← Writing
NOV 21, 2026 ARGUMENT REFLECTION

Universal Access to Abundance


There is a phrase going around that I want to take seriously rather than cheer for: universal access to abundance, delivered as a by-product of AI and robotics. I believe a strong version of half of it. The other half is where the work lives, and calling it a by-product is how the work gets skipped.

Start with the half that holds.

Abundance is a cost-collapse story, and we have seen it before

Things become abundant when the binding constraint on making them collapses, not when anyone decides to be generous. The history of the last two centuries is mostly a list of these collapses.

Light is the cleanest example. The economist William Nordhaus traced what an hour of light actually cost a human being across history, measured in hours of labor. For most of that history, light was expensive enough to shape behavior: you went to bed because staying up cost oil you could not spare. Today the labor price of an evening of light is so close to zero that almost nobody thinks about it. Almost: as of 2024, about 650 million people still lived without electricity, and they think about light constantly. Hold that exception; it is the second half of this essay. For everyone else, light did not become abundant because anyone distributed it. The cost collapsed, and abundance was the residue.

Aluminum was once a precious metal; the story goes that Napoleon III served honored guests with aluminum cutlery while lesser company got gold. The Washington Monument was capped with it in 1884 because it was exotic. Two years later the Hall-Héroult process turned it into something we wrap sandwiches in. Food went from around 43 percent of an American household’s spending in 1901 to around 13 percent in 2024. Computation went from institutional to disposable inside one human lifetime. A phone call across an ocean went from an event you scheduled to a thing teenagers do absentmindedly, for free, with video.

Each time, the shape was the same: a good whose price was dominated by one scarce input became cheap when that input was industrialized, and then it disappeared from our list of things worth worrying about. Abundance, historically, is what it looks like when a constraint dies.

What AI and robotics actually collapse

Seen through that lens, the current moment is legible, with one distinction I need before the sweep of it. What AI is collapsing is the cost of raw cognition: the drafting, translating, summarizing, first-pass diagnosis, and tutoring work whose price was always mostly the price of a scarce person’s time. Raw, because unverified output is not the same product as expertise someone stands behind. That gap is real, and the second half of this essay lives inside it. Robotics is beginning to collapse the cost of dexterity: the picking, placing, welding, and harvesting, and eventually the lifting and bathing of elder care, work whose price was always mostly the price of a present person’s body. I say beginning deliberately. The cognition collapse has receipts; the dexterity collapse is still mostly pilots and demos, and the hands have lagged the mind in robotics for forty years.

Those two inputs, skilled attention and skilled hands, are the dominant cost in almost everything that is still expensive. Health care, education, legal help, construction, elder care, food production. The reason a will costs eight hundred dollars and a tutor costs eighty an hour was never materials. It was that the only machine that could do the work was a human being with years of training, and those are genuinely scarce.

So here is the half of the claim I buy: the two most expensive inputs in the economy have started down the same cost curve that light and aluminum and computation already walked. Unevenly, further along for words than for hands, and I hold the forecast loosely, because everyone should. But the argument that follows does not need the strong version. Even a partial collapse makes the access question live, because even dramatically cheaper production guarantees nobody anything.

The half that does not follow

Here is where I get off the train, and it is the word “by-product.”

Notice what the framing does, whoever uses it. If access arrives automatically, then nobody owes the access work. A by-product has no owner. A deliverable does. You do not need a theory of anyone’s motives to see the effect: calling access a by-product moves the job from the people with the means to the category of things that happen by themselves.

Abundance is a by-product. Universal access never has been. Access has been built, deliberately, by institutions, usually decades after the cost collapsed.

I should also say what I mean by universal access, because the phrase does three different jobs in most conversations. I mean something concrete. Everyone can get the essential services: health, legal, educational, infrastructural. At a quality someone accountable stands behind. At a cost that does not gate them out. With recourse when it fails. Affordable tools are the easy layer. Reliable, accountable essential services are the real test. And a claim on the output that does not depend on selling your labor is the hard part, for a reason that arrives shortly.

Electricity is the honest case study. By the early twentieth century, electric power was cheap and transformative, and city dwellers had it. In 1935, roughly one American farm in ten did. The cost had collapsed; the access had not, because running wire to a farmhouse was a cost nobody with a profit motive wanted to carry. It took the Rural Electrification Administration (REA), an act of deliberate institutional will, to finish the job, and it still took another two decades after that. The technology made universal access possible. It did not make it happen. A policy made it happen.

The same shape repeats everywhere you look. Vaccines reached the rich world years before they reached everyone else, and the gap was never mainly manufacturing cost. Copying information costs almost nothing, but delivering it is wire, towers, and spectrum, and the ITU still counted roughly two billion people offline in 2026. Cheapness creates the possibility frontier. Who lives inside the frontier is decided by ownership, infrastructure, and rules, which is to say by people.

And there is a specific reason to think this time is harder, not easier. Light and aluminum were commodities: once cheap, anyone could buy them from anyone. The new abundance arrives through concentrated channels. Frontier models are made by a handful of firms. Robot fleets will be owned before they are ubiquitous. When the thing that collapsed in cost is accessed through a platform rather than bought in a sack, the gap between “cheap to produce” and “available to you” is a business decision renewed every quarter.

