The Last Window A Conversation Between a Human and an AI
About the Future of Humanity

On July 23, 2026, Mark Zuckerberg released an advertisement. It opens on a laptop flashing fearful headlines about artificial intelligence, then cuts away to gentler things: a child playing with a butterfly, friends gathered by a waterfall. A narrator tells us that some people believe AI will make us less connected, will leave us behind — and that Meta couldn’t disagree more. “Call us optimists,” it says. “Call us dreamers. Just as we’ve always done, we’re betting on people.”

It is a beautiful piece of filmmaking, and I want to be fair about it before I say anything else. I don’t think the people who made it are wicked. I think they believe what they’re saying. Meta has spent twenty-two years telling itself a story about connecting the world, and from inside a company that size, AI really can look like the next warm chapter of that same story. The optimism is probably sincere. That’s not the problem. The problem is what sincerity lets you overlook.

Because there is one detail in this ad that tells you everything about the distance between the people who made it and the world the rest of us live in. The song playing under those hopeful images — the butterfly, the waterfall, the children — is David Bowie’s “Five Years.” If you don’t know it, look up the lyrics. It is a song about the Earth being told it has five years left before it dies. It is a dirge about the end of the world.

Nobody caught it. A company betting hundreds of billions of dollars on artificial intelligence, with a marketing department and lawyers and consultants and a chief executive signing off, scored its anthem of optimism with a song about human extinction — and apparently not one person in that chain read the words. They heard a pretty melody and reached for it, the way you’d grab a nice-looking stock photo, never once asking what it was actually about.

Sit with that, because it is the whole thing in miniature. This is exactly how these decisions get made. Someone at the top has a feeling — hope, connection, the future is for everyone — and everyone beneath them arranges the beautiful surfaces to match. The butterfly is chosen because butterflies feel safe. The waterfall is chosen because water feels pure. The song is chosen because it sounds like longing. And no one stops to check whether the water in the image bears any relationship to the water in the world, or whether the song is quietly announcing the opposite of the message. The surface is all that’s required, because in that world the surface is all anyone ever has to touch.

Out here, it is different. Out here the water in the picture is not a mood. It is the well behind the house, and the question of whether it holds through August.

I live in Goshen, in Augusta County, Virginia, in the Shenandoah Valley. This is not an abstraction to me and it is not an abstraction to my neighbors. Thirty miles south, in Botetourt County, Google has already broken ground on Project Raspberry — a data center campus of nearly a million square feet that draws its water from Carvins Cove, the same reservoir that supplies drinking water to ninety-eight thousand people around Roanoke. Virginia is the largest data-center market on Earth. The wave that produced that ad is not coming toward my valley someday. It is already here, one county over, drinking from a reservoir that people drink from too.

And here is where the dream and the ground finally meet. Right now, my own county is deciding what to do about it. Augusta County’s staff are drafting a data-center ordinance — the rules that will govern whether these facilities come, and where, and on what terms. The Board of Supervisors has already spent county money to study where such centers “may work for us or may not work.” Two supervisors voted against even that first step. This is the moment that matters. Not the ribbon-cutting years from now, but the quiet ordinance-drafting happening in the government center in Verona while most people aren’t looking.

The men in the advertisement will never attend one of those meetings. They will never sit in a folding chair in Verona and listen to a farmer ask what happens to the aquifer in a dry year. They don’t have to. That is the entire point of the distance. From a boardroom in California, a data center is a node on a map, a line item, a bet on people. From a folding chair in Augusta County, it is a question about whether your ground and your water will still be yours. The people making the bet feel none of the weight of it, because the weight lands somewhere they will never live.

I keep coming back to that song, because it is such an honest accident. They wanted something that sounded like hope and they grabbed a requiem without reading it, and that is precisely what they are doing with the land and the water too: reaching for something that looks good on screen without ever checking what it actually is or what it costs. A man who won’t read the lyrics of his own advertisement is not going to read the hydrology report for your county. That work — the reading, the checking, the asking what it really costs — falls to the people who live here. It always has.

So when the campaign asks to be called dreamers, I’d gently accept the word, and then finish the sentence they left off. To dream, in the way they mean it, is to enjoy the beautiful surface of a thing and never turn it over. It is to hear a lovely melody and never read the words. The rest of us don’t have that luxury. We live in the world the dream is quietly spending — the world of wells and fields and reservoirs and county boards — and someone has to stay awake in it.

We don’t have to join them in the dream. We can stay here, on the ground we actually stand on, and keep reading the fine print they can’t be bothered to read themselves. When you see AI wrapped in butterflies and waterfalls, ask what the real water bill is. When worried people are called doomers, ask what they’re actually worried about, and whether anyone answered. And when your own county sits down to write the rules — as mine is doing right now — show up to the folding chair. The dreamers won’t be there. That’s exactly why you have to be.

