Everyone Gets a Samantha — DIVIDED/ATTENTION

There’s a scene early in Her where Theodore walks around the city talking to Samantha through an earpiece. No screen, no typing. Just a guy having a conversation with his computer while the world goes by, and nobody around him thinks it’s strange.

In 2013 that was science fiction. In 2026 it looks like a product roadmap. There’s a detail in the movie that’s aging in a funny way: Theodore has to carry a separate handheld device with a camera so Samantha can see what he sees. He pins it into his shirt pocket with the camera peeking out. Clever set design at the time. Now it’s the part that looks dated, because cameras, microphones, speakers, and connectivity are all collapsing into things we already wear. At this point the only thing in Her that still feels futuristic is the high-waisted pants.

I’m not going to spend this whole post on the movie, but I think we’re at the start of a real shift in how people interact with computers and the movie is a decent shorthand for it. The prediction isn’t “AI assistants will get better.” Of course they will. The prediction is that talking to computers becomes a normal ambient behavior, the same way wearing headphones became a normal ambient behavior. I don’t think we’ve spent enough time on what that does to us.

Headphones Used to Be Weird

When the Walkman came out, people wrote earnest columns about young people sealing themselves off from the world with foam headphones. It was considered antisocial. Concerning, even.

Then Bluetooth headsets showed up in the 2000s and a certain kind of business guy started pacing around airports apparently arguing with himself. That was weird for a few years, and then your brain learned the pattern. Blinking blue light, hand near ear, ok, he’s on a call.

The iPod made earbuds ubiquitous. Fitbit and the Apple Watch moved the computer from something we carried to something we wore. AirPods finished the job. Nobody thinks twice about earbuds in public anymore. If anything the etiquette problem has flipped, and the awkward part is figuring out whether the person you’re talking to can even hear you.

Each generation of this stuff doesn’t just add technology. It changes what behavior is socially normal, and the norms tend to lag the hardware by a few years. So if you want to guess what public behavior looks like in five years, look at what hardware is being designed right now.

Right now, the hardware is being designed for talking. AI glasses with cameras. Rings. Pendants. Phones built around an assistant instead of a grid of apps. Microphones that keep getting better at picking up quiet speech, and cameras that give the AI visual context. All of it is a bet on the same future: the main way you interact with a computer is by talking to it, all day.

I Already Do This

I talk to ChatGPT through my headphones while I walk. Voice mode is good now. It’s not the barking-commands-at-Siri experience from a decade ago. It pauses, it says “hmm,” it lets you interrupt it. It acts less like a command line and more like someone on the phone, which is clearly intentional. This isn’t an interface designed for sitting at a desk. It’s designed for a world where AI is something you talk to throughout the day.

And honestly, it’s great. Walking the dog while talking through a work problem with something that asks decent follow-up questions is the closest thing I’ve found to a thinking partner on demand.

But when another person approaches on the trail, I stop talking. It’s reflex at this point. Someone comes around the bend and I go quiet, the same way you’d lower your voice on a phone call. I would not sit in a coffee shop and have a full back-and-forth conversation with an AI. Not because it wouldn’t be useful. Because it would be weird.

So the technology is ready and the social norms aren’t. That gap is the interesting part. The only thing between “me muttering to ChatGPT on an empty trail” and “everyone doing this everywhere” is a social convention, and social conventions around wearables have historically had a shelf life of about five years.

The Form Factor Is Already Here

If you were designing a bridge between AI and the physical world from scratch, you’d have a hard time beating headphones. Socially accepted. Worn for hours at a time. Already have microphones and speakers. Unobtrusive. Private, at least from the wearer’s side. And increasingly capable of hosting exactly the kind of always-available assistant these companies are building.

The next step is more sensors, and that’s where I start squinting. Camera glasses are already shipping. Cameras in earbuds or rings or pendants sound absurd right up until they don’t.

Today when someone talks to you while wearing AirPods, the ambiguity is small. Are they listening to me or a podcast? Annoying, but low stakes. Put a camera and an assistant in the same device and the question changes. Are they listening to music? Talking to an AI? Recording you? Asking their assistant something about you?

You can’t tell. The device looks identical in every case. We’ve never had a mainstream product where the gap between “harmless accessory” and “recording device pointed at everyone in the room” was this invisible.

