In September 2026, a new AI assistant did something no other app had managed in years: it knocked ChatGPT off the top of the U.S. App Store free chart. Muse, Meta’s personal AI agent, launched on September 8 and reached the No. 1 spot on Apple’s U.S. App Store within a week. The surge was so dramatic that Meta’s stock jumped 11.4% in a single day, adding nearly $200 billion in market value.
The Numbers: A Launch Unlike Any Other
Muse’s early performance outpaced ChatGPT’s mobile debut in almost every measurable way. According to Apptopia data comparing the U.S. and Canada iOS markets, Muse recorded 1.8 million downloads in its first 12 days, compared to ChatGPT’s 1.3 million during the same period after its launch. Globally, Muse reached 2.8 million installs in 12 days.
The daily engagement gap is even more striking. Muse’s U.S. mobile daily active users hit 642,000, compared with ChatGPT’s 231,000 at the equivalent point after its debut—nearly three times higher. For a product from a company that had struggled to translate its massive AI investments into a consumer hit, these numbers represent a genuine turning point.
What Makes Muse Different? An Agent That Actually Does Things
Beyond Chat: Execution, Not Conversation
Muse is not another chatbot. Its core proposition is execution. Users tell Muse what they want done—book a hotel, find cheaper car insurance, fill out a form, cancel a subscription—and the agent navigates the web, fills in fields, and completes the transaction autonomously. There are no complex prompts to learn. As one analysis put it, “Users do not need to become prompt engineers, they only need to be ordinary people”.
The Cloud VM Advantage
What makes this possible is Muse’s architecture. Meta assigns each user a dedicated cloud virtual machine running a browser and file system. When you give Muse a task, it executes inside that isolated environment—opening tabs, comparing options, and completing purchases—even after you close the app. This “cloud computer” approach is what separates Muse from earlier agent products that required constant supervision.
Security as a Feature, Not an Afterthought
Handing an AI your email and payment access is a significant ask. Meta’s response is a layered security model: sensitive actions require explicit user authorization, and payments route through Stripe-generated one-time card numbers, so real banking details are never exposed to Muse. If Muse makes a mistake, the blast radius is contained within its virtual machine.
The Reality Check: Enthusiasm Meets Friction
Muse’s viral moment has not been without complications. Amazon blocked Muse from its shopping features, citing privacy and security concerns, while Shopify partnered with Meta to enable agentic commerce—a split that reveals how platform-level access will shape the agent economy. Security researchers also discovered a zero-day vulnerability in the macOS version, a reminder that giving AI autonomy carries inherent risks.
The deeper question is sustainability. Previous viral AI releases—including OpenAI’s Sora and Meta’s own Vibes AI—enjoyed similar download spikes before fading. Muse’s integration with commerce and productivity tools creates a higher barrier to exit than a novelty app, but the company must still convert curiosity into long-term retention.
The Bottom Line
Muse represents a meaningful shift in consumer AI: from tools that generate content to agents that complete tasks. Whether it becomes a lasting platform or a well-funded experiment, its launch has already changed the conversation. For the first time, a mainstream AI product is judged not by how well it answers questions, but by how reliably it gets things done.