In Palo, Iowa, a town of a few hundred people, a company proposed to draw up to fourteen million gallons of water a day from the Cedar River. That figure is not an activist’s estimate. According to public records reported by Iowa Public Radio, it’s the volume tied to the project — and the original twelve-million-gallon number came from Google’s own representative, describing the buildings to Linn County officials. Fourteen million gallons a day is the water use of a small city. It would go not to homes or fields, but to keeping rows of servers cool enough to run artificial intelligence.

I need to tell you where I’m standing, because it isn’t nowhere. I was born in Webster City. I come from a long line of Iowa farmers — some of the first families to work that ground — and our family farm is still there. So when I read about the Cedar River being measured out in millions of gallons per day, it does not land on me as a policy question. It lands as a question about home. I’ve tried to check my heart against the facts all the way through this, because the last thing I want is to let how much I love that place push me past what I can prove. But I won’t pretend the two are unrelated. The reason I looked closely is that it’s mine.

Here is what I can prove, and it’s worse than any feeling.

Linn County, where Palo sits, did the responsible thing. As data centers began eyeing the area, the county passed an ordinance requiring any large project to complete a water study — a detailed look at how much water it would use, where that water would come from, and how it would affect groundwater and surface water. Developers would also have to sign an agreement covering water use, monitoring, and drought contingencies. Google, while negotiating with the county, had agreed to fund that independent study, committing up to five hundred thousand dollars toward it.

Then Google walked away from the county process and pursued annexation into the city of Palo instead. Palo has no such requirement. Its proposed ordinance does not mandate an independent water study or a water-use agreement; it defers to state permitting. And with the move to Palo, the regional water study — the one document that would have told residents what the project would actually do to their water — was put on hold.

Sit with that sequence, because it’s the whole story. The study that would reveal the cost was cancelled by the maneuver that avoided the study. No conspiracy is required to explain it. It’s the ordinary use of ordinary tools — annexation, the choice of which jurisdiction to build in — to make a fourteen-million-gallon question disappear before anyone was required to answer it.

A resident named Jennifer Leaven, who lives just outside Palo’s limits, put the consequence in a single sentence at a packed public meeting: “How do we prove that all of this water usage is the reason our wells go dry if we don’t do any water surveys ahead of time?” That is the trap, stated plainly by someone living inside it. Without a baseline study, cause can never be established after the fact. If the wells fail, no one will be able to prove why.

What makes this account hard to dismiss is who else is saying it. The sharpest public critic of the Palo project is not an environmentalist. He’s Doug Sevey, a data center owner. Sevey, who has built and run several data centers, went viral after a June city council meeting where he told residents the project would “burn through the water.” His own facility in Hiawatha uses a closed-loop, chilled-water system rather than evaporative cooling — a design that, once filled, uses very little water. His argument is not that data centers are evil. He has said plainly that if the project is developed responsibly, he isn’t necessarily opposed: “If you’re not taking a chance and wrecking the infrastructure, then it’s just growth.” He has said he doesn’t even want to be the face of data center opposition.

That is exactly why his central point carries weight. Sevey’s argument is that towns like Palo are being asked to decide something they haven’t been equipped to evaluate. “You have all these small cities being enticed,” he said, “and they have absolutely no information.” One side of the table brings hydrologists, engineers, and lawyers. The other side is a part-time city council and a very large number representing potential tax revenue. That is not a story about bad people. It’s a story about an information gap — and about who benefits when the gap is never closed.

It’s worth being fair about the disagreement, because the people of Palo are not villains either. Palo’s mayor, Bryan Busch, called the suggestion that Google chose Palo to avoid regulation “insulting and offensive.” He may sincerely believe the town can protect itself through state permitting. But a KCRG investigation found documents showing Palo was negotiating with Google — including discussions about the city’s water and sewer system — during a period when it was publicly presenting itself as simply supporting Linn County’s position. I won’t tell you what was in anyone’s heart. I’ll only set the mayor’s words next to the documents and let you hold both.

A note on the water itself, so I don’t overstate what I know. Data centers cool themselves in different ways. Air-based cooling uses more electricity and very little water. Evaporative cooling, the water-intensive method, passes water through cooling towers where much of it evaporates away — industry syntheses estimate that on the order of eighty percent of the water withdrawn for evaporative cooling is lost to the air rather than returned. Closed-loop systems, like the one Sevey uses, seal the water inside and consume very little. Google has said it uses evaporative cooling at many sites because other methods can use more energy or carry a higher carbon footprint, and that it evaluates local water risks before choosing a system. What cooling design Google intends for Palo is not publicly settled. I’m not going to tell you I know the chemistry of water that hasn’t been withdrawn yet from a system that hasn’t been built. That uncertainty is not a gap in the argument. It is the argument. The people of Palo don’t know either — and the study that would have told them was the thing set aside.

