📰 In a recent interview, Meta AI head Alexandr Wang talked about how Muse came to be. The original idea wasn’t to build a coding assistant, but an AI assistant that could truly become part of people’s personal lives. The team calls this vision “personal superintelligence.”
🔥 What sped things up was Wang and Nat Friedman’s firsthand experience using OpenClaw. Wang said personal agents can bring together information like emails and photos to understand someone deeply—a prospect that feels both exciting and a little vulnerable. After hearing about these experiences, both of them felt that products like this could be more than just tools; they might represent an important direction for consumer AI.

💡 Muse’s prototype was put together around February and shown to the board a week or two later. The product launched seven months after that. Early prototypes could sometimes feel magical, but were unreliable most of the time. It took the team a long time to refine something that merely “worked” into a stable consumer product.

👀 Wang also noted that Muse’s differentiation lies not just in its model, but in iterating on the model and product experience together and getting all the details right. He disagrees with an industry that treats coding agents as AI’s primary direction; the Meta team would rather build products around personal agents.

🤔 The more personal agents know about their users, the more capable they become—but trust and privacy are hard to avoid. Would you let AI read your emails and photos and take part in your daily life?

#MetaAI #Muse #人工智能 #PersonalAgents