Everyone You Know Is Talking to an AI Now — and They're Talking About What It Said

2026-07-19

Published July 19, 2026. An observation from walking through cities and watching people live with AI.

I walk through Bogotá and see it everywhere. On the TransMilenio, a woman in business attire has a ChatGPT window open, scrolling through a thread — not asking one question and closing it, but reading. A conversation. On a café table in Chapinero, a student has Copilot pulled up next to a half-written essay, the two windows side by side like they belong together. In line at Éxito, a man is talking to his phone — not on a call, the screen is too bright for that. He's dictating something to an AI. "How do I cook the chicken so it doesn't dry out," he says, and the phone speaks back.

This isn't tech workers. This isn't early adopters. This is everyone.

My partner talks about what the AI told them. Not in the way people talk about a useful tool — "oh, I used the calculator" — but in the way people talk about a conversation. "ChatGPT said I should try restructuring the argument first, and then it gave me three examples." The verb "said." Not "generated." Not "returned." Said. As though there's a person on the other end delivering counsel.

A friend of mine, who has never written a line of code and has no interest in doing so, now spends twenty minutes a day talking to Claude. Not for work. Not for productivity. She asks it questions about her garden. She tells it about her day. She describes it as "useful" in the same tone she uses for her therapist. The AI is not a search engine with better formatting. It's a presence. One she consults regularly and references in conversation the way people used to reference what they saw on Twitter or what someone posted on Instagram.

Something has shifted.

The Social Evidence

For two decades, the standard image of someone on their phone in public was a person scrolling — Instagram, Twitter, TikTok, whatever feed dominated that year. The face was passive. The thumb moved up. The content washed over them.

What I see now is different. People aren't scrolling. They're typing. They're reading. They're looking at a screen and then looking up, thinking about what they just read, then looking back down and responding. It's not consumption. It's interaction. The physical posture is different — it's the posture of someone in a text conversation with a person they care about, not someone doomscrolling a feed.

I've started noticing it everywhere. The woman on the train who's clearly asking ChatGPT for advice before a meeting. The student in the library who has Perplexity open and is cross-referencing something across tabs. The couple at a restaurant where one person pulls out their phone, types something into an AI, and then shows the other person the response — and they both react, discuss it, argue with it.

The AI output has become something you show someone. The way you used to show someone a text from your mom, or a funny tweet, or a photo. "Look what it said." The AI's response is social currency.

Programming Was Never This Hot — Not as a Job, But as a Thing People Do

There's a strange reversal happening. For years, "learn to code" was career advice — a path to a stable job, a good salary, an in-demand skill. Coding was hot because coding was lucrative.

Now coding is hot because it's interesting. Not as a career — as an activity. As a thing people do for the same reason they watch cooking shows or follow sports. You might never write a line of production code. But you've used an AI to build something — a personal dashboard, a script that renames your photos, a website for your friend's wedding. You've watched YouTube reviews of AI coding tools the way you watch car reviews. You have opinions about which model "codes best" despite never deploying anything to production. You're a spectator, but the line between spectator and participant is thin enough that everyone has tried it at least once.

The AI coding influencers and the car review channels are the visible part of this, but they're riding a wave that's much bigger than they are. The wave is: interacting with AI has become a normal human activity. It's not a developer thing. It's not a tech thing. It's what people do when they're waiting for the bus.

When something becomes a normal human activity — the way texting became normal, the way googling became normal — it stops being a thing you do for a reason and becomes a thing you do because you have a minute. The friction is gone. The barrier is gone. The need for a specific use case is gone. You just... talk to it. About anything. About everything. Because it's there, and it answers, and people have started to care about what it says.

What People Actually Talk About

The conversations I overhear are not technical. Nobody on the TransMilenio is discussing the SWE-Bench Pro implications of GPT-5.6's tokenizer architecture. They're asking whether the chicken needs more time in the oven. They're preparing for a difficult conversation with their boss. They're trying to understand a medical term their doctor used. They're arguing with the AI about a sports statistic, and the AI is winning, and the person is laughing.