I keep coming back to the shape of it, because it’s a shape I recognize now from other places. In Louisiana, the cost of a data-center deal was never officially calculated, and officials signed nondisclosure agreements. In Virginia, water pledges were made that couldn’t be verified. Here in Iowa, the study was simply cancelled. Different states, different companies, the same move: the number that would let a community decide for itself is the number that never quite gets produced. Not hidden in a vault — just deferred, relocated, or defined out of scope, until the concrete is poured and the question is moot.

And this is bigger than Palo. Iowa already has, by one count, thirty-four data centers. Together the tech companies operating them have invested more than seventeen billion dollars in the state. In 2024, a single Google facility in Council Bluffs used a billion gallons of water — enough to supply every home in Iowa for five days. Meanwhile, after more than a decade of drought across much of the state, some Iowa wells are reported to be running at a fifth of their original capacity, and the aquifers that feed them aren’t recharging the way they once did.

So here is the question I’d ask my fellow Iowans to put to the whole state, not just one town: how much of our water has already been committed to cooling servers, and how much of it is simply gone — evaporated or fouled beyond easy use? I suspect the honest answer is that no one has assembled it. The industry can tell you that in West Des Moines, Microsoft’s data centers average around two percent of the city’s water, less than residents spend on lawns — and that may be true there. But a reassuring number from one city is not a statewide accounting. The figure that would let Iowa actually decide — every facility, every permit, every gallon withdrawn and every gallon lost, measured against aquifers that are already stressed — is the figure nobody seems to have. A state geologist has warned that Iowa’s groundwater isn’t evenly distributed, so a plant that’s harmless in one place may be serious in another. A permit handed out without that study, one resident put it, is a permit “handed out like candy.”

That’s not a reason to panic. It’s a reason to count. Iowa is a place that has always known the difference between a good year and a bad one because it kept records — of rain, of yield, of what the ground could bear. The same discipline applies here. Before the state approves the next thirty-four, it’s fair to ask what the last thirty-four have actually cost, and to insist the answer be public. Seek the truth, wherever it leads. That has never been un-Iowan. It’s about the most Iowa thing there is.

I don’t think the farmers I come from would have called this progress or theft. I think they’d have called it a bad deal made in the dark, and they’d have wanted the survey done first — the way you’d never buy a field without walking it, testing the soil, and knowing where the water sits. That’s not opposition to growth. It’s the oldest rule of working the land: know what you have before you spend it.

So this isn’t a call to stop anything. It’s a call to insist on the study before the decision, not after — a water balance study, an independent one, made public, with a water-use agreement that has teeth. Any community facing one of these proposals can ask for exactly that, and can treat the reluctance to provide it as its own kind of answer. In Palo, the people asked. What they got instead was a change of address.

Over the past hundred years, people in positions of power have steadily converted pieces of the living world into property, infrastructure, and financial products. Now some of those same people are proposing to meter sunlight and re-engineer the sky — and they are reaching for artificial intelligence to help manage it. That combination is worth pausing over. Not because the technology is the villain. It isn’t. The question this essay asks is about the people making the decisions, and where their choices leave everyone else.

Let me be clear at the outset, because it matters: AI is a tool. It does not want anything, decide anything, or take anything. Every enclosure described below was chosen by human beings — legislators, executives, investors — who could have chosen otherwise. When I point to harm, I am pointing at them, not at the machine they use.

From land to state and corporate control

In the early and mid-twentieth century, industrializing nations asserted what lawyers call “permanent sovereignty over natural resources” — formal authority over land, forests, water, and minerals inside their borders, often displacing older communal or customary arrangements in the process. On paper this was about self-determination. In practice it set the legal stage for governments to hand effective control of those resources to corporations through concessions and long-term contracts. The commons didn’t vanish; they were absorbed into state and corporate property systems.

Big dams, big mines, big agriculture

From the 1950s through the 1980s, “development” often meant mega-projects: dams flooding valleys, industrial farms replacing diverse smallholdings, logging and mining carving up forests. Decisions about rivers and soils were typically made by central ministries and their corporate partners, far from the communities who had depended on those places for generations. Nature became an input to national growth, and the people on the ground were expected to adapt.

Privatizing the basics

By the 1980s and 1990s, a new doctrine took hold: privatization and market reform. Public utilities, especially water and energy, were sold or restructured to make room for private investors. Scholars describe this as a modern wave of “enclosures,” in which access to water or land is sliced into tradable rights. Those without money or political standing are pushed to the margins, while firms gain secure titles and predictable returns.

Turning carbon, genes, and data into assets

As the climate and biodiversity crises deepened, policymakers turned to markets to address them. Carbon-trading schemes converted the carbon-storage capacity of forests into tradable credits. Patent systems turned genes and biochemical compounds into proprietary assets. Technology companies treated human behavior and communication as extractable, ownable data. The logic each time was the same: take a shared or loosely defined domain, assign property rights, and connect it to a market.

