Meta's AI Agent Muse: 2.8 Million Downloads in 12 Days, Knocking ChatGPT Off the App Store Throne
The marker for the end of the chatbot era isn't a better chatbot — it's a thing that can swipe your card. Meta's Muse proved the point in its first 12 days.
Three key facts
Meta's personal AI agent app Muse, launched September 8, collected roughly 2.8 million downloads in the U.S. and Canada in its first 12 days: 1.8 million on iOS, versus the 1.3 million ChatGPT managed in the same regions and timeframe back then (Apptopia data). The app briefly hit #1 on the U.S. App Store, displacing ChatGPT. The base version is free, with subscriptions for heavy users.
The watershed between Muse and a chatbot is "completing the transaction": sending emails, booking travel, executing multi-step tasks — all the way through to finishing the payment. It doesn't generate text for you to act on; it gets the thing done. Wall Street has already priced it as Meta's next revenue engine: the company's market value has grown by more than $200 billion since the app's release.
Behind the download counts sits the first real test of privacy and security: an agent that reads your email, moves your calendar and touches your payments carries a permissions list longer than any chatbot's. Multiple tech outlets made the same point in the same week — Muse's biggest test isn't growth; it's whenever the first security incident lands.
Wang Dou's Take
First, separate the padding from the substance in this "beats ChatGPT" headline: Muse launched on iOS and Android but only in two countries; ChatGPT launched globally but iOS-only — the comparison doesn't fully line up, and the media flagged that themselves. The substance still holds: a "does-things" app pulling 2.8 million downloads in 12 days across two mature platforms in mature markets tells you users really are switching lanes. People didn't download another chat toy — they hired a digital errand-runner.
The $200 billion market-value bump is the real headline here. What is the market pricing? Not Muse's current revenue — it has almost none — but the narrative that "agents are the new entry point." The search entry point fed Google; the social entry point fed Meta itself; the getting-things-done entry point is the next plot of land. Meta is pouring its most valuable asset — your social graph and behavioral data — straight into Muse's personalization, a dowry OpenAI simply cannot give ChatGPT.
But let's say the ugly part out loud: the business model of a task-executing agent is inherently "spending money on your behalf," which makes every misoperation a real-money loss rather than the mild embarrassment of a chatbot saying something wrong. Wrong trip booked, wrong email sent, wrong card charged — the compensation terms, liability assignments and actuarial math for these incidents are all blank pages right now. 2.8 million users have already handed wallet permissions to a system that is barely three weeks old. The prettier the growth curve, the louder the first incident will ring. Meta has seized the entry point; now it must answer a question a hundred times harder than growth: when it goes wrong, whose fault is it?
Source: Gulf News · Mint (Apptopia data) · KSL/Wall Street view · Meta