For Meta, Charm serves a specific business validation task: after Muse has cultivated a core user base willing to consistently use and deeply customize AI characters, test whether enough of them are willing to pay for a dedicated physical entry point for their AI. In other words, Muse has already answered "Is there anyone willing to use a personal agent?" and Charm now seeks to answer the next question: "Is there anyone willing to carry their personal agent with them?"

Author:Zhang Yaqi

Source: Wall Street Journal

Zuckerberg pulled a keychain-sized handheld device from his pocket and installed Meta’s most popular AI assistant, Muse, onto a 2-inch screen. This product, named Muse Charm, marks Meta’s first step into the independent AI hardware market following its bet on the “personal superintelligence”赛道.

At the Meta Connect annual conference on September 23 local time, Zuckerberg personally unveiled this device. The Muse Charm features a 5G modem, allowing users to wake the Muse AI assistant without unlocking their phone. This design clearly aims to bypass the operating system ecosystems of Apple and Google, creating a dedicated hardware entry point for Meta’s AI. Zuckerberg defined it as “the fastest way to interact with Muse” when not wearing smart glasses.

The timing of Muse Charm's release is significant. Since its launch on September 8, the Muse app has garnered approximately 730,000 downloads within five days, according to Sensor Tower, quickly surpassing 2.5 million downloads and exceeding the daily active user levels of ChatGPT at its launch, reaching the number one spot on the US iOS Free Apps chart. Investors are already pricing in this momentum—stock prices of Muse partners like Shopify have risen, while companies perceived as vulnerable, such as Booking and Intuit, have been sold off.

Led by Apple’s former design chief, developed in just nine months

The creation of Muse Charm is itself an unusual story of hardware development.

According to Meta CTO Andrew Bosworth, the project was not officially launched until March of this year, with an initial requirement as straightforward as: “We want to launch a piece of hardware this year that doesn’t yet exist.” Meta originally planned to release the device in 2027, but after reviewing early prototypes, Zuckerberg asked the team to accelerate the timeline to the end of 2026. Nine months later, the Muse Charm appeared on the stage.

This product was developed by two key design executives recruited by Meta from Apple at the end of last year—former Apple Interface Design Lead Alan Dye and Billy Sorrentino. Previously, they led the design of the iOS interface and Apple Watch interactions; their departure forced Apple’s internal design team into a major reorganization. After joining Meta, they were tasked with establishing the company’s new design lab, and Muse Charm is among the first outcomes of collaboration between this lab and Meta’s Superintelligence AI team.

The internal codename for this device is “Clawmagochi,” a name derived half from the globally popular 1990s electronic pet Tamagotchi and half from the team’s inspiration by the OpenClaw AI assistant. Bosworth admitted in media interviews that the speed at which the project progressed still feels “a little unbelievable” to him—the entire design and manufacturing process took place within a single calendar year.

Currently, Muse Charm is still in a very early stage: the final material for the casing has not been determined, and Meta has only produced a limited number of prototypes, but plans to officially launch before this year’s holiday shopping season in December.

Product form: A handheld AI companion

From a hardware specification standpoint, the design boundaries of Muse Charm are clearly defined—it knows it’s not a phone.

The device features a roughly 2-inch OLED touchscreen with a translucent casing, allowing the internal circuitry to be clearly visible through the back and sides, embodying what Zuckerberg calls an “upgraded hacker aesthetic,” while also exuding a distinct Y2K retro style. Users can carry it by attaching it to a keychain, securing it to a backpack with a lanyard, or wearing it directly on the wrist.

At the center of the screen is a fully customizable virtual Muse character, allowing users to set its name, appearance, clothing, voice, accent, and speech speed. A fingerprint sensor in the upper-right corner of the device instantly wakes up Muse, eliminating the need to unlock your phone or open an app. Built-in 5G modem enables access to Muse without relying on Wi-Fi, and supports music and video playback, with front and rear cameras, microphone, speakers, and a USB-C port.

When two Muse Charm devices come close to each other, they recognize one another, and the two AI virtual characters interact—Meta adds a physical-world social layer to AI agents.

However, Charm is not fully disconnected from the phone. Initial setup and data synchronization still require the Muse app on an iPhone or Android, and the pricing structure for 5G network services has not yet been disclosed.

Analysts remain cautious about Muse Charm's market potential but do not rule out the possibility of it gaining traction among specific user groups.

IDC Director Jitesh Ubrani noted that Charm targets users who prefer not to wear smart glasses and evokes collective memories of the Tamagotchi era. “Some people might remember Tamagotchi, and this product could bring back that feeling,” he said. “It might even run small language models locally on such a compact device.”

Linda Sui, founder of market research firm Smart Analytics Global, believes Muse Charm holds appeal for Gen Z and younger consumers—generations that tend to view AI assistants as companions or thinking partners rather than mere search tools, coinciding with the current resurgence of retro trends in digital cameras, flip phones, and MP3 players. “The combination of AI, fashion, and personalization could indeed attract younger users if priced affordably and designed for easy use.”

But Sui also warned that the path to mass market adoption is unclear: “I’m cautious about its mass market potential. Unless Muse Charm offers unique features that smartphones, watches, headphones, or smart glasses cannot easily replicate, it’s more likely to remain an ‘optional’ product for early adopters rather than a mainstream one.”

The same lesson applies. The Rabbit R1, a handheld device from AI hardware startup Rabbit Inc., garnered significant attention upon its release in early 2024, but ultimately failed due to severely overpromised AI capabilities and an unstable user experience.

Meta's true calculation: hardware experimentation, not a desperate gamble

In an interview after the release, Bosworth was surprisingly candid about Muse Charm’s positioning. He acknowledged that Charm is not a core pillar of Meta’s AI hardware strategy: “Clearly, we expect more people to use Muse through smart glasses and phones.” He characterized the product as an “exploration,” a “classic v0 version.”

For Meta, Charm serves a specific business validation task: after Muse has accumulated a core group of users who are willing to consistently use and deeply customize AI characters, test whether enough of them are willing to pay for a dedicated physical entry point for their AI. In other words, Muse has already answered the question, “Is there anyone willing to use a personal agent?” and Charm now seeks to answer the next question: “Is there anyone willing to carry their personal agent with them?”

According to Bloomberg, citing sources familiar with the matter, Muse Charm will be priced within the same range as smartwatches on the market. The initial batch will primarily target Muse’s core users and early technology enthusiasts, with future versions contingent on market response.

According to Meta’s roadmap, Charm is just the first step in bringing Muse off the phone screen. At the launch event, Zuckerberg stated that Muse will soon extend to Meta’s line of smart glasses, allowing users to assign tasks via voice while driving, cooking, or exercising—without needing to pick up any screen.

The founder, who previously made heavy bets on the metaverse and later shifted focus to AI, this time offers the following reasoning: Toys are often the initial disguise of the next computing platform.