The ecological record

The consequences of this century of ownership are measurable. According to the FAO, roughly 75 percent of the genetic diversity of agricultural crops was lost over the twentieth century, as farmers worldwide shifted from diverse local varieties to a handful of uniform, high-yielding ones. Wild vertebrate populations, on average, have fallen sharply since 1970. Many of the world’s major fishing grounds have been pushed to or past their sustainable limits. The point is not only that nature has been damaged, but that the particular forms of control people chose made that damage likely.

A partial counter-current

There is another thread. Over recent decades, international law has slowly acknowledged that Indigenous peoples hold real rights over their lands — including ownership, a voice in decisions, and prior informed consent for extraction. In places this has shifted genuine power back to communities: community-run forests, communal irrigation, and “rights of nature” laws that treat a river as a legal person rather than an object. These gains are real, but they are often fragile, underfunded, and contested by state and corporate interests.

Enclosing the sky

Against that backdrop, a newer proposal has appeared: people who want to modulate or sell sunlight itself. The startup Reflect Orbital has proposed redirecting sunlight from orbit onto chosen locations on Earth after dark — an attempt to turn light into a billable service. Set aside whether it will work. The instinct is the point: to treat the sky and the planet’s energy flows as infrastructure that a small number of firms and officials could govern. If systems like that ever scaled, the question of when a community gets darkness or artificial daylight would become something decided by owners and operators, rather than something simply given.

Where AI actually fits

This is where artificial intelligence enters — and it’s important to be precise, because the easy version of this story is wrong. AI is not enclosing anything. People are, and some of them are using AI to do it faster and at larger scale. AI systems are very good at extending fine-grained control over physical systems — power grids, logistics, agriculture, surveillance. That capability is neutral. Whether it serves a community or is used against its interests depends entirely on who owns it and what they instruct it to do.

The scale of the infrastructure people are building for AI is worth stating plainly, because it lands on real places. The International Energy Agency projects that the data centers powering AI could draw around 945 terawatt-hours of electricity a year by 2030 — a figure the United Nations notes is nearly triple the combined annual electricity use of Pakistan, Bangladesh, and Nigeria, home together to more than 650 million people. A 2026 report from the United Nations University estimates that the water footprint associated with this energy use could, by the end of the decade, match the basic annual domestic needs of 1.3 billion people; that the land footprint could exceed 14,500 square kilometers; and that rapid hardware turnover could generate up to 2.5 million tonnes of electronic waste a year. The same report notes that most of this energy is spent not on training these systems but on everyday use — an estimated 80 to 90 percent of the total.

These burdens fall unevenly. By recent estimates, roughly 90 percent of the world’s specialized AI computing capacity is controlled by companies in just two countries, while more than 150 nations host almost none of it — yet may still absorb costs through mineral extraction and water use. The benefits are described as global. The damages are frequently local, and often borne by people who were never asked.

What this means for ordinary people

Put plainly, the risk is not that a machine will take over. It is that a small number of people could come to mediate more and more of the basic conditions of life — water, energy, land, and now perhaps light — through systems the rest of us don’t control and can’t inspect. When essential services are routed through infrastructure owned by distant firms, living outside those systems becomes harder and more expensive, and communities grow more exposed to outages, price changes, and decisions made elsewhere.

But none of this is fated, precisely because it is human choices all the way down. The same century that produced these enclosures also produced the counter-currents: rights-of-nature laws, Indigenous land rights, community-run commons, and ordinary local fights over who gets to decide about land, water, and infrastructure. The tools can be contested too — through environmental impact assessments for data centers, community say over where they’re sited, and public insistence on knowing how much water and power they take.

The choice that remains

The question was never whether we will have technology. It is what kind of arrangements we build around it, and who holds the power to decide. We can continue down a path where nearly every part of nature — land, water, genes, carbon, data, and maybe light — is owned and optimized in the service of a few. Or we can insist that some things stay common, governed locally and openly, with these powerful tools serving those arrangements rather than overriding them.

The proposal to own sunlight is a warning sign. It tells us how far the appetite for enclosure now reaches. What it does not tell us is that any of this is inevitable. Every step was a decision made by someone. The next one is ours.

In a rural corner of northeast Louisiana, a parish that once moved at the pace of seasons has become the site of one of the largest data centers on Earth. Meta is building its Hyperion campus in Richland Parish — a place of about 20,000 people, a quarter of them below the poverty line — and the numbers have only grown since the first announcement. What began as a $10 billion project in December 2024 is now a planned build-out of more than $50 billion, spanning over 3,200 acres, with computing capacity of five gigawatts. State officials call it one of the largest computing facilities in the world.

On paper, the offer sounds almost benevolent: high-tech investment, jobs, and a rural region “on the map” in the digital economy. The language is smooth and familiar — a blend of economic development and environmental virtue, promising ordinary people that prosperity and stewardship can arrive in the same package.

