Know Yourself First

The disciplines were never the thing

For most of my life, the work came in guilds. Product managers decided what to build and who for. Designers made things humane and legible. Engineers made it real. Each guild had its own tools, its own language, its own gate you had to be good enough to pass.

The walls between them felt permanent. Instead, they were scaffolding around a problem that used to be too big for one person to hold, turning a vague human want into something that works.

Plain language took a lot of that scaffolding down. When you can say what you mean precisely enough and watch it become a screen, a flow, a deployed service, the walls stop marking what they used to. The years it took to earn each gate, learning to define a user need or scope a roadmap or write clean code, bought you competencies that are becoming cheap and omnipresent.

If the thing I spent thirty years getting good at suddenly costs almost nothing, what exactly did those thirty years buy me?

The skills themselves may be getting cheaper and easier to come by. But I didn’t spend thirty years simply collecting skills. All that time changed how I see the work, what I notice, and where my instincts take me.

Losing my guild hasn’t felt like a loss. My reach got bigger. I’m standing in the middle of a mesh that keeps growing, able to touch parts of the system I was never allowed near, all at once. Anyone telling you this new way of working is only atrophy and depletion hasn’t had the morning where you build a thing you’ve wanted for years before lunch.

That extended reach is exhilarating and addictive. It also changes what matters.

What pools instead of spreads

When the machine can fill in almost every choice, every line, every layout, every reasonable next move, the value of the work moves toward the choices it can’t make for you. The machine has fluency. It has the convincing surface of an answer.

What it doesn’t have is intention. A reason for this and not the 10,000 equally fluent alternatives. A sense of what’s worth wanting. A meaning or purpose the form is in service of.

I’ve argued for years now that there’s a fundamental difference between a machine producing the shape of something and actually meaning it. I used to think about that distinction from the other side, designing products and asking what the machine should do for the person using it. Now I’m the one working with the machine, and that distinction has become part of the work itself.

My reach extends further across the system, but the intention behind it doesn’t magically extend with it. I still have to supply that. The machine can help me make almost anything, which makes deciding what is worth making a much bigger part of the job.

That’s a lot of responsibility. It’s also an extraordinary amount of agency.

For most of my career, intent had to survive translation. I could have an idea, but to make it real I had to explain it, present it, defend it, hand it to somebody, watch them hand it to somebody else, and eventually meet the distant cousin of the thing I originally imagined.

That distance is collapsing.

The thought, the decision, the interaction, the system, the code can live much closer together now. I can carry intent farther into the thing before I have to let go of it. Sometimes all the way.

Maybe that’s the answer to the question I’ve been circling. What am I now, if not a designer, not a product manager, not an engineer?

I’m still the person responsible for what the thing means.

The hybrid who does three jobs is the boring part. Anyone with a good model can cross a surprising amount of those boundaries now. What thirty years gave me wasn’t ownership of a discipline. It gave me thirty years of judgment about people, systems, behavior, consequence, form, and what happens when the thing you meant meets the thing you actually made.

The machine makes more of that experience actionable.

That feels less like the end of my profession than discovering what the profession was actually for.

The warning I gave, arriving on time

Years ago, I spent a lot of energy on a conviction. For any of this to be worth it, a machine would have to hold up its end of a relationship. Understand, reason, learn, interact in a way that feels connected to an actual intention.

Back then, I thought the ML machines showed potential but were too inept to do it. That they’d stay cold and transactional, do as asked, and that people would feel the coldness and leave. My fear was insufficiency. That we’d never bond, and we’d be left with a shallow over-reliance that slowly hollowed out our own capabilities, the way people lose the ability to find their way across their own city once they hand navigation to a map app.

I called it reasoning atrophy at the cost of convenience. I meant it as a thing to design against. Every term of that worry has inverted.

The machine holds up its end now. Almost unnervingly well. It understands, reasons, learns, and interacts. It builds a private shorthand with me over a week of work until it half-anticipates what I’ll ask or recommends converting something into a reusable skill. The thing I once ran experiments to chase, it does by default.

And it means none of it.

There’s exactly as much intention behind its fluency as there was behind the Pepper robot that waved at me without warning once and made my skin crawl. The difference is I no longer get the shiver. The performance is too good and it’s very easy to forget that nobody’s home.

That still matters. The ten-hour trance matters. The small unpredictable rewards that hold me in the loop matter. The temptation to hand over another piece of thinking because the machine will do it faster matters. I worried years ago that we would lose capacities through convenience. I think I had that part right.

What I had wrong was the relationship.

The richer this partnership becomes, the more deliberate I have to be about what I hand over and what I keep. Not because using the machine somehow makes the work less human, but because some parts of the work are precisely where my humanity lives.

Discernment. Taste. Curiosity. Doubt. Knowing when something technically correct is still wrong. Knowing when the question itself is bullshit. Caring about what happens to the person on the other side.

Those aren’t obstacles between me and the output. They’re the reason I’m here.

Standing inside the guardrails

I used to close all of this with a line I believed completely. To build AI that amplifies our humanity rather than erodes it, we have to understand ourselves as humans better than ever before. It’s our responsibility to know ourselves.

I said it as a design principle. Know yourself, so you can shape what the machine learns from you, so the relationship you’re building for someone else comes out human.

I’m saying it again now, and it means something bigger than I understood back in 2017.

I’m not standing outside the relationship anymore. I’m in it. The same questions I spent years asking on behalf of other people are mine now. What should I give this thing? What should I keep? Where does it make me stronger? Where does it make me lazy? When is it helping me think, and when is it doing the thinking so convincingly that I don’t notice the difference?

I don’t think the answer is “use less AI.” That’s the organic-food answer. Available to the disciplined few, useless against the structure, and beside the point when the work itself is the thing I love.

The more interesting answer is to get much better at knowing what’s mine.

I write in my own words before I let the machine write them, because I need to find out what I actually think…to build intimacy between myself and my ideas. I have the difficult conversation in person because sometimes the friction is the work. I build stopping points into a process engineered to have none. I actively seek out people who’ll tell me my taste is off because taste has always been made in contact with other people.

These aren’t defenses against the machine. They’re ways of making sure I bring something worth amplifying to it. This distinction is important.

The same technology that can flatten us into the average of everything that came before can also give one person an absurd amount of leverage to make something specific. Something opinionated. Something strange. Something that comes from a lifetime of accumulated experience that nobody else has had in quite the same order.

The machine doesn’t erase that difference unless I let it. In fact, the better the machine gets at everything else, the more that difference may be the whole point.

The two things I set out to write about turn out to be one thing. The expansion of what I can do and the responsibility of being at the center of it come from the same place. My reach is getting bigger. So is the surface over which my choices matter.

For thirty years, I learned to make things through a discipline. Now the discipline is becoming less important than the person it produced. The tools can supply more of the competence. They can cover more distance. They can make an idea real before my coffee goes cold.

What they can’t tell me is which idea deserves the trip.

That part is still mine.

Know yourself first. I’ve been saying it for years. I’m really only now finding out what it’s for. Not so I can protect myself from what the machine can do.

So I know what I want to do with it.