# Carnegie Mellon researchers build $20 ESP32 gadget that gives objects Tamagotchi-like personalities

> **Key takeaway:** Carnegie Mellon researchers created Object-Agent Twins (OATs), sub-$20 ESP32-S3 devices with E Ink displays that use Google's Gemini to generate pixel-art characters expressing sensor-detected conditions as emotions.

**Published:** 2026-10-21T16:09:42+00:00 
**Source:** Hackster.io News 
**Category:** open-hardware 
**Canonical URL:** https://fosswire.org/news/carnegie-mellon-researchers-build-20-esp32-gadget-that-gives.html 

## Summary
Researchers Daehwa Kim and Chris Harrison at Carnegie Mellon University developed Object-Agent Twins (OATs), small open-hardware devices that attach to everyday objects and display AI-generated virtual pet characters whose emotions reflect real-world sensor data. Built around a Waveshare ESP32-S3 board with a 1.54-inch two-color E Ink display, temperature and humidity sensor, microphone, and optional add-on modules for motion, light, occupancy, and soil moisture, each unit costs under $20 to assemble. A 1,100 mAh battery provides roughly one week of operation. In a seven-day study with 12 participants, users successfully configured OATs for items including plants, refrigerators, and dumbbells in about 15 minutes, with some reporting changed habits to keep their object's character happy.

## How OATs work
- **Core hardware**: Waveshare ESP32-S3 dev board, 1.54-inch two-color E Ink display, SHTC3 temp/humidity sensor, microphone, custom expansion board
- **Optional sensors**: Motion, occupancy, light intensity, soil moisture modules (under $5 each)
- **Power**: 1,100 mAh LiPo battery, ~1 week runtime depending on sensor configuration
- **Local processing**: ESP32-S3 wakes every 2 seconds; local anomaly detection filters updates to reduce wireless traffic and save power
- **Cloud integration**: When conditions change significantly, sensor data goes to a cloud server where Google's Gemini selects an emotion from 12 options (excited, happy, frustrated, depressed, etc.)
- **Configuration**: Web interface for object description, sensor selection, and defining desirable/undesirable conditions; Gemini generates pixel-art character and behaviors automatically
- **Enclosure**: 3D-printed case

## Why it matters
The OAT design demonstrates how low-cost open hardware combined with accessible AI services can create novel human-computer interaction paradigms. The sub-$20 BOM and published hardware details make this replicable by makers and hobbyists, though the cloud dependency on Google's Gemini raises questions about long-term sustainability and privacy for a device intended to monitor home environments.