Omi is another AI companion wearable — but this one’s trying to read your mind

Nik Shevchenko closes his eyes and starts to focus intently. He’s spent the last half hour or so telling me about his new product, an $89 wearable called Omi that can listen to, summarize, and get information out of your conversations. Now he wants to show me the future. So his eyes are closed, and he’s focusing all his attention on the round white puck stuck to his left temple with medical tape. (Did I mention he’s had this thing on his face the whole time? It’s very distracting.)

“Hey, what do you think about The Verge, like as a news media website?” Shevchenko asks, to no one in particular. Then he waits. Fifteen or so seconds later, a notification pops up on his phone, with some AI-generated information about how reputable and terrific a news source The Verge is. Shevchenko is thrilled, and maybe a little relieved. The device read his brain waves to understand he was talking to it, and not to me, and answered his question without any prompting or switching.

So far, that’s all the brain-computer-interface stuff Omi can do. And it seems pretty fragile. “It just understands one channel,” he says, “it’s one electrode.” What he’s trying to build is a device that understands when you’re talking to it and when you’re not. And then eventually understands and saves your thoughts, which Shevchenko both waves off as total science fiction and says will probably be possible in two years. Whenever it happens, he thinks it might change the way you use your AI devices.

For now, the Omi’s actual purpose is much simpler: it’s an always-listening device (the battery apparently lasts three days on a charge) that you wear on a lanyard around your neck that can help you make sense of your day-to-day life. There’s no wake word, but you can still talk to it directly because it’s always on. Think of it as 80 percent companion and 20 percent Alexa assistant.

Omi can summarize a meeting or conversation and give you action items. It can give you information — Shevchenko offhandedly wondered about the price of Bitcoin during our conversation and got a notification from the Omi companion app a few seconds later with the answer. There’s also an Omi app store, which developers are already using to plug the audio input into things like Zapier and Google Drive.

For Shevchenko himself, though, Omi is a personal mentor above all else. “I was born in the middle of nowhere on an island near Japan,” he tells me, and always wanted access to the tech visionaries he grew up admiring. For years, he says he cold-emailed people like Mark Zuckerberg and Elon Musk asking for advice and mentorship on how to make it in tech but never got much response. With no real-life options, Shevchenko decided to build his own.

Omi already has a product called “Personas,” which allows you to plug in anyone’s X handle and create a bot that assumes their social network persona. When Shevchenko shares his screen with me, it shows he’s been chatting with an AI Elon Musk for a long time. “It helps me to understand what I should be working on tomorrow,” Shevchenko says. “Or when I’m talking to someone and I don’t know an answer to the question, it will give me a small nudge — it sometimes tells me I’m wrong!” His wearable heard him say he was sick a few days ago and has been reminding him ever since to get more rest. He asks it every month to give him feedback and tell him how to do better.

He gets a lot of notifications from the Omi app, including during our call, and not all of them make much sense — one was just a transcription of a sentence he’d said a minute earlier. Shevchenko acknowledges it’s early, but he doesn’t seem bothered by the system’s misses. The communication works for him.

Most people won’t use Omi this way, though. The product will ship widely in the second quarter of this year, but Shevchenko says the 5,000 people with an early version of the device are using it to help remember things, look up information, and perform other tasks common to AI assistants.

In that sense, Omi has a lot in common with devices like the Limitless Pendant and bears a striking resemblance to another wearable called Friend. When Friend launched last year, Shevchenko claimed Friend CEO Avi Schiffmann was stealing his work, and the subsequent beef included everything from sniping on X to a freestyle rap diss track. Omi was actually called Friend for a while, and Shevchenko says he changed the name both to avoid confusion and because Schiffmann dropped $1.8 million on Friend.com and subsequently dominated search results.

Shevchenko is confident that Omi can improve on those other devices. All of Omi’s code is open source, and there are already 250 apps in the store. Omi’s plan is to be a big, broad platform, rather than a specific device or app — the device itself is only one piece of the puzzle. The company is using models from OpenAI and Meta to power Omi, so it can iterate more quickly on the product itself.

For all their issues and underlying concerns, it’s clear that AI models are already good enough to feel like a true companion to millions of people. You can feel about that however you’d like, but from Omi and Friend to Character.AI and Replika, bot friends are quickly becoming real friends. What they need, then, is both more information about you and more ways to help you. Omi thinks the first answer is an always-on microphone, and the second is an app store. Then, I guess, comes the brain.