I was drawing long before I knew what Photoshop was.
As a kid, I drew constantly. Action scenes. Characters. Little flip books where I tried to make figures move from one page to the next. I wasn’t thinking about animation or visual communication as disciplines. I just knew there were things I could see in my head that I wanted to get out of it.
Then I encountered Photoshop at the art school I grew up in. I don’t remember thinking I’d discovered some revolutionary piece of technology. I remember something much simpler:
Wait. This thing can do things I can’t do.
I could manipulate an image. Move things around. Change them. Combine them. Try something, hate it, undo it, try something else. Suddenly, I had another surface to think on, one that could move considerably faster than my hands.
But here’s the part that matters: Photoshop hadn’t created the fascination.
I was already drawing. Already making flip books. Already trying to turn the things in my imagination into something I could see. The machine showed up in the middle of something that was already developing.
It didn’t give me the impulse. It gave the impulse more range.
And when I look back, that became my relationship with technology for most of my life.
Photoshop became other design software. Design software became websites. Websites became digital media, marketing platforms, business systems, automation, and eventually artificial intelligence.
The machines kept getting better. So did I. At least, that’s how the relationship seemed to work.
I’d learn something, then find a tool that let me do more with what I’d learned. The tool would expose me to something I didn’t know, so I’d learn that too.
Back and forth.
Person expands machine. Machine expands person.
That’s still what excites me about technology. I have no desire to return to some supposedly purer age before computers, automation, or AI. I’ve spent too much of my life making things with machines to suddenly pretend the machines are the problem.
Some of the most important things I’ve learned to make were possible precisely because technology gave me access to capabilities I didn’t have on my own.
But here’s where things have changed.
The machines aren’t simply becoming faster versions of the tools I grew up with anymore. Increasingly, they can perform larger portions of the work themselves.
They can write. Code. Design. Research. Analyze. Organize. Sometimes they can produce in seconds what once required years of accumulated technical ability.
Which is incredible.
And a little strange.
Because when I first sat in front of Photoshop, the machine expanded what I could do while I was learning how to do it.
Now a machine can sometimes produce the result before I’ve developed the capability that result once would’ve required.
That’s a different relationship.
And it’s no longer just something happening between me and the tools I use.
If you work with AI, automation, or increasingly intelligent software, you’re entering the same relationship. You can now reach capabilities you haven’t fully developed yourself. You can produce work that exceeds your current technical ability. You can hand over parts of thinking, making, deciding, and executing that previous generations had to learn how to do themselves.
That’s extraordinary leverage.
But it also creates a problem that’s easy to miss because the output can look so good.
A more capable machine does not necessarily create a more capable person.
So the question I’m interested in isn’t whether we should use these tools. We should. I’m interested in what happens to us as they get better.
What should we happily hand over? What do we still need to understand for ourselves? When does a machine amplify our capability, and when does it quietly begin substituting for the development of it? And if execution becomes increasingly cheap, what becomes more valuable on the human side of the relationship?
Because somewhere underneath all of this is a distinction I think we’re going to need.
We need machines that can do more for us without gradually making us less capable of understanding what they’re doing.
By the end of this, I want to give you a way to recognize that difference, and a principle for deciding what belongs in the machine, what still needs to remain in you, and how the two can become more capable together.
Because the question isn’t only:
What can the machine do?
It’s also:
What is the machine making you capable of becoming?
When Better Tools Made You Better
For most of human history, the deal we’ve made with technology has been pretty simple.
You take something I struggle to do and make it easier.
A hammer gives my arm more force. A camera captures an image. A calculator handles arithmetic faster than I ever will. A computer lets me process, manipulate, and organize amounts of information that would be ridiculous to manage by hand.
In a way, civilization is one long history of us finding new ways to make that deal.
Writing allowed us to store memory outside the mind. Books allowed knowledge to survive the person who possessed it. Machines absorbed physical labor. Calculators took over tedious arithmetic. Search engines made it unnecessary to remember where every piece of information lived.
We handed things over, then used the freed capacity to go somewhere else.
Nobody looks at a calculator and thinks humanity really lost its way when we stopped doing long division on paper. Some capabilities are worth preserving. Others we’re perfectly happy to let the machine carry.
That’s one of the fundamental promises of technology:
Give the machine what the machine can do, and expand what the person can do because of it.
