Building My Own AI Tamagotchi Pocket Watch With Code

Overview

Consumer hardware trends are increasingly leaning toward dedicated, wearable AI companions designed to act as ambient assistants. Rather than waiting for official commercial releases, developers and enthusiasts are turning to open-source toolkits and low-cost microcontrollers to build bespoke hardware. This hands-on approach highlights how rapid prototyping with modern LLMs and open APIs allows non-traditional developers to customize physical devices far beyond their out-of-the-box limitations.

The Hardware Foundation

At the heart of this DIY build is the Waveshare ESP32-S3 1.75C, a compact touchscreen computer pendant priced at around $50. Officially sanctioned by Meta as a reference platform for building so-called "Muse Gadgets," the device features a small display, push-to-talk functionality, and Wi-Fi connectivity. Unlike high-end smartphones or smartwatches, this minimal devkit lacks built-in cellular radios and advanced native processing, relying instead on cloud-based AI services for heavy lifting.

However, the out-of-the-box software experience proved sparse. Initial configuration required pairing the pendant with Meta's developer repository and companion smartphone apps. Straight from the factory, the hardware suffered from significant limitations: audio transcription errors, slow cloud responses, a lack of local text-to-speech rendering, and an inflexible avatar interface that offered little interactive value beyond basic static visuals.

Customizing the Edge Experience via Vibe Coding

To overcome these hardware and software bottlenecks, the builder utilized Claude Code to iteratively rewrite and patch the device's firmware and companion applications. This conversational programming workflow—often described as vibe coding—enabled the rapid addition of sophisticated features without requiring traditional software engineering expertise.

Key modifications implemented during the build process included:

• Voice and Audio Routing: Integrating external text-to-speech engines like ElevenLabs to deliver natural, audible responses directly from the pendant.
• Interface Expansion: Constructing dedicated conversation history logs that allow users to swipe through previous prompts and replies directly on the tiny display.
• Camera Control: Enabling remote shutter triggers and low-quality live video previews over Bluetooth for remote framing.
• Gamification: Programming a complete Tamagotchi-style virtual pet minigame featuring dynamic hunger, boredom, and cleanliness loops.
• Utility Upgrades: Adding functional watch faces and local media playback capabilities within minutes.

Architectural Implications for Wearable AI

This DIY experiment raises critical questions about the future of dedicated AI hardware form factors. Many early consumer-focused AI pendants struggle to justify their existence alongside the powerful smartphones already residing in users' pockets. If commercial devices are locked down and restricted to rigid vendor ecosystems, enthusiasts may prefer adaptable development kits that can be reprogrammed to suit individual workflows.

Ultimately, the ability to rapidly transform a generic $50 microcontroller into a personalized, interactive gadget demonstrates the power of combining open-source foundational APIs with agentic coding assistants. As physical computing and generative AI continue to converge, the true value of wearable hardware may lie not in closed proprietary polish, but in open programmability.