I've been using AI every day for a bit over three years now. This piece is a quiet observation, not a sales pitch. I'm not going to tell you AI is changing the world or that it's coming for your job. I'm going to tell you what I've actually noticed, in the small group of real people around me, when they start using it seriously.
The word itself
The thing that keeps coming back to me is empowerment. Not in the corporate-branding sense of the word, the one you see on HR slides. I mean it in the very specific sense that a lot of things people used to need permission to do, they now do on their own.
The close friend metaphor
The framing that works best for me is this. Picture a close friend who happens to know something about almost everything. This friend is always available. You can ask them anything, any time, and they'll answer. And then they'll answer the follow-up, and the one after that, and the one after that. They won't get annoyed when you ask something you probably should already know. They won't charge you for their time. They'll meet you wherever you are in your understanding of a problem and work their way up with you.
That's what a real working relationship with AI looks like once you spend time with it. It's not magic. It's a back-and-forth that gets better as you get better at asking.
The learning happens in the loop
The thing I see with people who are just starting is that they expect AI to give them the answer on turn one, and when it doesn't they get disappointed and conclude the whole thing is overhyped. The people who actually get value out of it are the ones who stay in the loop. Ask something, read the answer, notice what's off, ask again, refine, iterate.
That process isn't a workaround for AI's limitations. That process is the work.
And here's the part I think gets missed. That loop is also how you learn. Every question you ask sharpens how you ask the next one. Every answer you read teaches you something about the domain you're asking about. A few months into that habit and you don't just have outputs. You have a new set of instincts.
It’s just a tool
I keep coming back to this because the AI-is-magic and AI-is-scary camps both miss it. AI is a tool. You use a tool to achieve a specific goal. You use a hammer to put in a nail. You use AI to build, or fix, or understand, or just to have a second brain around while you think through something.
Some of the things I've watched people do with AI over the past year, some of it my own work and some of it people around me: building a multi-agent content pipeline that runs at the day job. Running a personal agent that coordinates across five messaging channels. Translating recipes, or asking for new ones. Building a small educational game for your own kids. Shipping a weekend project in an afternoon that would have been a month a few years ago. Different scales, different goals. All of them count.
The barrier is lower than it’s ever been
The observation that makes me most optimistic is this one. You can have a kid in their room running three or four AI subscriptions, iterating on different agents, building things that would have needed years of study and a computer science background a decade ago. The distance from "I have an idea" to "I have a working version of it" has collapsed in a way that isn't getting talked about enough, and it's the most interesting thing happening right now.
The people who get called "vibe coders" in the online discourse are often dismissed as shallow. I don't see it that way, honestly. A lot of the ones I've watched are going deep, building real systems, making decisions a professional engineer would make, and shipping things that work. The tools gave them the leverage. Their interest and their iteration did the rest.
The grain of salt
I'm not going to pretend AI empowerment has no ceiling. A kid in their room with ten AI subscriptions isn't going to build a surgical robot. Domains where quality, compliance, and qualification matter structurally aren't where this story plays out, and they shouldn't be. There are places where human expertise, professional review, and formal training are the whole point, and no amount of back-and-forth with a language model replaces any of that.
But empowerment doesn't require solving every problem. It requires moving the line on what a regular person can do on their own. And that line has moved a lot, in the three years I've been paying attention.
What it means for me now
When I started, I thought of AI as a speed multiplier. Something that would help me ship the thing I was already going to ship, just faster. That turned out to be wrong, or at least incomplete. What it actually became was a range multiplier. I ship things I wouldn't have attempted. I ask questions I wouldn't have asked. I go deeper into problems I would have walked away from a few years ago.
That's the empowerment I mean. Not the HR version. Not the TED talk version. The quiet version, where a regular person with a little access and a little patience can do more with their day than they could a month ago, and keeps getting better at it.