But for a long time, there was another part of this relationship we didn’t really have to think about.
The tool could extend your capability, but it generally couldn’t manufacture the appearance of capability on your behalf.
A better camera could help a photographer capture an image, but it couldn’t give them an eye. Photoshop could make extraordinary visual manipulation possible, but knowing where the buttons were didn’t give you taste. A word processor could make writing faster, but it couldn’t solve the unfortunate problem of having nothing worth saying.
The technology mattered. So did the person operating it.
You learned the craft. You learned the tool. The tool exposed you to new possibilities. Those possibilities pushed the craft further. Better tools arrived, and the cycle began again.
Better tools made capable people capable of more.
I think that relationship is still buried inside how we think about technology.
We want the better laptop. Better software. Better workflow. Better model. Better prompt. Better automation.
We upgrade the machinery around our work because, historically, that’s been one of the most reliable ways to increase what we can accomplish. And often, it still is.
I can do things today that would’ve required an entire team when I started working professionally. That’s not some philosophical trick. It’s leverage. Better technology has genuinely expanded the range of what one person can build.
But here’s where the old deal starts to get strange.
The machine can now do something most of our previous tools generally couldn’t.
It can produce the evidence of capability before the person has developed the capability behind it.
You can get the result first. The skill can come later.
Or it might never come at all.
And once that’s possible, a better result and a better operator stop meaning the same thing.
When the Output Became Better Than the Operator
Artificial intelligence pushes that separation further than most of the tools we’ve used before.
At first, it can look like another upgrade to the stack. Faster software. Better search. More automation.
But I think something bigger is happening underneath.
We’re entering an era where people can produce work whose sophistication exceeds the capability they personally possess.
Someone who isn’t an accomplished writer can generate competent prose. Someone with limited programming experience can produce working code. Someone without years of visual training can generate sophisticated imagery. You can summarize a field you barely understand, build a workflow you couldn’t construct manually, or produce an analysis using methods you’ve never learned.
That doesn’t make the work fake. It makes the leverage extraordinary.
One person can now reach into capabilities that once required years of specialization, expensive software, institutional infrastructure, or an entire team.
To me, this is one of the great promises of the Digital Renaissance: capabilities once locked inside organizations are moving into the hands of individuals.
Good. I want more of that.
But here’s where the same leverage creates a strange new condition.
For a long time, the quality of someone’s output told you something about the capability behind it.
A beautifully constructed piece of furniture suggested someone knew woodworking. Sophisticated software suggested someone knew how to program. A strong essay suggested someone had learned how to think and write well enough to produce it.
The output was evidence.
Not perfect evidence. But evidence.
Now that relationship is becoming less reliable.
There’s a 2025 field experiment from researchers at Harvard Business School and Stanford that helps show what I mean. They looked at how generative AI affected performance among workers with different levels of occupational proximity to a task: insiders, adjacent outsiders, and distant outsiders. They found that AI could substantially narrow expertise gaps in some situations, particularly for people who already possessed adjacent knowledge. But that advantage weakened as the knowledge distance increased and the tasks moved toward more demanding execution, a boundary the researchers described as the “GenAI wall effect.”
I find that more interesting than either easy conclusion.
AI doesn’t magically make everyone an expert. And using AI doesn’t make the expertise fake.
Something stranger is happening:
AI can compress parts of the distance to expertise without necessarily eliminating the distance.
What you already know still affects how far the machine can take you.
But here’s the problem: the better the machine becomes, the harder that remaining distance can be to see.
You ask for an answer. It sounds articulate. You ask for code. It runs. You ask for analysis. It looks convincing. You ask for an image. It looks beautiful.
Great.
Now comes the harder part:
Do you know if it’s actually good?
Not whether it looks good or sounds intelligent. Not whether it compiled, rendered, summarized, or arrived in a beautiful little box with bullet points and a confident conclusion.
I mean, do you understand enough of the territory to know what you’re looking at?
Can you see what’s missing? Recognize the edge case? Tell when the recommendation works perfectly for the average situation and fails spectacularly in yours?
Can you recognize when the system has reached the edge of what it knows?
Sometimes the machine really has carried you somewhere you couldn’t have reached alone. That’s the point.
But arriving somewhere is not the same as knowing where you are.
