Source time Edition date
ToolsThe Verge AI
I built an AI Tamagotchi before Meta could
我在 Meta 之前做出了一台 AI Tamagotchi
A $50 DIY build versus Meta's Muse Charm shows the real cost and effort gap behind cute AI gadgets, and whether hobbyists or buyers bear that tradeoff is the open question.
Judgment
Sean Hollister’s decision changes: he chooses a $50 Waveshare ESP32-S3 1.75C devkit plus Claude Code fixes over buying Meta’s Muse Charm, accepting no cameras or 5G and extra setup effort to get a lower-cost Muse gadget.
Basis
Hollister writes that he paid $50 for the 1.75-inch Waveshare ESP32-S3 1.75C, a push-to-talk AI devkit that Meta itself suggests for “Muse Gadget” builds using its open-source code and tokens to connect to the Muse AI cloud. He says setup was easy—plug into PC, point Claude at Meta’s GitHub, pair to the Muse phone app—but after installing Meta’s keys, Muse treated audio as transcription, replied slowly, did not speak aloud, showed only a few words onscreen, had a single Jolly avatar, and did not auto-discover smart-home devices. Meta AI engineers Anant Narayanan and Jeremy Payne include a sample app for streaming phone-camera video, but Hollister notes the lack of a remote shutter. He then used Claude Code sessions to add a conversation-history page and reprogram each desired function.
- Buy $50 devkitsetup
- Point Claude at GitHubpair
- Pair to Muse appfix
- Reprogram gaps
Unclear The source does not quantify total build time, battery life, privacy exposure, or whether the $50 kit matches Muse Charm’s durability, support, and long-term software maintenance.
Next Check whether Hollister publishes the full parts list and Claude Code prompts, or whether Meta’s Muse Gadget documentation explicitly lists the Waveshare ESP32-S3 1.75C.