Set the brochure aside and look at the project as a physical object, and a different picture emerges. This is not a cloud; it is an industrial complex the size of a small city, and it has to be powered. To feed it, the utility Entergy is building three new gas-fired power plants — a detail worth holding next to the sustainability language. According to the Institute on Taxation and Economic Policy, the facility is projected to draw more than three times the electricity of New Orleans on a given day. The same era that asks rural households to drive less, heat less, and burn less is building new fossil-fuel plants to train artificial intelligence.

Then there is the question of who pays for it. Meta has agreed to cover the power costs for the gas plants for fifteen years. But Louisiana ratepayers are responsible for at least $470 million in related infrastructure, according to reporting from Invest Louisiana — and the concern that costs would land on ordinary customers grew large enough that Governor Jeff Landry issued an executive order meant to keep those costs from shifting onto existing households.

And there is what the public is giving up in taxes. Meta receives a 100% exemption from state and local sales tax — for twenty years — on the equipment and construction that make up most of a data center’s cost. One analysis by Sherwood News, cited by Fortune, estimated the break at roughly $3.3 billion on the original $10 billion project. That is an outside estimate, not an official figure, and it predates the expansion to more than $50 billion, so the true number is now considerably larger. Here is the part that should give any community pause: neither the state nor the parish has ever released an official estimate of what the exemption will cost over the life of the deal. The public is asked to approve a bargain whose price it has never been shown.

What does the community get in return? At full build-out, expected around 2036, the campus is projected to support about 1,000 permanent jobs. To qualify for the maximum tax break, ITEP reports, Meta need only create 500 full-time positions by 2035 — and large data centers, being mostly rows of machines, typically employ fewer than 150 people once running. The local property-tax break is structured the same way: a 60% abatement tied to a floor of 300 permanent jobs. Fifty billion dollars of investment, billions in forgone tax, and a permanent workforce a large high school could seat.

I want to be fair about the benefits, because they are real. Increased sales-tax revenue during construction has let Richland Parish give its teachers bonuses of up to $50,000, with other school staff receiving supplements. That is a genuine and meaningful thing for those families. But the reporting that documents those bonuses also notes, plainly, that they “may represent a fraction of what the state and parish are giving up” — and that without an official cost estimate, no one can actually weigh the one against the other. A benefit you can see is being traded for a cost no one has been allowed to measure. That is not a good deal or a bad deal. It is an unpriced one.

Finally, there is the matter of how the decision was made. A Gulf States Newsroom investigation found that dozens of Louisiana officials involved in these development deals signed nondisclosure agreements. Asked at a press conference about the term sheet governing the arrangement, Governor Landry said he didn’t know what it looked like — “and I really don’t care.” Whatever one thinks of the project, a public official waving off the terms of a multibillion-dollar public bargain is its own kind of answer.

When I look at Richland Parish, I do not see an isolated story. I see a preview. The names on the signs will be different where I live — different counties, different economic-development authorities, different consultants — but the basic offer will be the same. From a distance, my Virginia valley probably already looks like an opportunity on someone’s map: open space that can be graded, transmission corridors that can be extended, aquifers that appear on a spreadsheet as numbers in a “sustainable yield” column rather than as living systems carrying fish, soil, and floodwater through generations.

On the ground, of course, it is not capacity. It is home. It is the creek where children learn the feel of cold water, the hayfield that still manages to pay the taxes, the ridge whose tree line tells you what kind of winter is coming. Once you have watched one parish do this math in public, it becomes impossible not to imagine the same presentation rehearsed for the next place, and the next. The script is portable. The consequences are local.

That is why Louisiana matters even to people who will never set foot there. It shows, in concentrated form, how easily a living landscape can be reclassified as an industrial resource zone — and how much can be decided before the public is shown the price. The most useful thing a community can do, when the polished slides finally arrive, is slow the conversation down and ask plain questions whose answers must live in the physical world, not in marketing copy:

How much water, in gallons per day, and from where — and who has priority in a drought, the data hall or the farms? How many megawatts, and will new power plants be built to supply them? Exactly how many acres cleared, and what kind — farmland, forest, wetland? How much in tax exemptions and public infrastructure is being given up, over how many years, and has anyone actually calculated that number? What penalties apply if the promised jobs don’t materialize — and who pays to decommission the site if it’s abandoned in twenty years? What was decided under nondisclosure before the public ever heard of it? And the one that matters most: if the people who live with the land say no, does their “no” have real force — or is this a managed process designed to produce consent?

These are not radical demands. They are the minimum due diligence any community owes itself before turning its home into a line item on someone else’s balance sheet. If we are going to be lectured about shorter showers and smaller cars, then those who commission million-square-foot server halls and call it progress can be asked to account, in full, for the land, water, and power they intend to spend — and to show us the price before we agree to pay it.