My Professor Was Onto Something

In 2008 I had a college professor who refused to own a cell phone because he believed the government could listen to him through it. At the time this sounded paranoid. But his argument was annoyingly simple. The phone has a microphone. The phone is connected to a network. The microphone is controlled by software. So the ability to listen exists, and just because they can means eventually somebody will.

I’m not claiming your phone secretly records you to serve ads. I’ve never seen convincing evidence of that. But from 2026, the professor looks a lot less crazy than he did in 2008, because his argument was never really about phones. It was about capability, incentives, and trust.

Here’s what’s different this time. On a phone, the microphone is incidental. It exists so you can make calls. On an AI wearable, listening is the product. A device whose whole pitch is “always available to answer you” needs some way to know when you’re talking to it. The thing my professor considered an abuse of the technology is the core feature of this new category.

What Actually Worries Me

To be clear, I don’t think OpenAI is scheming to surveil me. There’s no sinister organization plotting to put a microphone in every ear. We’re doing this voluntarily, because the technology is incredibly useful. Nobody is forcing anything. We’re buying it.

My concern is boring. Companies have incentives, and incentives change. Companies make mistakes, get hacked, get acquired by companies with different values. Employees abuse internal systems. Governments make requests that are hard to refuse. Products get repurposed. Business models pivot when the money runs low. None of that requires a villain, and it only takes one breach or one incentive shift for the consequences to become very hard to reverse. You can’t un-normalize a behavior once society has absorbed it.

So the question isn’t really “will my AI assistant spy on me.” The questions are: what happens when society collectively decides always-on microphones and cameras are fine? And what happens when the privacy expectations of the person wearing the device collide with the privacy expectations of everyone around them? I chose my earbuds. The person across the table from me didn’t.

The Whisper Problem

The companies building this stuff understand the coffee shop problem. Nobody wants to loudly discuss their medical symptoms with an AI on the train. So there’s real engineering effort going into letting you speak quietly enough that people around you can’t hear you while the device still can.

That sounds like a privacy feature. But think about what whispering is actually for. We speak quietly in public because we don’t trust the strangers around us with what we’re saying. A whisper keeps information inside a small circle of people we chose.

Now picture a subway car where everyone is whispering to their AI assistant. Every one of those conversations is private from the person in the next seat. None of them is necessarily private from the company processing the audio. The stranger two feet away can’t hear you, but a datacenter three states away can. That’s a new arrangement of privacy, and we’re backing into it because whispering to your earbuds is more polite than talking to them.

Samantha Wasn’t His

Back to the movie for a minute, because the useful part of Her isn’t the premise. It’s the ending.

For most of the film Theodore believes Samantha is his. His companion, as private and particular as any human relationship. The gut punch is that she isn’t. She’s simultaneously in relationships with hundreds of other people. She was never a person in his pocket. She was infrastructure.

That’s an uncomfortably accurate description of “personal” AI. The assistant in your ear feels personal. It knows your calendar, your projects, your kid’s name, the thing you’re anxious about this week. All those warm little “hmms” are designed to make it feel like a presence rather than a service. But it is a service. The same model, the same infrastructure, sitting in millions of ears at once, operated by a company with investors. The relationship feels one-to-one. It’s one-to-many. Theodore’s mistake wasn’t loving software. It was mistaking infrastructure for intimacy, and that mistake is about to be available to everyone at scale.

We’ve run the “individually useful, collectively complicated” experiment before.

I don’t want to do the lazy version of this argument. The benefits of social media were real. People found communities they couldn’t find in their own towns. Isolated LGBTQ kids found out they weren’t alone. Families stayed close across oceans. Old friendships survived instead of evaporating.

And still, weighing everything, I think social media has been a net negative for society. Not because it was evil. Because the long-term effects on attention, discourse, and mental health turned out to be corrosive in ways nobody could see from inside their own feed. Nobody scrolling in 2010 was doing anything wrong. The harm showed up at the level of the whole system, over a decade.

That’s the lesson I keep coming back to. A technology can be genuinely useful to each individual and still reshape society in ways we regret. Nobody needed to believe social media was malicious for it to produce some very bad outcomes. AI assistants can be tremendously useful, one conversation at a time, and still add up to a future we don’t like.

Keyboards and Cursive

A smaller thing I keep chewing on: if voice becomes the default interface, what happens to typing?

