A developer purchased a $50 Waveshare ESP32-S3 touchscreen pendant computer to experiment with Meta’s open-source Muse AI platform, creating a homemade alternative to the company’s upcoming commercial Muse Charm device. The hardware consists of a 1.75-inch screen housed in a pocket watch-style casing and is officially supported by Meta as a reference implementation for building custom Muse Gadgets.
Initially, the device had significant limitations. Meta’s base software was difficult to use, with slow response times, no audio output, and minimal on-screen text display that required users to reference a companion phone app for full conversations. The single included avatar provided little engagement, and expected smart home integration did not function automatically. The camera features highlighted in Meta’s documentation were hampered by the lack of a physical shutter button.
Using Claude Code, the developer systematically addressed these shortcomings without programming experience. Improvements included adding a conversation history interface, integrating a text-to-speech voice, implementing remote camera controls with live preview, creating multiple custom avatars, and developing a Tamagotchi-style minigame with varying pet care mechanics. Additional refinements fixed audio input issues and added a functional pocket watch display.
The developer noted that Meta’s servers occasionally experienced noticeable lag but acknowledged not trusting the platform with sensitive information, citing a previous incident where Muse leaked a reviewer’s home address. Security concerns led to restricting the device to guest Wi-Fi access.
The project demonstrates that while the Muse Charm may appeal to consumers seeking a dedicated AI companion gadget, the real value proposition for technical users lies in the hardware’s potential for customization rather than Meta’s default software implementation. The accessibility of the codebase allowed extensive personalization despite the developer’s lack of formal coding background.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI