How Alexandr Wang Pushed Meta Muse to No. 1 in the App Store, Ahead of ChatGPT

Alexandr Wang started Scale AI at 19 and now runs Meta's AI effort, which pushed its Muse agent app to the top of the App Store in September. The climb is real, but it is narrower than the headline story, and the details are worth getting right.

Some people build a company, and some people get hired to rebuild someone else's. Alexandr Wang has done a bit of both. He started Scale AI as a student, turned it into one of the most valuable names in AI data, and then moved across to Meta when the company paid for a large share of his work. What has happened since is less a single turnaround than a bet that the frontier of AI would be won by products people actually use every day, not just by the biggest models.

From MIT dorm project to a $29 billion company

Meta's own leadership page says Wang founded Scale AI in 2016 as a 19-year-old MIT student. The company sold the least glamorous part of AI, the labelling of images, text and sensor data that models need before they can learn anything. That sounds dull until you notice that every major lab depends on it. Under Wang, Scale AI reached a valuation of about $29 billion, according to Forbes.

He stepped down as chief executive when the Meta deal closed, but kept a seat on the Scale AI board. That detail matters, because it shows the arrangement was never a clean exit. Wang was moving from running a company that served the AI industry to running one of the industry's biggest builders.

Meta's bet and the job that came with it

Meta spent more than $14 billion for a 49 percent non-voting stake in Scale AI in 2025. The money bought a share of the company, and it bought Wang. He joined Meta that June as the company's first chief AI officer and leads Meta Superintelligence Labs, which covers research, products and the infrastructure behind them. A memo from June 30 paired him with Nat Friedman, the former GitHub chief executive, who oversees Meta's AI products and applied research.

Meta's Muse AI Agent Cracks the App Store's Top Three on Thin Downloads

Meta's AI agent Muse jumped to No. 2 on the US App Store days after its September 8 launch and just got a Mac app on September 17, but its download numbers trail far behind Threads and Meta's own chatbot app, and Android traction is weak. - Meta's AI agent Muse App Store ranking success - why Meta's Muse AI agent topped app charts

That split is worth noticing. Wang owns the research and the infrastructure, while Friedman owns what people actually touch. An AI lab can have brilliant researchers and still lose the market if nobody builds something a person wants to open twice. Meta seems to have decided that the gap was in the product, and the Muse app is the clearest test of that idea so far.

Muse at the top of the App Store

Muse is a separate agent app, not the Meta AI chatbot. According to Sensor Tower, it reached the top of the US App Store on September 18 with about 730,000 downloads, displacing OpenAI's ChatGPT from first place. Bloomberg reported more than 900,000 downloads in its first six days. Earlier in the year, the Meta AI app had climbed from No. 57 to No. 5 after the Muse Spark model launched, behind ChatGPT, Anthropic's Claude and Google's Gemini at that point.

Here is where the story needs care. CNBC reported on October 7 that Muse was still the top free app on Apple's iPhone store. A Startup Fortune report dated October 6 put it at No. 2 in the US, and the sources do not settle the difference. We also found no current App Store ranking for Claude or ChatGPT, so any claim that Muse has beaten Anthropic in the rankings right now is one we cannot confirm.

What a number one ranking does and does not prove

App Store rankings are a snapshot of installs and attention. A new app can climb quickly on curiosity, a launch event or a fresh feature, and then fall once people stop returning. Downloads are the easy number to watch. Daily use, retention and what people do after the first session are the harder ones, and Meta has not shared those figures publicly.

Wang has made a bolder claim. At Y Combinator's Startup School in October, he said that in some internal cases, when an agent loops and its evaluation and optimization metrics are sound, a group of agents can easily outperform 100 engineers. Meta has not named those cases or published results. It is a claim worth holding until someone can measure it, and it is the kind of statement that will be tested by the people who use the product, not by the people who make it.

What founders can take from it

The useful lesson is not that a famous name can buy a comeback. Wang's career shows a founder who turned a useful, unglamorous service into a large company, then took on a harder job at a much bigger one. The second half is the part most people underestimate. Scale AI worked because it sat under the whole industry, and Muse only works if people want to come back to it.

If you are building an AI product in a crowded market, the question to ask is not whether you can reach the top of a chart for a week. Ask whether a stranger would open your product again on a Tuesday morning, and what you would need to change if the answer is no. That is the question Meta now has to answer for Muse, and it is the same one every founder faces, however large the budget behind them.

Mark Zuckerberg has moved his desk into Meta's AI lab and is reportedly building an AI version of himself to handle the meetings he no longer attends

Meta CEO Mark Zuckerberg has moved his desk into the company's AI lab, coding daily alongside Superintelligence Labs co-leads Alexandr Wang and Nat Friedman. The move follows a sweeping restructuring that cost $15 billion and produced Muse Spark, Meta's first major model release in the new era, which lifted the company's stock 8% last week. Meta... - mark zuckerberg moved desk to meta ai lab - ceo building ai version of himself for meetings