You forgot to water the plant. Instead of seeing yellowing leaves, you see a cute plant-shaped pet on a tiny screen, looking sad and dejected. Maybe that’ll give you the push you needed to grab the watering can. As you pour water into the soil, the plant pet on the screen gets excited and happy.
That is the concept behind OATs, or Object Agent Twins. These are micro-agent AI avatars on small E Ink displays designed to pair with everyday household objects. Daehwa Kim, a fifth-year PhD student in Carnegie Mellon’s Future Interfaces Group, created the technology. She presented the project at the ACM Symposium on User Interface Software and Technology in Detroit in early November. The forum brings together academics and industry experts to explore human-computer interfaces. Kim received two Honorable Mention awards for her work.
You might remember Tamagotchi, the popular egg-shaped toys from the 1990s that housed digital pets requiring real-time care and attention. Kim’s goal was to bring a similar experience to physical objects in our lives. “We bring this joy of caring from virtual to physical objects,” she tells WIRED. While people have encouraged Kim to sell OATs kits online, she has not decided whether to pursue it as a business opportunity.
Kim states everyone spends all day staring at mobile devices. AI agents, even if they are designed to handle tasks on your behalf, still demand attention from your smartphone screen. OATs were a way to bring those interactions out into the real world.
Giving AI a physical body is not new. Companies rush to create “the next smartphone,” whether that is failed attempts like the Humane Ai Pin, OpenAI’s upcoming hardware created in collaboration with famed ex-Apple designer Jony Ive, or Meta’s Muse Charm. The pocket-sized gadget houses Meta’s popular Muse AI agent and is expected to launch in December. Increasingly, though, companies have leaned on “cuteness” in their strategy for widespread adoption.
OATs have an E Ink display connected to internal and external sensors. These sensors can track things like moisture, temperature, humidity, light, occupancy, and motion. The hardware is relatively cheap, around $20 for the main module. The sprites are created with Google Gemini’s Nano Banana image-generation service, creating a suite of emotions for various states based on the sensor data. You can set all of this up through a web portal, where you can name your OAT and choose which sensors to use for the object you are attaching it to.
The web app will generate “Positive” and “Negative” behaviors based on your object’s name and description. So if the OAT is attached to a plant, positive triggers could be “consistent air temperature readings” and “low humidity levels” if the plant of choice prefers dry environments. However, you can customize these behaviors to suit your specific plant type.
It can be used for very simple things. Want to use your skateboard more often? Attach the OAT to your board, and if it has not detected motion for a few days, the avatar will look sad and lonely. You do not want that, do you? As Kim says, “These kinds of inanimate objects can be more expressive.” When you hop on the board, the OAT avatar will be happy, with fun faces.
The catch is that the system’s battery life is really short, showing why this is more of a proof of concept than anything else. OATs last around three to four days, but paired with the sound level sensor, they will only last a day. It has a USB-C port for charging the lithium-ion polymer battery.
Kim says she distributed OATs to many people and found that people use them in three different ways. One is, “Am I being used?” Like when you attach an OAT to a coffee machine and have the avatar get excited when the machine starts grinding beans. The second is for warning people near an object. For example, 3D printers release toxic fumes during use, and attaching an OAT to one can warn people that the device is operational. And the third is to track their own behaviors. Someone attached an OAT to a dumbbell to track how often they lifted weights, encouraging themselves to exercise regularly so they would not disappoint the AI pet.
The demo at the symposium was popular, and Kim turned it into a competition to see who could come up with the most interesting ideas about what objects to pair with the OATs. Kim’s website features a gallery with sprites people made, from an animated conference badge to a panda avatar for an Owala water bottle.
But the demo also ignited a huge discussion about anthropomorphizing artificial intelligence, and whether it is proper to give these sprites human-like traits and emotions to make people feel connected with them.
It is already a topic in play within the AI industry. Companies often refer to their chatbot behaviors in human terms, like when they “hallucinate.” Anthropic announced a “dreaming” feature this summer where its AI agent cleaned up its local memory as a background process. These anthropomorphized terms can distort moral judgments about AI and cause people to overly trust the tools.
Kim says that is why OATs have very limited interactions restricted to expressive emotions and the E Ink display.
“Anthropomorphism can be very useful in certain ways, like helping people to change their behaviors,” Kim says. “People name their car, and they name their bike. I rarely see people who name their iPhone. I feel like in the future of computing, we will have more of a relationship with AI, but we should definitely be careful not to abuse this.”
What it means
The project does not change how customer products function, account privacy, or API behaviour. It is a research demonstration showing how sensor data can drive visual feedback on a separate display. Users can attach the device to objects to track usage or receive warnings, but the system runs on a short-lived battery and requires manual charging. There is no indication that this technology will be integrated into existing consumer electronics or cloud services.