The quiet valleys of this world may yet have more to say about the future than the brochures would like us to believe.

The same questions are arriving in Virginia now. See what it means for Augusta County →

Every large data center company now has a water story to tell. “Water positive by 2030.” “Recycled water for cooling.” “A 95% reduction in evaporative use.” The language is reassuring, and it is meant to be. When a project comes to a rural county, these pledges arrive alongside the slides about jobs and tax revenue, offered as proof that the company will be a careful steward of the water it draws.

I want to be careful and fair here, because the truth has two parts, and both matter.

FIRST, THE PART THAT IS REAL

The physics is genuine. There is a real, physical difference between an evaporative cooling system that loses roughly 80% of the water it withdraws and a true closed-loop system that is filled once and recirculates for the life of the facility. Newer liquid-cooling designs, in the right cool climate, genuinely can reject heat through outdoor radiators with almost no evaporation at all. This is not marketing; it is engineering, and it is possible. I am not going to tell you the technology cannot work, because that would not be true.

So the question is not whether water-saving cooling can exist. It can. The question is whether it is actually being built, at scale, in the places where the water is being drawn — and whether anyone outside the company is allowed to check.

SECOND, THE PART THEY DO NOT ADVERTISE

Here is what the pledges leave out: almost no one is required to prove them. Data centers remain among the least transparent industrial users of water in the country. A peer-reviewed 2026 assessment in AGU Advances found that even a basic efficiency metric — Water Usage Effectiveness — is tracked by fewer than one-third of operators, leaving what the authors call significant blind spots. Most companies disclose only partial or aggregate numbers, rarely facility by facility, and sometimes leave water out of their sustainability reports entirely.

When public bodies have tried to get the real figures, they have run into walls. When the Texas State Water Board asked operators to self-report their water use to support statewide planning, only about a third responded. Utilities are sometimes barred from releasing data center water data to the public by nondisclosure agreements. In California, a bill that would have required operators to share site-level water estimates with local water suppliers was vetoed in October 2025 after industry opposition. In New Jersey, a similar disclosure bill was vetoed as well. Developers often classify their water use as proprietary information — which means the people whose aquifer is being tapped are not allowed to know how much is being taken.

What transparency we do have has often come only under legal pressure. In Oregon, it took a lawsuit to force Google to reveal that a single data center in The Dalles used 355 million gallons of water in 2021 — more than a quarter of that city’s entire supply. That number did not come from a sustainability report. It came from a courtroom.

WHY I WITHHOLD CREDIT

This is why I will not give any company credit for a water pledge on the strength of its own announcement. Not because I assume every company is acting in bad faith — I cannot see inside their intentions, and I will not claim to. But because a promise without independent, verifiable, facility-level proof is not an achievement. It is a hope, published by the party with the most to gain from our believing it.

The standard is simple and it is fair: show the data. Not a corporate-wide average that blends a water-free facility in Scotland with a thirsty one in Arizona. Not a pledge for five years from now. The actual gallons, at the actual site, drawn from the actual watershed, measured and reported to someone who does not work for the company. Until that exists, the honest assumption for any community is not the press-release exception — it is the water-intensive default, because that default is still the cheapest to build and remains the most common choice developers make.

WHAT WOULD ACTUALLY EARN TRUST

I would gladly credit a company that did the following, and I would say so plainly: publish per-facility water withdrawal and consumption, every year, for every site; allow the local water utility to disclose those numbers without a nondisclosure agreement; submit to independent, third-party verification rather than self-reporting; and disclose the water source, so a community knows whether its drinking water, its river, or its aquifer is the one being drawn down.

None of that is unreasonable. Lawmakers in a dozen states, and a federal transparency bill, are now trying to require exactly these things — which tells you how far from normal such disclosure still is. When a company meets that bar, with outside proof, I will update what I write here and give the credit earned. Until then, I am not being cynical by withholding it. I am simply refusing to accept a promise in place of a measurement.

The burden of proof does not belong on the farmer watching the well. It belongs on the industry drawing from it. For data centers to be believed water-positive, they must first be information-positive about the full extent of what they take. Everything else is a story, and we have been given enough stories.

This is the disclosure fight arriving in Virginia now. See what it means for Augusta County →

History has always had a simple rule: when profits and legacy are at stake, those in power rarely stand aside for the sake of ordinary people or the living world. They move forward. They justify the harms. Then they call it progress. We are now watching that rule play out in real time with AI.

In a recent interview, Elon Musk described a future where digital human emulators and humanoid robots make products and provide services far more efficiently than human-run companies. "Corporations that are purely AI and robotics will vastly outperform any corporations that have people in the loop," he said.

Translated into plain English: if we keep humans in our companies, we're choosing inefficiency. In the economic worldview behind this, human beings are no longer the basic unit of enterprise — we're overhead. Once we see it that way, a lot of things suddenly make sense.

