Meta has quietly launched Muse Gadgets, an open-source hardware program that lets hobbyists connect off-the-shelf microcontroller boards to its Muse AI assistant. The pitch is straightforward: grab an ESP32 or Raspberry Pi, download an SDK, wire up screens, speakers, sensors, and buttons, and you have a physical device that talks to Muse. The code ships under Apache 2.0, the community lives on Discord (94 members online at time of writing), and the firmware is on GitHub. The featured hardware roster spans seven devices from third-party manufacturers — Waveshare's ESP32-S3-Touch-AMOLED-1.75, Seeed's reTerminal E1002, M5Stack's StickS3, an AiPi Lite desk buddy, an ideaspark ESP32 with a 1.9-inch display, a Home Assistant Voice Preview Edition, and a Raspberry Pi 5. An HDMI TV stick is listed as "coming soon." Meta explicitly disclaims endorsement of any of these third-party products. The strategic logic is transparent. Meta doesn't need to manufacture hardware to colonize physical space — it needs developers to do it for them. By providing SDKs and letting the maker community build the peripheral ecosystem, Meta gets thousands of edge devices feeding data and engagement back to Muse at essentially zero hardware cost. The Apache 2.0 license is genuinely permissive, but the SDK connects to Meta's proprietary backend, which is the actual value capture layer. This is the Android playbook applied to IoT: open the edges, own the cloud. Every gadget a hobbyist builds is another endpoint for Muse, another surface where Meta's AI becomes the default interface between a user and their home, their TV, their desk. The Home Assistant integration is particularly telling — it positions Muse as the voice layer atop an already-thriving open-source smart home ecosystem. The disclaimers are worth reading carefully. "Side effects of tinkering may include bricked boards, voided warranties, brownouts, or bankruptcies" is playful copy, but it also means Meta accepts zero liability for a hardware ecosystem it is actively encouraging. The devices are sold by other companies; Meta provides only the software glue. This is extraction through abstraction — the hard costs (hardware, support, returns) sit with third parties and users, while the platform value accrues to Meta. For the maker community, the immediate value is real. ESP32 and Raspberry Pi developers get a new integration target, Apache 2.0 means genuine code freedom, and the Discord suggests Meta is at least attempting community engagement rather than a pure dump-and-run open-source release. But the 94-member Discord count signals this is extremely early — more announcement than ecosystem. The 20-year question is whether this seeds a genuinely open hardware ecosystem around AI assistants or whether it follows the familiar pattern: open at the periphery, proprietary at the core, with the community doing unpaid R&D for a platform that can change its API terms at any time.
