The Capability Conundrum: Are We Ready to Slip the Leash?
- Rich Washburn
- 9 hours ago
- 12 min read
Something strange happens when capability stops being scarce.
The question changes.
For most of human history, the interesting things you might want to do had a fairly substantial wall sitting between the idea and the execution.
Want to build software? You needed a programmer. Want to model a business? Find somebody who understands finance. Need a contract analyzed? Call a lawyer. Need research? Hire researchers. Need engineering? Find engineers.
Design, marketing, statistics, translation, video, logistics, manufacturing, medicine — pick your problem and somewhere behind it was a person, usually several people, who had spent years acquiring some narrow slice of capability you didn't have.
That's not a criticism. It may be one of civilization's greatest inventions.
We learned to specialize. And then we learned to systemize the specialists.
The modern company is essentially a machine built out of human capabilities. Accounting does this. Engineering does that. Legal handles this corner. Marketing owns that one. Management sits somewhere above the whole mess trying to keep everybody pointed vaguely in the same direction.
The organization became the system. The individual became a component inside it.
That worked spectacularly well. It also trained us to think like components.
AI is about to make that a problem.
The really consequential thing happening right now isn't that AI can write an email or generate some code. That's the parlor trick. The bigger shift is that capability itself is becoming increasingly callable.
You don't have to become a programmer to call programming. You don't have to become a graphic designer to call design. You don't have to spend three weeks buried in research before you can begin interrogating a subject at a meaningful level.
The expertise hasn't disappeared. Neither has the need for actual experts, particularly when the stakes get high.
But the barrier between you and functional capability is collapsing.
Fast.
And once enough capabilities become callable, something moves. The premium shifts away from knowing how to execute every individual part and toward understanding how the parts fit together.
That's where this whole conversation about "systems thinking" suddenly stops sounding like management-school wallpaper and becomes incredibly important.
Because systems thinking may be the skill of the AI age.
And here's the good news: you already know how to do it. You may just need to remember.
THE TRICK IS ZOOMING OUT
"Systems thinking" sounds more complicated than it is.
Academics have been arguing over systems theory, cybernetics, complexity, emergence and various related flavors of this stuff for decades. There are entire libraries devoted to it.
Fine. Useful.
But you don't need any of that to start thinking this way.
The trick is zooming out. That's basically it.
Zoom out far enough that the thing you're looking at stops appearing isolated.
Look at a tree up close and you see bark, leaves, branches and roots. Pull back and the tree becomes part of something larger. Sunlight matters. Rain matters. Soil matters. Fungi matter. Insects matter. Animals matter. Wind matters. Temperature matters. Other trees matter.
The tree didn't change. Your frame did.
And suddenly you can see causes and constraints that were invisible when the tree filled the entire picture.
That's systems thinking.
Now zoom out from Earth. From down here, gravity is just a fact of life. You drop something, it falls. Pull your perspective back to the solar system and you start seeing relationships. Orbits. Mass. The sun. The geometry of everything moving around everything else. What seemed like a simple property from one perspective reveals itself as part of a much larger arrangement.
Again, nothing changed except where you were standing.
That's the skill. The ability to mentally change altitude. Zoom in when detail matters. Zoom out when context matters. Then move sideways and ask what else is connected to the thing you're looking at.
Most bad decisions aren't made because somebody lacked intelligence. They're made because somebody locked the camera in one position. They optimized the department and damaged the company. Fixed the immediate problem and created a larger one six months later. Cut the cost without understanding what that cost was protecting. Introduced the technology without thinking about how people would react to it. Changed the incentive and then acted surprised when behavior changed with it.
The spreadsheet was right. The system wasn't.
THEN YOU TIME TRAVEL
There's another part of this that we somehow don't talk about enough.
Systems thinking isn't only about changing scale. It's about changing time.
You play the movie forward.
We do this constantly in engineering and science. We build models. We simulate. We don't simply construct a complicated thing, turn it on and hope everybody has a nice afternoon.
We model airflow before an aircraft flies. We model structures before we load them. We model orbital paths before we throw hardware into space. We model weather, traffic, electrical grids, molecules, financial systems and countless other things because complex systems are expensive places to discover that you forgot something.
Simulation is just disciplined imagination.
Given what I know about the system, if I change this, what probably happens next? Then what? Then what happens because of that?
That's second- and third-order thinking, stripped of the consultant language.
You can do the same thing in your head. Humans do it naturally. You walk into a room and sense tension before anybody explains it. You can imagine what happens if you say the wrong thing to your spouse. You know that if you leave the house twenty minutes later, traffic might change the rest of your morning. You look at dark clouds and mentally rearrange the afternoon.
That's simulation.
We're prediction machines. On a leash.
We constantly take incomplete information, combine it with experience, run a crude model of the future and act based on the output.
Systems thinking is doing that intentionally.
The difference between first-order thinking and systems thinking is often just one more question: and then what happens?
You don't need perfect information. You won't get it anyway. You need enough information, enough imagination and enough humility to construct a reasonable model of what might follow.
Then you update the model as reality gives you better data.
That's the whole game. Not omniscience. Navigation.