And that’s the real break from the old relationship.
The machine can now cross parts of the competence gap for us. What it can’t guarantee is that we’ve developed the judgment required on the other side.
Which leaves us with a question that would’ve sounded ridiculous with most of the tools I grew up using:
If I can get the result without developing the capability, do I still need the capability at all?
The Wrong Fight
Once machines can produce the result without requiring the same capability from the person, there are two easy conclusions you can reach.
The first is that human skill matters less.
Why spend years developing capabilities software can provide in seconds?
Why learn to write if AI can write? Why learn to code if you can describe what you want and have a model build it? Why memorize information when nearly all of it can be retrieved on demand?
Follow that logic far enough and you end up somewhere strange:
Maybe becoming capable is becoming inefficient.
The second conclusion goes in the opposite direction.
If machines are beginning to substitute for capabilities people once had to develop, then maybe the answer is to resist them. Protect the craft. Preserve the friction. Do things manually. Treat technological assistance as somehow less authentic than producing the work yourself.
You can already see versions of this argument forming around AI.
Real writers don’t use it. Real artists don’t need it. Real programmers write their own code. Real thinking happens without assistance.
But I think both positions make the same mistake:
They assume the machine and the person are competing for the same capability.
They aren’t.
We’ve always built ourselves partly through tools. The camera changed how artists learned to see. Recording technology changed how musicians made music. Search engines changed how we find information. Software changed how designers work. Simulators let people practice situations too expensive, dangerous, or impractical to repeatedly experience in the real world.
The tool doesn’t have to sit outside the learning process. Sometimes it can become part of the learning process.
AI can work that way too.
There’s a 2026 randomized experiment involving more than 6,000 middle-school students that gives us a useful example. Researchers tested AI-supported mathematics practice inside different learning structures. AI helped students recover more effectively from mistakes, while the most encouraging delayed-learning results appeared when AI was embedded inside a mastery-based workflow.
The findings were modest, and I think that’s important. The larger point isn’t that AI suddenly made everyone learn better. It’s that access to AI alone wasn’t the intervention. How the technology was structured around the learning process mattered.
What the person does with the machine matters.
Sometimes the machine is part of how you build the person.
But the opposite can happen too.
A systematic review of 74 studies across healthcare, aviation, human-computer interaction, military, and other settings found a recurring problem known as automation bias. Useful automated systems can improve performance while also creating new errors when people over-rely on their recommendations. In a subset of healthcare studies, erroneous automated advice increased the risk of an incorrect decision by 26 percent.
Think about how strange that is:
The system can get better while the person gets worse at questioning it.
And I think that’s the part the human-versus-machine debate misses.
Technology can teach or bypass learning. It can extend judgment or encourage us to stop exercising it. It can expose us to capabilities beyond our current level or conceal how little of the underlying territory we understand.
So the danger isn’t delegation.
I delegate things to machines every day. So do you. Civilization would become absurdly inefficient if every generation insisted on personally retaining every capability technology had made unnecessary.
I don’t need to prove my humanity by beating a calculator at arithmetic.
The danger is cognitive abdication.
That’s different.
Here’s what I mean.
Cognitive abdication happens when we hand over enough of a process that we gradually lose the ability to understand what’s happening inside it.
We can no longer evaluate the answer. We can’t recognize the edge case. We don’t know where to intervene. We stop learning from the process because we no longer participate meaningfully in it.
Eventually, we may not possess enough understanding to recognize when the machine itself has reached the edge of its competence.
And yet the consequences still have to belong to someone.
If an AI recommends a business decision, someone has to decide whether to act on it. If it generates code, someone has to decide whether that code is safe to deploy. If it produces an argument, someone has to decide whether the argument is true. If it designs a system, someone has to decide whether that system should exist in the first place.
The machine may perform most of the execution.
Responsibility does not disappear simply because execution moved somewhere else.
So I don’t think the useful question is whether human beings should surrender capabilities to machines. Of course we should. We always have.
The harder question is deciding which capabilities we can afford to surrender.
If the machine is going to do more of the work, we need to become more deliberate about what remains on our side of the relationship.
Not everything needs to stay. But something does.
And this is the question I keep coming back to:
What capability do I need to retain in order to remain an intelligent operator of what I’ve automated?
