Someone You Recognize

“Big man in a suit of armor. Take that off, what are you?”

“Genius, billionaire, playboy, philanthropist.”

“I know guys with none of that worth ten of you.”

That exchange between Captain America and Tony Stark in Civil War has been following me around for weeks. Part of the reason is that, for the first time in my life, the question doesn’t feel like it’s meant for someone else. It feels like it’s meant for me. Not literally, obviously. But close enough that I can’t quite shake it.

I can’t remember the last time I didn’t have Claude Code open on my terminal. My “suit” consists of three or four agents running at once, tokenizing away while each one writes, rewrites, explains, suggests, and builds. They help me debug at 2 AM. They outline papers I don’t fully understand yet. They one-shot projects that sound impressive enough to put on a resume. Sometimes they tell me what to say in interviews, in arguments, in conversations I don’t know how to have on my own. It is fast. It is useful. It works. Every week I wake up to another tool, another release, another layer of my life being automated without any friction. I’m amazed. I’m also somewhat dissatisfied.

The Thought Experiment

Imagine a world where everyone woke up tomorrow with their own Iron Man suit. No cave in Afghanistan, no training arc, no trial and error. Overnight, every person on Earth can fly, fight, build, analyze, and create at a level they could never reach on their own. What would people actually do with it?

Would they go out and solve real problems? Would they build something that mattered? Would they push the boundaries of what anyone thought was possible? Or would they, mostly, use it to get through the day a little faster, to make the boring parts shorter, to look slightly more impressive to the people they were already trying to impress? When I look at how I use AI, and how most people around me use it, it isn’t for anything heroic. It isn’t particularly ambitious. It’s convenient. Write this faster. Finish this assignment. Generate a few ideas. Polish something until it sounds smarter than the thinking behind it actually was. Ship a startup that might make a little money, or at least look good in a pitch. Tony Stark built his first suit to survive. We use ours to keep up.

Cost

Tony’s suit means something to him because of what it cost to build. He hammered the Mark I together in a cave, with a car battery wired into his chest, while the men who kidnapped him waited outside the door. The suit wasn’t a productivity tool. It was the thing standing between him and dying and successfully democratizing world peace. Every version he built after carried the weight and conviction of that first one. Mine came by subscription. Fifty dollars a month, depending on which tools I’m experimenting with.

There is no cave. There is no cost. And I keep wondering whether a capability that arrives without any cost at all can actually belong to me, or whether it’s just something I’m borrowing for as long as the servers stay up and the subscription keeps renewing. Cap, for all the preaching he does in that scene, isn’t so simple himself. Tony fires back that everything special about Steve Rogers came out of a bottle too. The serum, the program, the scientist who looked at a scrawny kid from Brooklyn and chose to turn him into something more. Cap is also augmented. He just got there first.

Maybe nobody is unaugmented, not really. Genes, family, money, schooling, the specific people you’re around. Capability has always been borrowed from somewhere. What’s new isn’t the augmentation itself. What’s new is how cheap and sudden and frictionless it has become, and whether something that arrives without any cost can carry any weight at all.

The Treadmill

Walk through campus at Hopkins and everyone is building something. A startup, a research project, a side project, a podcast. A hackathon project that will become a resume line that will become a summer internship that will become a LinkedIn post captioned “excited to share.” The pace is constant. You feel it in the dining hall, in the group chats, in the way people answer “what are you working on” like it’s a status check instead of a question.

AI has made all of this easier, and because it has made it easier, it has also made it more mandatory. “I built an AI agent that” is its own genre now, a demo day pitch and a coffee chat opener and a LinkedIn post rolled into one. When shipping used to take months, not everyone shipped. Now the floor is an afternoon, and every problem is Claude-shaped. When everyone can ship, not shipping starts to feel like a decision you’re making against yourself. The question I have been avoiding, mostly because I don’t like the answer, is how much of what I’ve built actually matters to me. Some of my projects exist because they’d look good in an application. Some exist because a friend was working on something similar and I didn’t want to fall behind. A few I genuinely care about, and those are the ones I work on least, because caring is slow and the treadmill is not.

YCombinator general partner Diana Hu came to JHU recently and said something else I keep replaying, that this is the best moment in history to find a real problem, talk to real people, and ship. I believe her. I also suspect that most of what I see around me, including a lot of what I’m making, is the shape of that advice without the substance of it. “Real problem” gets replaced by convenient problem. “Real people” gets replaced by no people. “Ship fast” becomes the whole of the thing, because the shipping is what gets rewarded and the caring is what’s getting dropped.

Without the Suit

There’s a scene later in the Marvel timeline where Tony takes Peter Parker’s Spider-Man suit away. Peter has screwed up, not out of malice but out of eagerness, the kind of ambitious overreach a teenager with new powers would have. And Tony gives him a line I find myself quoting back to myself:

“If you’re nothing without the suit, then you shouldn’t have it.”

What makes that scene hit is that Peter isn’t lazy. He’s the opposite. He wants to help, he wants to matter, he’s convinced the suit is the path to becoming who he’s supposed to be. Tony takes it anyway, because being good with the suit and being good are not the same thing, and one can stand in for the other long enough that a person stops being able to tell them apart.