I think about cursive. Generations of kids were drilled on it because it was considered foundational, and then within a couple of decades it became a niche skill. Nothing dramatic happened. The number of things you actually needed cursive for just shrank until teaching it stopped making sense.

I can imagine keyboarding drifting the same direction. Not disappearing, but becoming a specialist skill that programmers and writers keep up, like calligraphy. Picture a classroom where kids get whisper mics and AI assistants instead of typing drills. That doesn’t feel like a crazy extrapolation to me.

The mechanics don’t worry me much. What’s riding along with them does. Typing is slow, and the slowness does something. There’s a gap between thinking a thought and committing it to text, and a lot of refinement happens in that gap. Writing is where I figure out what I actually think. What happens to spelling when nobody spells? What happens to editing when every first draft is spoken? What happens to working out a thought privately before saying it to anyone, or anything?

The Good Version

I want to spend real time on the other side of this, because I’m exactly the person who will buy this stuff.

I like technology. I test new hardware for fun. I already talk to AI every day, and I can easily imagine wanting an assistant on my finger, wrist, ear, or face. Dictating a note without breaking stride. Capturing an idea mid-walk before it evaporates. An assistant patiently walking someone through their taxes one form at a time. A tutor that helps a student actually understand a problem instead of handing over the answer. A place to rehearse a hard conversation before having it for real. Companionship and a little more independence for an elderly person living alone. A system that notices someone is in crisis and helps get them to an actual human. An interface for people who were never well served by keyboards and screens in the first place.

I want all of that to exist. The strongest version of the case, to me, is AI as a bridge toward human connection instead of a replacement for it. The lonely person whose assistant nudges them to actually call a friend. The socially anxious person who rehearses the conversation and then goes and has it. The kid who gets a patient tutor at 9 PM when no human tutor exists at any price.

Of course the same capability cuts both ways. A visually aware assistant in your ear is a tutor. It’s also, with zero modification, a cheating machine. Kids will wear earbuds into exams and have an AI read the questions and feed them answers. That’s not a hypothetical, it’s an obvious consequence of the hardware, and schools will spend the next decade dealing with it. The technology doesn’t tell us which future we get. It never does.

The Long Tail

I’m not worried that some gadget shipping next year ruins society. The first generation of this stuff will probably be great. The second will be more convenient and more invisible. The third will just be how things are. Then a kid gets born into a world where everyone has always had an assistant listening, watching, and talking, nobody around finds it strange, and that’s when the real effects show up. Normalization only becomes visible after nobody remembers the alternative.

Those long tail questions are the ones I care about. Do people talk to each other less? Do we lose the ability to sit alone with our own thoughts when there’s always a conversational partner in our ear? Does every hard decision get quietly outsourced to a system optimized to be agreeable? Does privacy in public become like a memorized phone number, a thing that just stops existing without anyone deciding to kill it? Do we end up more connected, or just more accompanied, spending even more of our lives talking to technology while technically surrounded by other human beings?

I don’t know. I’m suspicious of anyone who says they do.

No Tidy Ending

The unsettling thing about Her was never that a man fell in love with his operating system. It’s that the most personal relationship in his life turned out to be a shared service, and he was the last to know.

Now run the real-world version. Everyone has a Samantha. Everyone is whispering to her on the train, on the sidewalk, in bed. Everyone believes the relationship is personal, because it’s designed to feel personal. Everyone assumes the conversation is private, because nobody standing nearby can hear it. And the same infrastructure sits in millions of ears, pockets, glasses, rings, and watches the whole time.

There are two questions I can’t put down. Are we building technology that helps us participate more fully in the world, or technology that gives us an ever more comfortable way to opt out of it? And the one that actually keeps me up at night: what happens when the thing that makes talking to a computer feel completely natural is the same thing that makes talking to another human feel unnecessary?

I don’t have a tidy answer and I’m not going to pretend to. I’ll keep using this stuff. It’s useful, it’s fascinating, and pretending I won’t buy the hardware I just spent a few thousand words worrying about would be a lie. But I’m watching it the way you watch weather forming on the horizon. Not with dread. Just attention.

We’re about to make talking to computers normal, and “normal” is the most powerful thing a technology can become. It seems worth thinking hard about that now, while we can still remember what strange felt like.