Delegate Execution. Retain Judgment.
If the problem isn’t automation itself, then the answer can’t simply be to use less of it.
That’s too easy.
The better question is how we build a relationship with technology that increases what we can do without quietly decreasing what we’re capable of understanding.
And I think that requires a different definition of technological sophistication.
The most sophisticated operator isn’t the person with the largest collection of AI tools, the most elaborate automation stack, or the highest percentage of work delegated to machines. But it isn’t the person proudly doing everything manually either.
Both can become forms of technological immaturity.
The better operator understands which capabilities belong in the machine and which still need to exist in the person.
I think of the principle simply:
Delegate execution. Retain judgment.
Now, retaining judgment doesn’t mean knowing how to manually perform every operation inside a system.
I don’t need to understand every electrical process inside my laptop to write on it. I don’t need to perform long division every time I use a calculator. And I certainly don’t need to rebuild every piece of software I depend on before trusting myself to use it.
That’s not capability. That’s a terrible use of a Tuesday.
The point isn’t to preserve capability for capability’s sake. It’s to preserve enough understanding to direct the system, evaluate what comes back, recognize when something is wrong, intervene when necessary, learn from what happens, and remain responsible for the result.
There’s a distinction I’ve found increasingly useful here.
A technology can function as a Capability Amplifier or a Capability Substitute.
A Capability Amplifier allows you to accomplish more while preserving or strengthening your ability to understand, evaluate, intervene, learn, and take responsibility.
The machine becomes more capable. And because of your relationship with it, so do you.
A Capability Substitute can produce an equally impressive result. Sometimes a better one.
The difference is what happens to you as the operator.
The performance of the system begins replacing your own development.
You get better results while becoming less capable of judging them.
And for a while, that arrangement can look fantastic.
Everything works. The output is good. Productivity goes up. You start wondering why anyone ever did this the old way.
Then the model hallucinates. The automation encounters an edge case. The market changes. The recommendation conflicts with something you’ve learned from experience.
Or the system gives you three perfectly plausible options, and there is no prompt you can write that will relieve you of the responsibility of choosing among them.
Suddenly, your capability matters again.
This is why I don’t think every capability deserves the same level of protection.
If a task is routine, reversible, and low-consequence, there’s little virtue in preserving unnecessary friction. Automate it. Delegate it. Let the machine carry as much of the load as it reliably can.
But as the work becomes more ambiguous, consequential, strategic, ethical, irreversible, or identity-defining, the value of developed human judgment rises.
The question isn’t whether you can ask the machine to make the decision. You probably can.
The question is whether you possess enough judgment to know when you shouldn’t accept its answer.
And judgment is different from information. You can’t simply download it.
Judgment develops through contact with reality.
You try something. Reality responds.
You discover that the brilliant idea sounded considerably more brilliant before reality got involved.
You make the wrong decision and experience the consequence. You notice something you couldn’t have noticed before seeing the pattern five times. You adjust. Then you do it again.
Participation creates feedback. Feedback develops capability. Capability becomes judgment.
That’s why I think experience may become more valuable, not less, in an environment where information is abundant.
A machine can give you access to enormous amounts of accumulated knowledge. It can help you interpret that knowledge and expose you to patterns you might otherwise miss.
But you still need enough contact with the territory to know when the map is wrong.
I’ve seen this in my own relationship with technology.
Photoshop became more useful as I became more visually capable. The software gave me possibilities, but years of drawing, designing, making things, getting things wrong, and developing taste changed what I could ask it to do.
More importantly, it changed what I could recognize when it gave something back.
AI makes this relationship more powerful. But I don’t think it makes the underlying principle disappear. If anything, it makes the principle more important.
The more capable the technology becomes, the easier it is to become obsessed with learning the machine.
The models. The prompts. The workflows. The agents. The automations.
Learn all of it.
But that’s only half the work.
The New Operator has to know what the machine should do, what the human still needs to know, where supervision belongs, and who remains responsible when the two work together.
That’s the real architecture of leverage.
Not maximum automation. Not maximum human control.
Maximum leverage without surrendering authorship or responsibility.
Build Them Together
So maybe the title of this essay is slightly wrong.
Build the person before the machine makes it sound like there’s a clean sequence.
