Alexandr Wang, 29, Meta's chief AI officer and head of Meta Superintelligence Labs, said he persuaded Chief Executive Mark Zuckerberg that the company needed a "personal AI" when the Muse project began. Meta says Muse Wonder, the top-tier AI model it recently released, delivers performance on par with offerings from OpenAI and Anthropic.
In an interview with podcast host David Senra on the 11th, Wang described the development process that took Meta from his arrival to the Muse breakthrough in seven months.
Wang, a Chinese American, founded Scale AI, a company that labels data for training AI models. Born in 1997, he dropped out of the Massachusetts Institute of Technology during his freshman year in the mathematics and computer science department before launching Scale AI. His connection to Meta began when the company was struggling after the failure of its Llama 4 model last year. Five to six weeks after Zuckerberg first reached out, Meta brought Wang on board by acquiring a 49% stake in Scale AI. He took charge of MSL and simultaneously became Meta's chief AI officer.
Wang recalled that at the time, AI coding agents — led by Anthropic's Claude Code — were the industry's most pressing topic. The expectation was that whichever company built the best-optimized AI coding agent would lead the field. Wang instead focused on the concept of "personal superintelligence," an AI that handles the tasks people have to do.
"I told Mark that if we achieve AGI or ASI, Meta will be a truly special company," Wang said. "Our lives are divided into things we want to do and things we have to do. AGI handles the things we have to do, while Meta's existing products also help with the things we want to do." The logic is that Meta can command a powerful position if it both frees people from work through AI and provides the services they turn to with the leisure time that creates.
"Meta's apps are used by 3.5 billion people a day, so Meta's success depends on consumer AI," Wang said. "Fundamentally, Meta is about the individual."
Meta's focus on everyday consumers is paying off. Two weeks after Muse launched last month, it overtook OpenAI's ChatGPT to become the No. 1 free app on the U.S. and Canadian app stores. Consumers have been drawn to its convenience as an agent embedded in daily life — one that drafts and sends emails, books travel and helps with online shopping.
On the 12th, Meta released its flagship model series, Muse Wonder, along with a premium version, Muse Wonderful. The company plans to monetize by offering Wonder first to paying subscribers. On the composite AI performance index compiled by U.S. research firm Artificial Analysis, Muse Wonderful outscored the latest models from OpenAI and Google.
Wang also explained Meta's emphasis on minimizing bureaucracy. "The founding idea of Superintelligence Labs is to recruit the smartest and most capable people and to build an environment where everyone can perform at their best," he said. "My role at Meta is closer to that of a traditional coach." Superintelligence Labs has about 200 staff, and all of them report directly to Wang. The team unveiled Muse seven months after submitting a planning memo on the AI model to Meta's board in February.