WE DIDN'T LOSE THIS. WE OUTSOURCED IT.
Here's where things get interesting.
I don't think systems thinking is some new cognitive capability humans need to develop because AI showed up. I think it's ancient. Probably much older than anything we would recognize as civilization.
Imagine trying to survive ten thousand years ago without systems thinking. Weather, terrain, animals, food, water, season, shelter, threats, other people — none of those things could be understood independently.
A hunter didn't need a course in systems theory to know that the animal, the weather, the terrain and the season belonged to the same problem. A farmer couldn't divide rain, soil, insects, sunlight and crops into separate departments and hope someone scheduled a meeting.
The system was right there. Life demanded that you see it.
Then we got extraordinarily good at civilization. Complexity exploded. Nobody could know everything anymore, so we did something brilliant.
We distributed cognition.
Instead of requiring one person to understand the whole thing, we divided the whole thing into specialties. One person learned medicine. Then medicine itself became too large, so we divided that again. One person learned engineering. Then that split into electrical, mechanical, chemical, civil, software and a hundred other specialties.
Companies did the same thing. Accounting. Operations. Legal. Sales. Marketing. IT. Human resources.
Each person became incredibly capable within a narrower domain. And because the domains had to work together, we built layers of management, process, reporting structures and institutions to coordinate them.
We systemized the many because we had to.
That may have been the defining organizational trick of the industrial age. But it came with a cognitive side effect.
The more civilization carried the whole for us, the less often the individual had to.
Stay in your lane. Do your job. Trust the process. Ask the expert. Follow the procedure. Call the department. Turn on the television and somebody will explain what happened. Open the app and it'll tell you where to turn. Type the question and search will hand you the answer.
None of these things are inherently bad. They're incredibly useful.
But muscles don't particularly care why you stopped using them. They just weaken.
We outsourced more and more of the work of constructing the larger picture. And now, at exactly the moment when AI is handing enormous capability back to individuals, we're discovering that the part we may need most is the part we've been exercising least.
THE INDIVIDUAL IS BECOMING THE SYSTEM AGAIN
This is where the rubber hits the road.
AI isn't turning everybody into a lawyer, programmer, engineer, accountant, scientist and designer. That would be the wrong way to understand what's happening.
It's making those capabilities more available to the person who can orchestrate them. That's a very different thing.
You don't have to contain every skill. You have to understand enough to assemble the skills into a functioning system.
Think about what that means.
We spent generations building organizations because complex execution required multiple people. The capabilities had to live somewhere, so they lived in specialists. The specialists had to coordinate, so they lived inside institutions.
Now some portion of that architecture can begin collapsing back toward the individual.
One person with a decent mind and the right tools can move from research to architecture to code to financial modeling to design to analysis to implementation without stopping every twenty feet to assemble another department.
Again, that person isn't magically an expert in everything. That's not the point. The expertise can remain distributed. The orchestration doesn't have to be.
That's the shift.
We systemized the many because we had to. Now the tools may let us systemize the one.
And if that sounds like a subtle distinction, it isn't. It changes what an individual is capable of being.
The old labor model rewarded people for becoming valuable inside someone else's system. Learn a specialty. Get good at the function. Become a better component.
There's still going to be plenty of that. But another class of value is rapidly appearing above it: people who can construct the system itself.
People who can look at a messy objective, figure out what capabilities are required, call those capabilities, connect them, evaluate their output, identify what is missing and keep moving.
That doesn't require knowing everything. It requires seeing the whole.
Which means the skill that used to distinguish architects, founders, exceptional managers and a certain breed of obsessive generalist suddenly becomes useful to almost everyone.
Systems thinking isn't the executive suite anymore. It's literacy.
WE'VE SEEN THIS KIND OF MIND BEFORE
You don't have to go digging through prehistory to find examples.
Go back to the founding of the United States. Look at the cognitive range of the people involved.
They weren't simply politicians in the modern sense. Benjamin Franklin bounced around printing, science, invention, diplomacy, civic organization and political philosophy. Thomas Jefferson moved through law, architecture, agriculture, government, philosophy and science. Hamilton was thinking simultaneously about finance, governance, military affairs, trade and institutional design.
They weren't all-knowing superhumans. They were flawed people working with far less information than we carry around in our pockets.
But they were expected to range. They moved between domains because the problems in front of them demanded it.
They weren't just writing rules for a government. They were trying to imagine a system.
What happens if power concentrates here? What happens if these branches collide? What happens when ambition meets ambition? What survives one generation? What happens fifty years from now when none of us are here?
That's simulation. That's second- and third-order thinking. That's changing altitude and playing the timeline forward.
And, for all their human flaws — those same damn ones you and I also have — that kind of systems thinking produced an architecture durable enough that we're still living inside it almost two and a half centuries later. This is us, right?
I wrote before about cognitive sovereignty — about recovering some of the intellectual independence and range that people like this represented.
This is the practical companion to that idea. Because sovereignty without capability isn't much sovereignty at all. And capability without judgment can become a weapon pointed in any direction.
THE CONUNDRUM
That's where I keep landing.