The content of these interactions reveals something the benchmark race obscures: most people using AI are not evaluating its reasoning capability. They're evaluating its presence. Is it helpful? Does it understand what I'm asking? Does it talk to me like a person or like a manual? The technical benchmarks matter to the people building the models. The vibe matters to everyone else.

The strangest thing about this normalization is how fast it happened. In 2022, using an AI assistant was weird. In 2024, it was a developer thing. In 2026, my partner talks about what ChatGPT said the way they talk about what their friend said. The verb "said" implies a speaker. And people have started to hear one.

The Great Equalizer

There's something about AI that distinguishes it from every previous wave of digital access, and it's this: you don't need to be anything to use it.

Email was universal — anyone could sign up for Hotmail or Yahoo or AOL. But email required someone to email. The internet was universal — anyone could browse. But browsing required knowing what to search for. Social media was universal — anyone could post. But posting required an audience, a performance, a version of yourself you were willing to broadcast.

AI doesn't require any of that. You don't need to have friends on the platform. You don't need to know the right search terms. You don't need to present yourself. You explain what you want, in whatever words you have, and the AI meets you there.

A grandmother asking how to cook chicken so it doesn't dry out. A developer asking for a React component with specific state management. A teenager asking for help with a history essay. A mechanic asking for the torque specs on a 2015 Corolla. The AI does not care who they are. It does not change its tone for age or gender or profession. It treats a question from a 14-year-old and a question from a CEO the same way: as text that means something, that deserves an answer.

This is not how most technology works. Most technology requires identity. Gaming requires you to be a gamer. Beauty apps require you to be into beauty. Coding tools require you to be a coder. The shape of the tool defines the shape of the user. AI is the first tool with no shape at all — it takes the shape of whatever you ask it.

The result is a tool where the playing field is, for the first time, almost genuinely flat. A grandmother in Bogotá with a free ChatGPT account and a developer in San Francisco with a $200/month subscription are using the same underlying capability. The paid user gets faster responses, a higher rate limit, access to the latest model. But the free user can still ask the same question and get back something useful. The gap between tiers is a gap of convenience, not capability.

This is different from every previous technology wave. The internet gave access to information, but finding it required literacy — digital literacy, language literacy, search literacy. Social media gave access to community, but participating required social capital. AI gives access to doing things — generating text, analyzing information, solving problems, creating content — and the only prerequisite is knowing what you want and being able to explain it. That's a lower bar than anything that came before.

We're not at true universality yet. You need internet access. You need a device. You need to speak one of the languages the models support, and the supported languages still skew heavily toward English and a handful of others. The free tiers have limits. But the direction of travel is unmistakable: toward a tool that anyone can use, for anything, just by asking.

The grandmother and the developer are equal in the eyes of the machine. That's not a technical achievement. It's a social one. And it's why, when I walk through the city and see ChatGPT windows open on every kind of phone held by every kind of person, I'm not looking at a technology being adopted. I'm looking at a technology that has become infrastructure — something everyone uses, regardless of who they are, because the alternative of not using it has stopped making sense.

This is an observation piece grounded in both lived experience and verifiable data. ChatGPT surpassed 900 million weekly active users and 1 billion monthly active users in mid-2026 (DemandSage, July 2026), receiving approximately 2.5 billion prompts per day (TechnologyChecker.io, July 2026). India became its second-largest market at 100 million weekly active users, growing 41% year-over-year; Brazil, Indonesia, and Southeast Asia are expanding rapidly (Omnibound, 2026). The platform has some of the highest engagement metrics of any web platform — deep, habitual usage rather than casual drive-by traffic. The scenes described here — people with ChatGPT open on transit, in cafés, in grocery store lines — are drawn from direct observation across Bogotá and from conversations with people across different demographics who have integrated AI into their daily rhythm.