Meta launched Muse with an aggressively practical pitch: not another chatbot, but an agent meant to shop, haggle, cancel subscriptions and take the small, repetitive tasks off users’ screens. The strategy plays to Meta’s scale and consumer reach rather than a claim that it has won the raw-model race.¹

Within a week to 10 days of launch, that pitch had found an audience. Muse rose to No. 1 among free iPhone apps in the United States, ahead of ChatGPT, Gemini, Claude and Instagram — an early signal that users may value an AI that acts as much as one that answers.¹ Meta says the assistant can research, fill out forms, make purchases and reservations, and connect to services including email, calendars and Spotify, seeking approval before sensitive actions.²

Early anecdotes have supplied the momentum. One user said Muse found equivalent car insurance that cut $3,500 from the annual bill, bought the policy and cancelled the old one in roughly five minutes.² Another said it found a jacket, applied live discount codes and completed the purchase; such stories are central to Meta’s argument that an agent can deliver tangible value, not just clever conversation.²

But the demonstration is also the warning. An agent useful enough to manage purchases, emails, calendars and eventually phone calls must be granted an unusually intimate view of a person’s life. That is a difficult ask for Meta, whose record on privacy and safety has left it with what one account calls a “formidable trust deficit.”¹ The product’s limits are already visible, too: one user reported that Muse could not summarize a saved Instagram workout video because it saw only the first frame.²

Meta is nevertheless broadening the bet. Zuckerberg said, “Muse for Mac is out today,” adding that it works across apps, files, calendar, notes and messages, while users control what it can access.Mark Zuckerberg (finkd)³ The downloads are a promising start; earning permission to become a stand-in for users will be the harder campaign.