A friend of mine bought his home decades ago.
There is a lot of equity in it now, built quietly over years of rising prices he did not engineer and did not fully anticipate. He wants to sell eventually, and he wants the price to be good.
He also has children. They are looking at the same market and finding themselves on the outside of it. The numbers do not work the way they worked for him. The entry price has moved to a place where the math requires either inherited help or an income that most jobs do not produce.
He holds both of these things honestly. High prices are good for him. Lower prices would be good for his children. He cannot fully want both. The system put him in that position. He did not design it. But he benefits from it, and he is not confused about that.
I have been thinking about that conversation in the context of AI.
Technology Is Deflationary
Every significant technology wave in history has done the same thing. The printing press collapsed the cost of producing text. Industrial machinery collapsed the cost of producing goods. Computers collapsed the cost of information processing. The internet collapsed the cost of distribution.
In each case, the technology made something that was previously expensive and scarce into something cheaper and more available. That is what technology does. It compresses cost. It is structurally, inevitably, deflationary.
AI is doing this to cognitive labor. Writing, analysis, synthesis, legal drafting, customer service, code, research, translation. The cost of producing these things is collapsing. Not eventually. Now.
This should be good news. Cheaper cognitive labor means more people can access services that were previously priced out of reach. The small business owner who could not afford a marketing consultant. The student who could not afford a tutor. The community organization that could not afford a grant writer. In theory, the deflationary benefit of AI is democratic. It reaches everyone.
I believe that. I have seen it work. I use these tools every day, and the access they have opened is real. A person of any background, any income level, any geography can interact with models that a year ago would have required institutional access. That openness is genuine and it matters.
But perceived openness and actual openness are not the same thing. And I have been sitting with the gap between them.
The System Cannot Allow Deflation
A debt-based economy requires growth. Not optional growth, not preferred growth, but structural, continuous, nominal growth, because debt is fixed and the economy must expand fast enough to service it. Deflation is the one condition a debt-based system cannot absorb. When prices fall, the real burden of debt increases. Asset values shrink. The balance sheets that the entire financial system is built on begin to crack.
This is not a conspiracy. It is a mechanical constraint. Central banks exist largely to prevent deflation. The entire post-2008 monetary architecture was constructed on that premise. Interest rates were held near zero for a decade to keep asset prices rising. Quantitative easing poured money into financial markets specifically to prevent asset price deflation. The housing market is not a side effect of this system. It is one of the primary mechanisms through which wealth is stored, protected, and transmitted across generations.
My friend's equity is not a windfall. It is the output of a system designed to produce that result, because the alternative, falling house prices, would solve the affordability crisis for his children and simultaneously destroy the balance sheets of everyone who borrowed against that equity, which is most of the people who own anything.
Something has to break. He knows it. The system knows it. Nobody in a position of power wants to be the one holding the lever when it does.
Where AI Enters This
AI arrives into this environment as the most powerful deflationary force in a generation. The question of who captures the deflationary benefit is not a technical question. It is a structural one.
Richard Cantillon observed in the eighteenth century that when new money enters an economy, it benefits those who receive it first, before prices adjust to reflect its arrival. By the time the benefit reaches the people furthest from the source, prices have already risen to absorb it. The gain is real. The distribution is not equal.
The same principle applies to deflationary technology. The people closest to the new capability, the companies building the infrastructure, the organizations with the capital to deploy it at scale, the individuals who can afford the better models and the developer access, capture the productivity gain first. They reinvest it before the benefit reaches the wider market. By the time the deflationary effect of AI shows up as lower prices for everyone else, the early capturers have already moved to the next layer of advantage.
This is what I mean when I say there are layers beneath the perceived openness.
Yes, anyone can open a browser and interact with an AI model. That access is real. But the gap between free-tier access and API access, between consumer products and enterprise deployment, between using a tool and shaping how it is built, that gap is not closing. It is widening. The concentration of computational resources, training data, and the talent to work with both is accelerating, not dispersing.
The student who cannot afford a university whose graduates end up at the companies building these systems will interact with AI. But the terms of that interaction, what the model knows, what it does not say, whose language it centers, whose history it reflects, are being decided by people in rooms she will never enter.
The Humanities Students Were Not Wrong
There were students at a graduation ceremony who booed a speaker talking about AI. People read that as resistance, as fear, as the reflex of disciplines under threat.
I read it differently.
The students who spend four years in philosophy, history, classical studies, and literature are trained in a specific capacity: the ability to notice when a system is serving one interest while claiming to serve another. They are trained to read power, to trace the distance between stated purpose and actual function, to ask whose story gets told and whose gets left out.
That capacity is not hostile to AI. It is exactly what AI most needs in the room.
An educator I spoke with recently started our conversation with indigenous students lacking access to traditional teaching methods that include elder knowledge. It moved into Black Wall Street. The systemic erasure of economic and cultural infrastructure that threatens the established hierarchy. He was not changing the subject. He was naming the same subject from different angles.
AI will not automatically repeat that pattern. But it will not automatically avoid it either. The training data reflects the documented world. The documented world reflects who had the resources and the institutional backing to document. The gaps are not random.
What Genuine Openness Would Look Like
I want to be clear about what I am not saying.
I am not saying AI is a tool of oppression. I use it every day. I have built a practice around it. The access it has opened is real, and for people who bring genuine capability to the tools, the amplification is genuine.
I am not saying the deflationary benefit will not reach people who need it. It will reach some of them. Probably many of them. The question is how much of the benefit will have already been captured by the time it arrives.
What I am saying is that the people best equipped to notice the difference between genuine openness and managed openness are the ones currently being told to stay out of the room. The humanities students, the critical theorists, the educators who have spent careers watching institutional power decide what counts as knowledge. These are not AI's opponents. They are its necessary critics.
And genuine openness would require something the current system is not designed to produce: a deliberate redistribution of the deflationary benefit before it concentrates. Not as charity. As design.
Back to My Friend
His equity is real. His children's problem is real. Both things are true and they are in genuine tension and the system that produced both is not going to resolve that tension on its own.
AI could change the terms of that equation. If the deflationary benefit reaches wages before it reaches asset prices, if the productivity gain shows up as time and income for people who do not yet own assets before it shows up as higher returns for people who do, the math changes.
That is not what deflationary technology has historically done. But it is what it could do if the people with the capacity to insist on it are engaged rather than dismissed.