Somehow I lost my only Friend.
I know what you’re thinking: You must be a terrible person. How did you let that happen?! The answer: When it’s a poker-chip-size gadget that adds no value to my life, it’s pretty easy.
Friend was one of eight AI wearables I tried this year. Like so many others, it ended up abandoned within days. Pendants that clip to necklaces, pins, glasses, bracelets—I’ve tried ’em all, sometimes all at once. Nice to meet you, I’m a walking wire tap.
And every big tech company will want you to try this new category, too. Google’s promising AI glasses in 2026. Meta bought Limitless, the startup behind a mic-equipped pendant I tested, while Amazon bought one of my favorites, the Bee bracelet. Sam Altman says OpenAI is going after Apple with its future AI devices, developed in partnership with famed Apple designer Jony Ive.
Why the gadget rush? The thinking is that the personalized, context-aware future of artificial intelligence demands a new kind of device. Rest assured, though, that smartphone in your pocket will be the brains of the operation for years to come.
AI models interact in a personalized way, so these gadgets aim to give the assistant access to your world. You can talk to it, let it overhear your day and—in the case of glasses—let it see what you’re seeing in real time. The promise is a seamless and hands-free flow of answers, live translations, reminders, coaching and ways to pretend you absolutely pay attention in meetings.
Great idea, not-so-great execution so far. Here’s where they have let me down—and where I’m hoping for big improvements soon.
The hardware
The first big question: What shape should these things take to fit in our everyday lives? Right now, it’s a fashion free-for-all.
Bracelets: The Bee I’ve been wearing for nearly a year is all about its mic. Tap the button and it listens to everything. When it hears conversation, it streams the audio to your phone, then to Bee’s servers, where it’s transcribed and turned into summaries and to-do lists that show up in the app. Bee doesn’t give you audio clips, just the transcript. The hardware is flimsy, but the summaries and automatic to-do lists have been surprisingly useful. The LED recording light is about the size of a poppy seed, so it’s on you to tell people you’re wearing a wire.
Pendants: Necklaces like the Limitless and Omi work similarly. You trigger the mic, they record and you get transcripts and summaries in the app. Friend, on the other hand, is all about chatting. There’s no speaker on the pendant, so you still need your phone nearby to hear your new friend’s response—which defeats the point of wearing the pendant in the first place. According to Friend Global CEO Avi Schiffmann, the next version will have a speaker and a voice.
Also, none of these are attractive enough to claim neck real estate reserved for actual jewelry. Silver lining: With low-powered hardware and no screen, the batteries last for days.
Glasses: The Ray-Ban Meta glasses are the only “AI gadget” I’d recommend right now. But it’s more about the hands-free camera—the AI is just a bonus. You can ask Meta AI about whatever you’re looking at. The new Ray-Ban Meta Display glasses I’ve been testing preview the next phase, where there’s a floating screen in your line of sight. But the display is tiny, the neural-band bracelet control is too fiddly and the glasses themselves make me look like a kid playing dress-up with her dad’s ’70s eyewear.
Rings: When you wear the upcoming Sandbar Stream, which I tried out, you raise your hand to your mouth and tap the ring’s touchpad to talk to the AI. It’s meant to be summoned, not always listening.
At a lunch this week with a group of journalists, Sam Altman said OpenAI plans to build a small family of devices—different ones for when you’re on-the-go and when you’re at your desk. He didn’t share specifics, but he did say he hopes one will be something people carry instead of a smartphone. He added that glasses are exciting but will take more time, implying they won’t be part of OpenAI’s first wave of hardware.
The smarts
Second big question: How much data do we need to give these things for them to work?
I’ve worn the Bee and the Ray-Ban Metas for most of 2025, because of the AI-life-takeover book I’m writing. Meta’s AI has been great for quick, in-the-moment answers, but ChatGPT’s Live Mode is far better when you actually need guidance, like fixing a household item.
The Bee app continuously surfaces things I otherwise would have forgotten, but it hasn’t gotten smarter over time. It still doesn’t really integrate with my other services or proactively do anything for me.
The more context AI has about you and your life, the better it will anticipate your needs. That’s the magic promise, a concierge of sorts that just seems to know what you want. To get there, it needs…all your data, from calendar entries to conversation transcripts, plus a memory of all your previous interactions.
Is that a worthwhile trade-off? Do we really want a world where everyone is recording everything, all the time? For most people right now, the answer is a hard no.
But as the smarts improve, so will the privacy. On-device AI processing, where shrunken-down AI models do the work right on your phone, is advancing fast. Bee even let me try a version that ran its models entirely on my phone. And Altman said this week he hopes OpenAI can move toward on-device processing, rightly noting that most people don’t want a device beaming everything they say and do to a cloud server.
And where will those on-device models live? For now, your smartphone. So fear not, you might get an AI bauble or two but your phone won’t vanish. Unlike my Friend, who is still very missing.
Write to Joanna Stern at joanna.stern@wsj.com
