Wang, the 29-year-old co-founder of Scale AI who now heads Meta Superintelligence Labs, revealed details of the memo during an appearance alongside Zuckerberg on the Cheeky Pint podcast hosted by Stripe co-founder John Collison.
The first principle in that note was simple: success in AI is not about having the largest workforce, but about having the highest concentration of top talent, according to a Fortune report. Days before accepting an offer to lead Meta’s artificial intelligence efforts, Meta’s AI chief Alexandr Wang sent Mark Zuckerberg a lengthy memo outlining how he would approach the company’s AI ambitions. The discussion offered a rare glimpse into the conversations that preceded one of the biggest executive hires in the AI industry.
Zuckerberg acknowledged that the company was unhappy with the direction of Llama 4 and was prepared to reorganise its AI efforts if necessary. The Muse app launched in the US on September 8 and quickly became the most popular free app in Apple’s App Store. It topped 2.5 million downloads in about two weeks, according to Sensor Tower data cited by CNN. Zuckerberg said late last year that Meta was discussing spending as much as $100 billion or $200 billion, and recalled asking Wang whether he was sure it would work. Meta expects 2026 capital spending of $130 billion to $145 billion, nearly double last year’s $72.2 billion.
Because failure would be very bad in financial terms, wang said Meta CFO Susan Li pulled him aside at an executive gathering in Hawaii to stress that the model has to work,.
“The very first thing is that you don’t need a lot of people, and talent density is the thing that matters,” Zuckerberg said. The second principle was being willing to rebuild from scratch, according to the Meta CEO. Wang said it is state-of-the-art on his team’s measurements for personal and professional agent tasks. On the Artificial Analysis Intelligence Index, the flagship version lands roughly level with leading models from OpenAI and Anthropic but a couple of points behind the leader, according to AI Weekly.
The lab’s first model, Muse Spark, arrived in April. On Monday, Meta released Muse Wonder, the first model from the lab it calls frontier-class.