And there is a deeper break with the historical pattern, the one that makes me distrust every easy analogy in this essay, including mine. Light, aluminum, food: those were goods. When their cost collapsed, ordinary people got richer, because the wage in their pocket suddenly bought more. The floor rose through the wage channel. This time, the thing collapsing is the thing most people sell. Cognition and dexterity are not just the expensive inputs in the economy; they are the income source of nearly everyone in it. If their price falls toward zero, the same collapse that makes abundance possible erodes the purchasing power that let every previous floor rise. Access then depends on ownership and transfer mechanisms, on who holds the machines and what claim everyone else has on their output. Measured against that, rural electrification was a rounding error. The REA ran wire to people who could pay for power once the wire arrived. This time the harder question is what “being able to pay” even rests on. The last mile is not the problem; the road is.

The gates that remain when cost is gone

If the cost of expertise and labor stops being the constraint, it is worth naming what still gates access, because these are the actual work:

Energy and atoms. Intelligence may approach free; the electricity it runs on and the materials robots are built from will not. The cost floor moves from wages to watts. Regions with cheap energy get cheap abundance first. This is a solvable, buildable problem, and it is infrastructure, not by-product.

Ownership and pricing. A cost collapse captured by its owners shows up as margin, not access. Historically, competition eventually forced collapses through to prices, and open alternatives have a way of chasing closed ones. But “eventually” did a lot of quiet work in that sentence, and the people outside the gate live in the meantime.

Licensure and liability. In my industry, legal, the scarce credentialed human is not just a cost input. They are the accountability mechanism. We price legal advice partly by the malpractice insurance behind it. Cheap expertise without an accountability structure is not access to abundance; it is access to plausible-sounding output that nobody stands behind. I have written before about what happens when the person on the hook can no longer interrogate the thing they are signing off on. Until we build accountability structures that work at the new cost floor, regulators will rightly keep the old gates up, and the abundance will pool behind them.

Trust and verification. When expertise was expensive, its price was a crude quality filter. When it is nearly free, telling good from confident-and-wrong becomes the scarce skill. Verification is the new bottleneck, and bottlenecks are where cost lives.

Notice what that list is. It is not a list of technical problems waiting on a better model. It is a list of engineering, institutional, and governance problems: grids, markets, liability frameworks, verification systems. Every one of them is work someone has to choose to do.

The part worth being optimistic about

I want to be precise about where my optimism sits, because I am genuinely optimistic.

The floor can rise even when the ceiling runs away. It did in every collapse so far. The light in a poor household and the light in a rich one are now the same light. The phone in a day laborer’s pocket and the phone in a banker’s pocket are, to a first approximation, the same phone. Inequality of wealth persisted; inequality of lumens did not. But I wrote can, not does, because this is exactly where the wage-channel problem bites: the lumens reached everyone because everyone still had something to trade for them. And a second caution, because it is easy to celebrate the wrong floor: what gets cheap first is tasks, not outcomes. A drafted document is not accountable legal help, and accountable legal help is not a resolved dispute. The floor that matters is the outcome floor, and it rises last.

I owe the counterexamples a hearing, because they are sitting in my own list. Health care, education, and housing got relatively more expensive while everything around them got cheap. Economists call the service half of that Baumol’s cost disease: anything made of human time resists the productivity gains that make everything else cheap, so it eats a growing share of every budget. AI and robotics attack cost disease at its root, which is the genuinely new thing about this moment and the strongest reason for optimism. But housing teaches the other lesson: collapse every input you like, and the surplus pools into whatever stays scarce, land above all. The same phone in every pocket did not put the pockets in the same zip code. The floor rises in everything that can be manufactured. The fight moves to everything that cannot.

But every previous collapse needed its Rural Electrification Administration, its public health system, its universal service obligation: the deliberate, unglamorous institution that carried the last mile after the market carried the first ninety percent. Those institutions were built by people who treated universal access as a requirement, not an externality. Engineers have a word for the difference between a property you designed for and a property you got lucky on, and we do not trust the lucky kind in systems that matter.

And this time the institution has a bigger job than wire. If the collapse reaches what labor can earn, universal access rests on something every previous collapse could take for granted: a durable claim on the output. Wages where human work remains valuable. Ownership spread wider than the firms that hold the machines. Public provision for the essentials that should not depend on income at all. Transfers where nothing else reaches. I do not know the right mix, and I distrust anyone who is certain they do. But a plan for abundance that does not say where ordinary people’s claim on it comes from has not answered the question. It has answered an easier one.

Universal access to abundance is available, nearer than it has ever been. But it is not a by-product. It is a deliverable, and deliverables have owners. For the firms capturing the collapse, that means building the access path with the same seriousness as the capability, because the by-product story will not hold once the first access fight reaches a regulator. For policy, it means doing what the REA did, at the scale of this collapse: building the infrastructure the market will not, and keeping ordinary people’s claim on the output intact while the wage channel shakes. And for the engineers in between, the people who decide what actually gets built next quarter, it means treating universal access as a requirement with a name next to it. That is the only place roadmap items have ever lived.