Cheeky Pint — searchable YouTube transcripts & AI Q&A

Muse Wonder: Zuckerberg and Wang's $100B AI

Meta's new Muse Wonder model is a frontier-class personal agent, trained end-to-end for consumer use cases like shopping, coaching, and memory, rather than just coding. The speakers argue that the model's strength comes from scaling both pre-training and RL, with a focus on internal evals and product-specific behaviors like active memory and discretion. They also discuss the massive upfront costs and long planning cycles required, noting that the bet is risky but justified by the potential for broad consumer adoption and new revenue streams.