Meta旗下Muse用户破千万 25%开通用户保持日活

Summary

Muse is Meta's personal AI agent. In its first month, more than 10 million users completed account setup. It is available only in the US and Canada; the main reason for not launching an international version is high operating cost and insufficient compute to support global scale.

1. About 25% of users who completed setup use it on any given day, and retention among early adopters continues to rise with longer usage.

2. Shopping is a core use case: more than a quarter of active users make purchases through it, with an average order value above $200. Personal productivity is also frequent; users can link email and calendar accounts to handle daily tasks.

3. Most ordinary users use it for free, while professional users pay for advanced models and more tokens; the platform takes a commission on commercial transactions.

4. Privacy protection includes Muse Secure VM, and the forthcoming Muse Confidential VM will prevent even Meta from viewing usage content. However, there have already been reports of Marketplace anomalies and private conversation leaks, and a Canadian bank executive said it is considering blocking financial data access.

Summary

Muse's first-month user scale and shopping data offer brands a new AI-agent touchpoint.

1. More than a quarter of active users complete purchases through shopping on Muse, with an average order value above $200. The scale of Plaid payment-connector activity on Muse has not appeared in other services integrated with Plaid.

2. In the future Muse network, users will be able to shop without redirecting to merchant websites, with the user's Muse communicating with merchant-side AI agents to complete transactions. Merchants will need to adapt to agent interaction; Meta's CFO said all participants will need to adapt to this change.

3. Commercially, most ordinary users use Muse for free, Meta takes a commission on platform transactions, and professional users pay for advanced models and more tokens. Meta expects revenue per Muse user to exceed the hundreds-of-dollars annual level of its social product portfolio.

4. Risks include privacy disputes and Canadian banks considering blocking financial data access. Muse will also become central to Meta's smart glasses strategy, with Reality Labs losses expected to narrow from $19.1 billion in 2025 to $15 billion in 2026.

Summary

Muse creates an incremental AI-agent shopping channel while bringing integration and compliance risks.

1. In its first month, more than 10 million users completed setup; 25% of those who completed setup are daily active. It is available only in the US and Canada, and international expansion is constrained by compute and cost.

2. More than a quarter of active users make purchases through Muse, with an average order value above $200. Plaid payment-connector activity is unprecedented in scale, indicating real demand in the transaction chain.

3. The opportunity lies in the Muse network: a user's Muse can communicate directly with merchant AI agents to complete transactions without redirecting to the merchant website. Sellers can focus on agent-side product communication and transaction fulfillment.

4. Most ordinary users use it for free, the platform takes a commission on transactions, and professional users pay. Risks include privacy disputes, Marketplace anomalies, and banks potentially blocking financial data access; merchant willingness to integrate is still evolving.

Summary

Muse brings design and commercial opportunities for hardware products and e-commerce transaction chains.

1. Muse will become central to Meta's smart glasses strategy. Reality Labs is responsible for smart glasses, VR, and the metaverse, and its loss is expected to narrow from $19.1 billion in 2025 to $15 billion in 2026.

2. In shopping, more than a quarter of active users make purchases, with an average order value above $200. Plaid payment-connector activity is unprecedented in scale, showing that AI agents can support transactions.

3. In the future Muse network, a user's Muse will communicate with merchant AI agents to complete transactions without redirecting to merchant websites, creating agent-collaboration requirements for product digitization and e-commerce fulfillment.

4. Compute costs are high. Meta aims to improve cost efficiency by more than 10x within a year and build compute reserves to serve more than 1 billion users in the future. On privacy, it has Muse Secure VM and will launch Muse Confidential VM.

Summary

Muse shows that personal AI agents have reached scale, and service providers can focus on transaction, productivity, and security solutions.

1. In the first month, more than 10 million users completed setup, with 25% daily active. In shopping, more than a quarter of active users make purchases, with an average order value above $200, and Plaid connector activity is unprecedented in scale.

2. Personal productivity is a high-frequency use case: users link email and calendar accounts to handle daily tasks, indicating that agents need to connect accounts and workflows.

3. Customer pain points center on privacy, security, and cost: there have been reports of Marketplace anomalies and private conversation leaks, and a Canadian bank executive is considering blocking financial data. Meta uses Muse Secure VM to protect data, and the forthcoming Muse Confidential VM will prevent even Meta from viewing content.

4. Compute consumption is high. Meta expects to improve cost efficiency by more than 10x within a year and build compute reserves to serve more than 1 billion users. The Muse network will upgrade agent-to-agent interaction and collaboration.

Summary

Muse's first-month data and Meta's operational priorities provide a reference for platform merchant recruitment, operations, and trend management.

1. In its first month, more than 10 million users completed setup, with 25% daily active. It is available only in the US and Canada, and an international version is constrained by operating cost and compute.

2. Product-market fit has been achieved, and Meta's focus is shifting to scale expansion, revenue growth, and compute-cost optimization. Most ordinary users use it for free, the platform takes a commission on transactions, and professional users pay for advanced models and more tokens.

3. Meta expects to improve cost efficiency by more than 10x in the next year and become profitable after scale. It is building compute reserves to serve more than 1 billion users in the future. The Muse network will upgrade from independent single-user use to interaction between user agents and merchant agents.

4. Risks include privacy disputes, Marketplace anomalies, reports of private conversation leaks, and Canadian banks considering blocking financial data access. Merchant willingness to integrate needs cultivation; the CFO said all participants need to adapt to AI-agent interaction.

Summary

Muse is a new sample for observing the industrialization and governance of personal AI agents.

1. In its first month, more than 10 million users completed setup, with 25% daily active, and early retention rises with usage duration. It is available only in the US and Canada, showing that demand and compute constraints coexist.

2. In shopping, more than a quarter of active users make purchases, with an average order value above $200, and Plaid payment-connector activity is unprecedented in scale. Personal productivity connects email and calendar, indicating that agents are entering transactions and workflows.

3. The business model is free for ordinary users, commission-based for transactions, and paid by professional users for advanced models and more tokens. Meta expects revenue per user to exceed the level of its social portfolio. The compute-cost target is more than 10x efficiency improvement within a year and capacity to serve more than 1 billion users.

4. New issues include privacy and security, Marketplace anomalies, private conversation leaks, and banks considering blocking financial data access. The Muse network will promote agent-to-agent collaboration, and the smart glasses strategy and the narrowing of Reality Labs losses are also worth tracking.