Link
I’ve been wearing a wire everywhere since February.
I’ve got all the transcripts: important meetings, arguments with my kids, chats with disgruntled employees, late-night bathroom routines. There’s plenty more that I can’t share if I want you (and my bosses and my family) to keep liking me.
No, I’m not an FBI informant. I willingly wore a $50 Bee Pioneer bracelet that records everything I say and uses AI to summarize my life—and send me helpful reminders. I also tested two similar gadgets: the $199 Limitless Pendant and the $159 Plaud NotePin. These assistants can recall every dumb, private and cringeworthy thing that came out of my mouth.
Is this the dawn of the AI surveillance state? Absolutely. Is it also the dream of hyper-personal, all-knowing AI assistants coming to life? Also absolutely.
Within hours of wearing the Bee, I was blown away at how quickly it turned ramblings and random chatter into useful, actionable information. Yet, allow me to quote myself from Feb. 24 at 5:15 p.m.: “This bracelet is really f—ing creepy.”
For years, we’ve been told there’s no way social-media apps are secretly listeningthrough our phones’ mics to target ads. Picking up all that audio? Too hard! Processing it? Too intensive! Storing all the data? Too much!
Yet both the Bee and Limitless do that, with small microphones that listen for voices—particularly your voice. When they detect dialogue, they stream the audio to your phone via Bluetooth, then to company servers where it’s transcribed. AI models take the transcription and generate summaries, which appear in the apps within minutes.
The Bee doesn’t save the audio after transcription; you can’t listen back to anything that’s been said. Limitless keeps the audio, letting you play back full recordings.
Both have physical buttons to quickly cut the mics. The Limitless displays a small light when recording. The Bee only shows a red light when it’s not recording. Nothing says “Don’t worry, it’s off” like the universal symbol for recording! The company says a coming update will reverse this, so you see a light when it’s listening.
With multiday battery life, they capture big stuff—that important pitch meeting—and minutiae, like finding your missing TV remote under the dog.
Plaud is different: It’s a meeting assistant you manually start and stop, then decide if you want to create an AI transcript and summary.
With massive transcripts of your life, the AI in these apps can:
• Summarize: These apps recap your conversations, often reading like a bad biography. Bee summary from April 9: “Joanna’s day was a blend of familiar responsibilities and intense professional engagements…She ended the day listening to music by Sting.” Riveting stuff. Can’t wait for the movie adaptation.
The transcriptions themselves aren’t all that accurate but the summaries usually are. Well, except for March 24: “Conversation with Johnnie Cochran about trial evidence.” Yep, just a casual chat with a deceased celebrity lawyer. (I was watching the new O.J. documentary.) Ethan Sutin, co-founder and chief technology officer of Bluush, Bee’s creator, says a recent update has improved the ability to tell recorded voices from real, living ones.
• Remind: Turns out, I promise to do a lot of things without putting them on a to-do list. Bee listens for action items and adds them to a suggested list. It’s repeatedly reminded me of important tasks, like calling the plumber or following up on work stuff. But it also hilariously adds things I’d never put on a list, like “check in on your sick son”or “schedule a follow-up with your hair stylist to discuss your haircut.”
Analyze: Both Bee and Limitless have chatbots so you can ask about your recorded life. I asked Bee for a detailed breakdown of my cursing habits. (Daily average: 2.4 curses.) But it can be genuinely helpful. “Look through my chats with Ethan from Bee and tell me what AI models it uses.” The answer: a combo from Anthropic, Google and Meta.
Nobody I talked to over the past few months would have known I was recording them, if I hadn’t told them. It’s a little fun, like I’m a low-budget Ethan Hunt. Mostly, though, I just felt like a creep.
And, depending on the state, I might have been breaking the law.
Most of my recordings were in New Jersey and New York, which are one-party consent states. Only one person—me—had to agree to the recording. But if I were in one of about a dozen states that require two-party consent, I’d need permission from everyone in earshot or end up with a possible civil-liability case.
“I would make sure everyone has consented verbally,” says Ashton Kirsch, a lawyer with Matthiesen, Wickert & Lehrer, which maintains a state-by-state list of recording laws. While the risk might be low, he adds, “we could never recommend people take that risk.”
Yet the most unsettling part was realizing the soundtrack of my life was stored on some companies’ servers. Sure, much of our identities are already in the cloud—photos, health records, etc.—but normally, we have some control over what’s there and how it’s stored.
The creators of Bee and Limitless say they encrypt your data, delete recordings when you delete conversations or your account, and don’t train AI models on the feeds. They also say they don’t sell your data to advertisers; they get revenue via hardware and subscriptions sales. Plaud doesn’t sell data either.
Bee’s makers are also testing on-phone AI processing. I tried it out. It wasn’t nearly as good.
So, should you bug yourself—and everyone else—just for convenient to-do lists and daily summaries? For now, that privacy trade-off doesn’t seem worth it. But as these AI assistants get more helpful and more humanlike, all bets are off.
Besides, it’s only a matter of time before this always-on recording is baked right into our existing smartglasses and smartwatches. Just remember to mute before you hit the bathroom, OK?
I’ve been wearing a wire everywhere since February.
I’ve got all the transcripts: important meetings, arguments with my kids, chats with disgruntled employees, late-night bathroom routines. There’s plenty more that I can’t share if I want you (and my bosses and my family) to keep liking me.
