse took a lot of time and energy. Tao Wang's method was to use a table with more than 100 lines, each line corresponding to a capability that Muse must have, with evaluation criteria and launch thresholds. At a certain stage, 80 items were marked green while 20 were still marked red, and the product would not be released. It was not until one version of Muse Spark 1.3 got all green marks for the first time that the team approved the release.
Vu Tran, a member of the Muse team, later recalled that the product was basically completed as early as May 2026, but Zuckerberg insisted on postponing the launch to prioritize improving the security mechanism. Vu Tran actually held a different opinion: he was used to the idea of "release first, iterate later", and hoped to launch before competitors, but Zuckerberg insisted on waiting.
However, in August, a 14-person startup called Instinct launched a competing agent product. Zuckerberg convened a meeting with Tao Wang and Friedman, and made the final decision to release. On September 8, Muse went online.
After Muse was launched, according to Sensor Tower's estimates, it reached 5 million downloads in the U.S. in about 22 days, faster than the historical data of ChatGPT, Grok and Claude in the same period. It once topped the free chart of the U.S. App Store and ranked first for many consecutive days. Meta's market value soared by about 234 billion U.S. dollars within five trading days.
Muse quickly topped the free chart of the U.S. App Store after launch
From founding Scale at 19 to taking over Meta's AI business at 29, Tao Wang, as a new generation of Chinese scientific innovation practitioner, has provided a pragmatic growth sample.
There is an easily overlooked symmetry in his story. At the age of 19, he chose a path that no one was optimistic about — data labeling. At that time, the whole industry was competing to iterate algorithms, and no one believed that the tough work of feeding data could produce anything valuable. Nine years later, Scale AI became an unavoidable infrastructure for the AI industry.
At the age of 29, he chose another path that no one was optimistic about — instead of making coding agents, he made a personal assistant for ordinary people. When the whole industry was flocking to Coding Agent, he insisted that Meta's advantage lies in its 3.5 billion daily active users. Muse topped the App Store ten days after its launch, which once again seemed to prove his judgment.
When he stepped onto the stage wearing a 30-dollar second-hand camouflage T-shirt and Crocs, making meme posts during meeting breaks, and saying "the internet rewards taking risks", these gestures seemed casual, but in fact they were consistent with his product development philosophy: to achieve the ultimate in the areas he can control, and accept the costs in the areas he cannot control.
At the age of 29, with a net worth of 3.2 billion U.S. dollars and regarded as a potential successor to Zuckerberg, these labels look dazzling when put together, but they may not be enough to define Tao Wang's future.
This article is from the WeChat official account "Zinc Scale" (ID: znkedu), author: Zinc Scale, published with authorization from 36Kr.