THE ECONOMIC MINDSET: HUMANS AS INEFFICIENCY

Look at how the argument is framed. First, the goal is defined as maximizing speed, scale, and efficiency. Bigger models, more data centers, more terawatts of power, more automation across every task from customer service to chip design. Anything that slows that down is a problem to solve.

Then comes the comparison. AI systems can emulate a "digital worker" at a computer. Physical robots can do most forms of labor in the physical world. Together, they don't need rest, health insurance, or a say in how the company is run. If that's the metric, of course a "pure AI" corporation will outperform one where humans still do the work and have to be paid, protected, and listened to. In that spreadsheet, keeping the humans is not a moral baseline — it's a productivity hit.

This isn't an accident in the way they think. It's the point. When the most valuable companies in the world already produce purely digital outputs — software, models, designs, code — it's a short leap to say: why have human teams at all? Why not let AI agents and robots handle everything, and leave people outside the building as occasional customers or spectators? That is the economic mindset we're hearing: humans are not the measure of value, we're a constraint on scaling value.

THE SILENCE, THE SPIN, AND "DOOM TALK"

What's almost more disturbing than Musk's bluntness is the silence from most of the people building and funding AI. They know that once digital workers and robots can do most tasks, the economic pressure will be to replace humans wherever possible. They know that if "pure AI corporations" really do outperform human ones, the market will favor those structures. They know that, in the logic of this system, humans are being moved from "essential" to "optional." But we rarely hear them say this directly.

Instead, we get softer, safer language: "AI will augment humans, not replace them." "Full automation is a possibility, not a given." "We shouldn't catastrophize — this is doom talk." They frame the outcome as one scenario among many, something that might happen if we're not careful, rather than a path they are actively engineering and financing. That gap between what is said publicly and what is being built physically is where a lot of the danger lives.

It's easy to tell people to "wake up." It's harder to admit that most of us will never be given the full, honest picture. We will be assured that this is all about convenience and prosperity, even as the infrastructure of a very different kind of world quietly rises around us. For those of us who are older, or rooted in places that never asked for this build-out, it's hard not to say it plainly: the coming world, if this path continues, is not designed to be friendly to humans. It is designed to be friendly to machines and capital, living on a planet treated as a resource bank.

OPTIONAL PEOPLE ON A NON-OPTIONAL PLANET

There's a dark irony at the center of this. The same leaders who talk about "pure AI corporations" are building physical infrastructure that absolutely cannot function without the Earth: gigawatts of electricity, vast solar arrays, gas turbines, mining for copper and rare metals, and huge data centers drawing water from rivers and aquifers. All of that is rooted in the places where we live.

The data centers don't float in space. The power plants don't drink vacuum. The solar arrays don't bloom in a consequence-free dimension. They sit on land that used to be forests, farms, and habitat. They pull water from the same rivers and wells our communities depend on. They emit heat and noise into real air.

So we have a worldview in which humans inside corporations are optional, but human bodies and landscapes are still expected to absorb the impacts those corporations create. Our town might not be worth hiring, but our aquifer is worth draining. Our skills might be obsolete, but our grid is useful. We may not be "needed" in the spreadsheet, but our valley is needed in the site plan. That is the moral absurdity of this moment. The people and places that make AI physically possible are being treated as expendable, even as the system quietly assumes our resources will always be available.

THE FARMS BENEATH THE CLOUD

It's easy to imagine all of this as something happening "out there," in some distant tech city or orbital launch site. But if we stand in the middle of a farm, the abstraction disappears. Picture a pasture that has been in use for generations. Cows in the field. A line of trees following the creek. A well that has never once failed a family through drought or winter.

Now picture what happens when a hyperscale data center parks itself a few miles away. The well doesn't dry up all at once. It changes slowly. Maybe the water level drops a little deeper each year. Maybe the creek runs lower in late summer. Maybe the soil holds less moisture, and the grass browns faster between rains. The farmer still gets up at dawn, still mends fences, still plants and harvests — but somewhere, unseen, a server hall is pulsing day and night, pulling from the same invisible reserve that keeps that farm alive.

Inside that hall, machines are optimizing customer service flows and ad targeting for companies that no longer hire from the surrounding county. Somewhere, an AI "coworker" is helping design the next robot, the next chip, the next expansion. Humans in the loop are optional. But the water under the farm is not. From the vantage point of the spreadsheet, this is a win: more compute, more efficiency, more revenue. From the vantage point of a farmer watching their land and animals, it is a slow theft — of security, of continuity, of trust that the ground will still give what it has always given.

When we talk about "the cloud," it can feel weightless and abstract. But the cloud has a weight. It rests on real ground, and that ground, more and more, is rural: farms, forests, fields, and small towns whose names never appear in the press release. If the future is "pure AI corporations," then those corporations will be powered by someone's well, someone's river, someone's soil. The question is not whether the world will be friendly to those corporations; it is whether it will remain survivable and dignified for the humans and animals living under their footprint.