First, develop yourself. Become sufficiently capable. Then, once you’ve done the appropriate amount of human homework, you’re allowed to touch the powerful technology.
That’s not how this works.
There is no finished person waiting on the other side of some developmental threshold. We learn through the environments we enter, the problems we encounter, the people we work beside, and the tools we learn to use.
And the machine can be part of that development.
It can help you learn. Expose you to ideas you wouldn’t have encountered. Show you possibilities beyond your current skill. Remove mechanical work that no longer deserves your attention. Give you feedback. Help you experiment. Let you attempt things that would’ve been impossible at your current level of resources or experience.
So the more I think about it, the more accurate principle isn’t:
Build the person before the machine.
It’s:
Build the person while you build the machine.
Your capabilities allow you to make better use of the system. The system exposes you to possibilities that develop your capabilities. Those new capabilities allow you to direct more sophisticated systems.
Back and forth.
Person and machine develop together.
But here’s the important distinction: their responsibilities aren’t interchangeable.
A machine can execute an intention without deciding why that intention matters. It can generate possibilities without living with the consequences of choosing among them. It can assist judgment without becoming responsible for what that judgment produces.
That responsibility still has to land somewhere.
Which brings me back to something I’ve been thinking about throughout this essay.
Every time we ask, What can this machine do? I think there’s another question we should be asking alongside it:
What is this machine making me capable of becoming?
Is it helping me learn faster? See patterns I couldn’t see before? Ask better questions? Attempt more ambitious work? Develop better judgment?
Is it helping me spend less time on mechanical execution and more time deciding what deserves to exist? Or is it simply allowing me to produce more while understanding less?
That’s the uncomfortable part.
Because from the outside, those outcomes can look almost identical. Both can generate impressive work. Both can increase productivity. Both can make one person appear capable of what once required many.
You don’t really see the difference until the system stops knowing what to do.
And as machines become better at execution, I think that’s where the human role becomes more important, not less.
Judgment. Taste. Meaning. Intention. Curiosity. Discernment. Relationships. Responsibility. Lived experience.
These aren’t consolation prizes left behind after machines take all the interesting work. They’re increasingly the upstream capabilities that determine what the machines should be doing in the first place.
And that brings me back to Photoshop.
What made the computer meaningful wasn’t simply that it could manipulate an image faster than I could by hand. It expanded something already alive in me.
Then, by using it, I learned things I couldn’t have learned without it.
The machine expanded the person. The person learned to ask more of the machine.
Back and forth.
Decades later, the machines around me are almost unimaginably more capable. They can write, analyze, generate, organize, automate, and execute things the kid sitting in front of Photoshop couldn’t have imagined.
But I don’t think the standard has changed.
The technology should expand the person using it.
Build the machine. Use the machine. Learn from the machine. Let it extend you.
Just make sure that as the machine becomes more capable, you aren’t accidentally designing yourself out of the process that makes you capable too.
Because the future doesn’t need less capable people surrounded by increasingly capable systems. It needs people capable enough to know what those systems are for.
Build machines that multiply the person, not machines that replace the need to become one.
Garett
PS: Know someone who would benefit from this? Send them this link → subscribe.garettcampbellwilson.com
Want more insights on mastering the creator economy? Follow me on Instagram @gcamwil and stay updated on the latest strategies.
Start Here: The Digital Renaissance Manifesto
The system wasn’t built for creators. The traditional career path is collapsing, and the future belongs to those who create, not just those who comply. But how do you transition from being trapped in the old system to thriving in the new one?
That’s exactly what I break down in The Digital Renaissance Manifesto: your essential guide to understanding how creativity, technology, and ownership are merging to create the biggest wealth shift of our time.
Read The Digital Renaissance Manifesto – If you’re ready to stop trading time for money and start building leverage, this is where you begin.
Keep Learning: Related Reads
- YOU DON’T NEED A PERSONAL BRAND. UNTIL YOU NEED ONE.: How to package your knowledge, point of view, or process into digital assets that don’t expire when your shift ends.
- HOW TO TAP INTO THE WEALTH TRANSFER NO ONE TALKS ABOUT: There’s a silent wealth transfer happening. It’s happening in human attention.
- THE 9 TO 5 IS DEAD. NOW WHAT?: Why some are waking up to the fact that relying on a single employer for financial security is too risky.