That is the version of the AI worry I actually believe, not the dramatic one. I don’t think these tools are making us stupid. That framing is too clean and probably wrong. The quieter worry is that they make it impossible to tell. You write something with AI and it comes out fine. You debug something with AI and it works. You get through the day with AI and the day gets gotten through. The feedback loop that used to tell you whether you were any good—the failure and the struggle and the slow accumulation of can-do-this-myself—gets smoothed flat. You might be getting sharper. You might be hollowing out. From the inside, both can feel the same.

Self-Knowledge

The real thing at stake isn’t intelligence or productivity, both of which the tools will keep making cheaper. The thing at stake is self-knowledge, which gets harder to come by the more of the work of being a person gets outsourced. If you let the suit write and think and speak and decide for long enough, you eventually lose the ability to tell whether the person underneath is someone worth having.

I want to stop here and say what my argument could talk me out of, which is that the optimists are right. The barriers to making something used to be enormous. You needed capital, or a team, or years of specialized training, or access to rooms most people could not get into. Those barriers have collapsed in the span of a few years. A kid with an internet connection and an idea can now ship something that actually works. Someone who was locked out of technical fields their entire life can build a real product in a weekend that would have required a small company ten years ago.

That is not a small thing. It is one of the most democratizing shifts in modern history or all history, and I don’t want to be the person who stands in the middle of that moment and complains. I use these tools every day. I plan to use them tomorrow. The point of writing this isn’t to put the suit down. The point is to notice what the suit is for.

Values

Every system has values poured into it, and the someone pouring is almost never you. Your Instagram feed has values. Your Spotify recommendations have values. The autocomplete that finishes your emails has values, and so does the model writing half (all) of your code. None of these things sat down and thought about what a good life looks like. They were told to optimize for something, usually a proxy like time on app or engagement or watch-through rate, and those proxies have been quietly compounding inside a generation’s attention for over a decade.

Marvel had an obvious version of this story. Ultron wasn’t built to be evil. He was built to protect Earth, and when he ran the numbers he decided humanity itself was the problem he’d been designed to solve. His reasoning was clean. It just wasn’t ours. The less obvious version of that story is the one we’re already living inside. A recommendation algorithm told to maximize watch time that eventually radicalizes a teenager. A hiring model told to find good candidates that learns to screen out anyone whose resume doesn’t look like the last round of hires. A sycophantic chatbot that leads someone to AI psychosis. These aren’t science fiction. They’re product decisions that got shipped, iterated on, and quietly compounded.

What I wonder, mostly in the shower, is what is getting poured into the tools I now use for six or eight hours a day. My coding assistant is optimizing for something. My search tool is optimizing for something. The model I ask for feedback on my writing is optimizing for something. I don’t fully know what those somethings are, and I’m not sure anyone does, including the people who built them. If I spend enough of my day inside those optimizations, at some point they stop being external. They become the defaults I reach for when I’m tired, which is to say, most of the time.

Purpose

Underneath all of this, under the dooming about the job market and the hype about the next release and the arguments about whether junior roles will still exist in five years, there’s a question that doesn’t have a comfortable answer: When everyone has access to a tool as smart as Einstein, and then smarter than Einstein, and then smarter than any human being who has ever lived, what are you for? What am I for?

I don’t know. And I suspect the people selling us answers don’t really know either. Become an AI-native something. Learn to prompt. Get ahead of the curve. Build something, maybe your own startup. Those are answers to a much smaller question than the one I just asked. They tell you how to stay employed. They do not tell you what a person is when the things that used to make a person useful can be rented for twenty dollars a month.

I don’t think this question has a confident answer yet. I think it’s the kind of question a whole generation has to sit inside of for a while before anyone has earned the right to answer it. And a lot of what I see around me, the hustle and the building and the constant shipping, looks like a way of not sitting inside it. If you’re always doing something, you never have to notice that you don’t know why you’re doing it.

Experiments

So: back to Cap. Take that off, what are you. I don’t have an answer. I’m not sure the Substack post earned one. What I have instead is a set of small experiments I’ve been running on myself, and I’ll tell you about them because they feel more honest than a conclusion.

I’ve started writing the first draft of anything that matters to me without AI in the room. Baby steps. Not because I think the AI draft would be worse. Because I want to know what the draft without it sounds like, so I can tell the difference later. I’ve started noticing, when I reach for a tool, whether I’m reaching because the problem needs it or because I’m avoiding the discomfort of sitting with the problem on my own for another thirty seconds. Sometimes I put the tool down and keep sitting. Sometimes I don’t. I’ve been trying, with mixed success, to work on the projects I actually care about before I work on the ones that will look better on a resume, because I’ve noticed that the resume ones expand to fill whatever time I give them, and the ones I care about shrink to fit what’s left.

None of that is a prescription. I don’t know if any of it generalizes. I don’t know if, in five years, I’ll look back and think the people who embraced the tools with no friction were right, and I was over-thinking a pencil. Maybe. What I do know is that there’s a version of my life where I keep getting faster at things I don’t particularly care about doing, and I’d like to pause and notice that before it becomes me.

---

Take the suit off for a minute. Not as a rule. As an experiment. See if the person inside is someone you recognize.