We're obsessed with how capable AI is becoming. Reasoning scores. Coding benchmarks. Context windows. Agents. Robots. Models generating video, designing molecules, writing software and operating computers.
Fine. I'm obsessed with that stuff too.
But there's another capability curve we ought to be watching.
Ours.
Because we're removing friction from human intention at a rate civilization has never experienced.
The gap between "I want to do this" and "I can meaningfully attempt this" is collapsing across entire categories of work.
That's liberating. It's also terrifying. Because friction was doing more than slowing us down. Sometimes it was protecting us from our own stupidity.
When action is expensive, you tend to think before acting. When building something requires capital, specialists, organizational approval and months of work, there are lots of opportunities for somebody to notice that the idea is terrible.
When a capable individual can move from idea to implementation in an afternoon, the blast radius of bad judgment increases right along with the leverage of good judgment.
That's the capability conundrum.
The machine can increasingly help with the how. It can even help explore the consequences. But somebody still has to decide what the system is for.
Somebody has to decide which variables matter. Somebody has to notice when the optimization target is stupid. Somebody has to look beyond the immediate output and ask what happens downstream. Somebody has to recognize that technically possible and actually wise are two very different categories.
This is where systems thinking meets judgment. And I think those may become the two foundational human skills of this era.
Not because nothing else matters. Because more and more of everything else can be called.
JUDGMENT IS THE GOVERNOR
Systems thinking by itself isn't automatically virtuous.
A casino is a system. So is a surveillance state. So is a scam. So is an addiction algorithm.
Understanding human behavior, incentives, feedback loops and leverage points can be used to free people or manipulate them.
A brilliant systems thinker with terrible judgment isn't somebody I necessarily want equipped with unlimited machine capability.
That's why judgment is the governor.
Systems thinking asks: what happens if I do this?
Judgment asks: should I?
One maps the terrain. The other chooses the destination.
AI makes the distinction more important because the machine is increasingly capable of traversing terrain once you point it somewhere.
We've spent enormous energy worrying about AI alignment. Fair enough. We might want to spend a little more time worrying about human alignment too.
What happens when millions of people gain access to capabilities once reserved for institutions, governments, universities and corporations?
What do we do with it? What do we build? What do we optimize? What do we value? What do we refuse to optimize at all?
Those aren't software questions. They're human ones. And there isn't going to be a model release that answers them for us.
SO PRACTICE THE TRICK
This doesn't need to remain philosophical. You can start using it today.
Next time you're looking at a problem, mentally grab the camera. Pull back.
What system is this actually part of? What looks unrelated from here but becomes relevant from farther away? Who else is affected? What are their incentives? What constraint am I treating as permanent that might actually be changeable? What am I optimizing? What happens if I'm successful?
That last question is wildly underrated. People spend so much time thinking about failure that they never simulate success.
Okay, your plan worked. Now what?
You doubled the customers. Can operations handle them?
You automated the job. What happens to the people who used to supply the judgment hiding inside that job?
You made something cheaper. Did you also destroy the thing that made it valuable?
You increased engagement. Congratulations. What behavior did you just incentivize?
Move the clock forward. A week. A year. Ten years if the problem deserves it.
Then zoom back in. Now what detail matters that you couldn't see from altitude?
Do that enough and systems thinking stops being something you read about. It becomes how you look at things.
That's the part I want people to understand.
There's no priesthood here. You don't need a systems-science degree. You don't need to memorize diagrams. You don't need to start using words like "emergence" at cocktail parties and make everybody regret inviting you.
Change scale. Find the relationships. Run the simulation. Update when you're wrong.
That's the tool. You've had it the entire time.
ARE WE READY TO SLIP THE LEASH?
I've been thinking a lot about what happens as automation loosens the old relationship between human labor and human capability.
The leash is getting slack.
Not gone. Physics still wins. Capital still matters. Time matters. Real expertise matters.
Reality has a wonderful habit of punching people who confuse a chatbot with omniscience.
But something has undeniably moved.
An individual can reach farther than an individual could reach before. Soon, much farther.
And the greatest mistake we could make right now is treating that as merely a productivity story.
It's an agency story.
We're taking capabilities that civilization spent centuries embedding in organizations and making portions of them available directly to individuals. That's an astonishing transfer of power.
But the cognitive habits required to wield that power don't automatically arrive with the software. We have to reclaim them.
The ability to pull back. To understand relationships. To hold competing ideas in the mind. To imagine consequences. To move across disciplines without becoming trapped by any one of them. To simulate the future before charging into it.
And then, after all of that, to exercise judgment.
Maybe that's what this moment is really asking of us.
Not that we become more like the machines. Quite the opposite.
The machines are becoming extraordinarily capable at the parts. It's time for us to get better at seeing the whole.
Because eventually the hardest question may no longer be what we're capable of doing.
It may be what happens when capability is no longer the leash.
And before we celebrate slipping it completely, we should probably make damn sure we remember how to walk without one.
Rich Washburn is a technologist and strategist working at the intersection of AI, infrastructure, and capital. He is Managing Partner and Chief AI Officer at Eliakim Capital.