Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
IndustryInsight · On RadarWired→ original

Carnegie Mellon Built an AI Tamagotchi for Objects: The $20 Module Lasts 3–4 Days

Carnegie Mellon PhD student Dahwa Kim built OATs, an AI Tamagotchi for everyday objects. A small E Ink screen with an avatar attaches to a plant, a skateboard, or a coffee machine. Forget to water the plant and the pet looks sad; water it and the pet cheers up. Sensors for moisture, light, and motion read these states. The module costs about $20, but the battery lasts only 3–4 days, so this is still a prototype. For product teams, the takeaway is that simple reminders can be delivered through a pet's face instead of a push notification.

AI-processed from Wired; edited by Hamidun News
Carnegie Mellon Built an AI Tamagotchi for Objects: The $20 Module Lasts 3–4 Days
Source: Wired. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceWired ↗
Editor-in-Chief
Key metric #1The core module costs about $20 and runs 3–4 days (one day with the microphone)
Three use cases"Is someone using me?", a warning near a 3D printer, and self-monitoring
Key metric #3Two Honorable Mention awards at the ACM UIST symposium in Detroit

You forget to water the plant, and instead of yellowing leaves, you see a sad plant-shaped pet on a tiny screen. This is how OATs (Object Agent Twins) works, a project by Dahwa Kim, a fifth-year PhD student in the Future Interfaces Group at Carnegie Mellon University (CMU). In essence, it is an AI Tamagotchi for things.

A tiny E Ink screen (the same "electronic paper" used in e-readers) with a drawn character attaches to a plant, a skateboard, or a coffee machine and shows the object's state through an emotion. The core module costs about $20, but the battery lasts only 3–4 days, or one day with the sound sensor on. So for now this is a demonstration of an idea, not a finished product.

Kim presented the work in early November at the ACM User Interface Software and Technology Symposium (UIST) in Detroit and received two Honorable Mention awards. The idea dates back to the 1990s: Tamagotchi required real-time care, and OATs transfer that sense of caring from virtual creatures to the real objects around us. "We are moving the joy of caring from the virtual world into physical objects," Kim says.

How OATs Work

Inside is a small module with an E Ink screen and a set of internal and external sensors. They track soil moisture, temperature, air humidity, light levels, human presence, and movement. In terms of hardware, it is simple and cheap, which is why the core module costs $20.

AI handles the rest. The images for the different emotional states are drawn by Nano Banana, an image generation service in Google Gemini. Setup happens through a web portal: you name your pet and choose the sensors for a specific object. Based on the object's name and description, the app suggests "positive" and "negative" conditions on its own. For a plant that needs dry air, low humidity can be the happy signal. You can adjust these rules for your particular plant species.

Everyday examples are just as simple. Want to skate more often? Attach an OATs module, and if there's no movement for several days, the avatar gets bored and sad. Step onto the board, and a cheerful face appears on the screen.

How People Actually Use It

Kim gave the devices to many people and saw three use cases. The first is "Is someone using me right now?": the pet on a coffee machine perks up when the grinder starts. The second is a warning for people nearby: 3D printers release toxic fumes, and the avatar signals that the device is running. The third is monitoring your own habits: one person attached OATs to a dumbbell to train regularly and avoid disappointing the AI pet.

At the demonstration, Kim ran a contest for the most interesting idea about which object to pair the device with. The project website has a gallery, ranging from an animated conference badge to a panda on an Owala bottle.

Limitations and the Debate Over Anthropomorphism

The main drawback is battery life: three to four days of operation, or only a day with the sound sensor. The device charges via USB-C. Kim has not yet decided whether to sell the kits online, although she has been advised to.

The demo also sparked a heated debate about anthropomorphizing AI: is it right to give such avatars human traits and emotions so that people feel a connection to them? For the industry, this is not an abstract question. Companies often describe chatbot behavior in human terms, for example, saying they "hallucinate." This summer, Anthropic announced a "dreaming" feature in which an AI agent cleans up local memory in the background. Such terms can distort moral judgments and lead to excessive trust in the tools.

Kim responds that this is why interaction with OATs is deliberately limited to expressing emotions on the E Ink screen. In her view, humanization can be useful, for example, for changing behavior: people give names to cars and bicycles but rarely name their iPhone. There will be more relationships with AI in the future, and they should not be exploited.

What This Means for Developers and Product Managers

Many companies are looking for "the next smartphone," from the failed Humane Ai Pin to OpenAI's upcoming device with Jony Ive and Meta's pocket device Muse Charm, which is expected in December. Increasingly, the bet is on cuteness. OATs show a cheap version of the same idea: the agent doesn't demand attention on a phone screen but lives next to the object and reports its state through a facial expression.

The practical conclusion: for simple notifications like "water me" or "the printer is running," an emotional avatar on E Ink can replace a push message. But power consumption and the ethics of anthropomorphism remain the key questions for anyone who decides to turn this prototype into a product.

Story

Key Questions Answered

How OATs Work?

Inside is a small module with an E Ink screen and a set of internal and external sensors. They track soil moisture, temperature, air humidity, light levels, human presence, and movement. In terms of hardware, it is simple and cheap, which is why the core module costs $20. AI handles the rest. The images for the different emotional states are drawn by Nano Banana, an image generation service in Google Gemini. Setup happens through a web portal: you name your pet and choose the sensors for a specific object.

How People Actually Use It?

Kim gave the devices to many people and saw three use cases. The first is "Is someone using me right now?": the pet on a coffee machine perks up when the grinder starts. The second is a warning for people nearby: 3D printers release toxic fumes, and the avatar signals that the device is running. The third is monitoring your own habits: one person attached OATs to a dumbbell to train regularly and avoid disappointing the AI pet.

What are the key details on Limitations and the Debate Over Anthropomorphism?

The main drawback is battery life: three to four days of operation, or only a day with the sound sensor. The device charges via USB-C. Kim has not yet decided whether to sell the kits online, although she has been advised to. The demo also sparked a heated debate about anthropomorphizing AI: is it right to give such avatars human traits and emotions so that people feel a connection to them? For the industry, this is not an abstract question.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…