No, I’m not an FBI informant. I willingly wore a $50 Bee Pioneer bracelet that records everything I say and uses AI to summarize my life—and send me helpful reminders. I also tested two similar gadgets: the $199 Limitless Pendant and the $159 Plaud NotePin. These assistants can recall every dumb, private and cringeworthy thing that came out of my mouth.
Is this the dawn of the AI surveillance state? Absolutely. Is it also the dream of hyper-personal, all-knowing AI assistants coming to life? Also absolutely.
Within hours of wearing the Bee, I was blown away at how quickly it turned ramblings and random chatter into useful, actionable information. Yet, allow me to quote myself from Feb. 24 at 5:15 p.m.: “This bracelet is really f—ing creepy.”
For years, we’ve been told there’s no way social-media apps are secretly listeningthrough our phones’ mics to target ads. Picking up all that audio? Too hard! Processing it? Too intensive! Storing all the data? Too much!
Yet both the Bee and Limitless do that, with small microphones that listen for voices—particularly your voice. When they detect dialogue, they stream the audio to your phone via Bluetooth, then to company servers where it’s transcribed. AI models take the transcription and generate summaries, which appear in the apps within minutes.
The Bee doesn’t save the audio after transcription; you can’t listen back to anything that’s been said. Limitless keeps the audio, letting you play back full recordings.
Both have physical buttons to quickly cut the mics. The Limitless displays a small light when recording. The Bee only shows a red light when it’s not recording. Nothing says “Don’t worry, it’s off” like the universal symbol for recording! The company says a coming update will reverse this, so you see a light when it’s listening.
With multiday battery life, they capture big stuff—that important pitch meeting—and minutiae, like finding your missing TV remote under the dog.
Plaud is different: It’s a meeting assistant you manually start and stop, then decide if you want to create an AI transcript and summary.
HOW THEY’RE HELPFUL
AI is nothing without data. When you feed it everything you’ve said for days, weeks and months, it gets infinitely more useful. Also, yes, it becomes a lot like a “Black Mirror” episode, but we’ll get to that.With massive transcripts of your life, the AI in these apps can:
• Summarize: These apps recap your conversations, often reading like a bad biography. Bee summary from April 9: “Joanna’s day was a blend of familiar responsibilities and intense professional engagements…She ended the day listening to music by Sting.” Riveting stuff. Can’t wait for the movie adaptation.
The transcriptions themselves aren’t all that accurate but the summaries usually are. Well, except for March 24: “Conversation with Johnnie Cochran about trial evidence.” Yep, just a casual chat with a deceased celebrity lawyer. (I was watching the new O.J. documentary.) Ethan Sutin, co-founder and chief technology officer of Bluush, Bee’s creator, says a recent update has improved the ability to tell recorded voices from real, living ones.
• Remind: Turns out, I promise to do a lot of things without putting them on a to-do list. Bee listens for action items and adds them to a suggested list. It’s repeatedly reminded me of important tasks, like calling the plumber or following up on work stuff. But it also hilariously adds things I’d never put on a list, like “check in on your sick son”or “schedule a follow-up with your hair stylist to discuss your haircut.”
Analyze: Both Bee and Limitless have chatbots so you can ask about your recorded life. I asked Bee for a detailed breakdown of my cursing habits. (Daily average: 2.4 curses.) But it can be genuinely helpful. “Look through my chats with Ethan from Bee and tell me what AI models it uses.” The answer: a combo from Anthropic, Google and Meta.
HOW THEY’RE CREEPY
Feb. 23, 5:15 p.m., in conversation with my mom: “This bracelet has nothing to do with fitness. It records everything that’s being said.”Nobody I talked to over the past few months would have known I was recording them, if I hadn’t told them. It’s a little fun, like I’m a low-budget Ethan Hunt. Mostly, though, I just felt like a creep.
And, depending on the state, I might have been breaking the law.
Most of my recordings were in New Jersey and New York, which are one-party consent states. Only one person—me—had to agree to the recording. But if I were in one of about a dozen states that require two-party consent, I’d need permission from everyone in earshot or end up with a possible civil-liability case.
“I would make sure everyone has consented verbally,” says Ashton Kirsch, a lawyer with Matthiesen, Wickert & Lehrer, which maintains a state-by-state list of recording laws. While the risk might be low, he adds, “we could never recommend people take that risk.”
Yet the most unsettling part was realizing the soundtrack of my life was stored on some companies’ servers. Sure, much of our identities are already in the cloud—photos, health records, etc.—but normally, we have some control over what’s there and how it’s stored.
The creators of Bee and Limitless say they encrypt your data, delete recordings when you delete conversations or your account, and don’t train AI models on the feeds. They also say they don’t sell your data to advertisers; they get revenue via hardware and subscriptions sales. Plaud doesn’t sell data either.
Bee’s makers are also testing on-phone AI processing. I tried it out. It wasn’t nearly as good.
So, should you bug yourself—and everyone else—just for convenient to-do lists and daily summaries? For now, that privacy trade-off doesn’t seem worth it. But as these AI assistants get more helpful and more humanlike, all bets are off.
Besides, it’s only a matter of time before this always-on recording is baked right into our existing smartglasses and smartwatches. Just remember to mute before you hit the bathroom, OK?