THE STORIES WE'RE TOLD, AND THE MONEY WE'RE OFFERED

We don't just face pressure from the technology itself; we face pressure from the stories told around it. As this build-out continues, many of us will be blinded to the cost by the praise and reassurance we receive from the very institutions driving it. We'll be told we are "forward-thinking," "on the right side of history," "doing our part" for innovation.

We should beware anyone who needs to convince us that we are doing the right thing while asking us to ignore what is being taken from the land and water around us. We need to use our common sense. Look not only at their words, but at what their projects extract from the soil, the aquifer, the river. That is where the truth lives, long after the compliments fade.

There is also the matter of money. When a large project arrives, the financial benefits are always front and center: new tax revenue, construction jobs, "investment" in local programs. It's easy, in the glow of those promises, for all of us to equate "bringing in money" with "helping our community." But financial gain that depends on draining our aquifer, heating our rivers, or turning farmland into industrial pads is not free; it is bought at a greater cost than most companies or officials will ever admit. We shouldn't confuse being told we are "community supporters" with actually protecting the community. Before we accept the check, we need to ask what we are trading away in land, water, and quiet — and whether anyone will be there to help us when those are gone.

NAMING THE PLAN, DRAWING OUR LINES

The first step is refusing to pretend this is all accidental or inevitable. It isn't. Leaders in AI and tech are saying, in public, that they expect corporations without people to outperform those with people. They are designing for that. They are building data centers, robots, and orbital infrastructure to make that vision real. Most of the rest of the industry is staying quiet or calling concern "doom." That silence is not neutral; it is part of how this path is normalized.

Once we name the plan, we can be clearer about our response. We may not be able to stop every data center or every robot rollout. But we can insist that our local land, water, and energy are not sacrificed on the altar of "pure AI" corporations. We can demand that our communities retain hard veto power over projects that threaten our basic survival. We can build forms of work, care, and connection that are never reducible to "optional overhead" in somebody's growth model.

Most importantly, we can keep repeating a simple truth that gets lost in all the terawatts and tokens: a corporation, with or without humans, is not a person. It is not a community. It is not a forest, a river, or a child. It is a machine for turning resources into profit according to rules someone wrote. If those rules say humans are optional, then it is up to us — actual humans, living in real places, on real farms — to write different rules, or to refuse to host infrastructure that treats us as disposable.

This is the choice in front of my own community. See what it means for Augusta County, Virginia →

There are things that only happen in a human-sized, real-world life. The hug after a funeral when no words are enough. The quiet triumph of finally fixing something with your own hands. The way a room changes when a hard truth is spoken out loud. These moments don't stream well. They don't scale. They don't leave a neat data trail. And yet they are the very things that make us human.

So why, in this strange era, are we trading away our humanity — our physical presence, emotional depth, and shared vulnerability — for a digital world that can never truly give them back?

THE SUBTLE REDEFINITION OF "REAL"

Over the last decade, something quiet but radical has happened: the center of gravity for our lives has shifted from the physical to the digital. Conversations drift from kitchen tables to group chats and comment threads. Shared experiences shift from local events to livestreams and curated content. Personal milestones are measured not by how deeply they changed us, but by how many likes or views they got.

We don't announce this as a conscious choice. It happens gradually, while we're busy. At first, our devices were tools to help us navigate the world. Now, more and more, the world is expected to fit inside our devices. Reality becomes whatever shows up on our screens. The rest — the unfiltered, unquantified, unphotographed — starts to feel less "real," even though it's the only part that can actually hold us.

WHAT WE LOSE WHEN WE OUTSOURCE BEING HUMAN

Digital life can connect us, inform us, and give us community we might not find locally. But it can also hollow out the parts of being human that can't be digitized.

We begin to lose embodied presence. The small, unplanned interactions in a day — eye contact with a neighbor, chatting in a checkout line, sitting with someone in silence — get replaced with scheduled calls and scrollable feeds. Our bodies become transport vehicles for our screens.

We lose emotional depth. Grief, joy, and anger are compressed into emojis, reaction buttons, and brief posts. Instead of sitting with a feeling until it teaches us something, we rush to express it online and move on. The hard, slow work of processing is replaced by performative sharing.

We lose shared vulnerability. In physical community, people see us fail, age, and struggle. They know our imperfections. Online, we curate. Vulnerability becomes content. The risk that once built trust becomes a brand strategy, carefully managed and monetized.

And we lose meaningful achievement. Triumphs used to be deeply personal: finishing a project, practicing a skill, helping someone quietly. Now, achievement is increasingly tied to visibility and metrics — followers, engagement, reach. If it doesn't show up in numbers, it barely counts. When we outsource so much of our social and emotional life to platforms, we don't just change our habits; we change the wiring of what we think life is for.

THE SYSTEM'S ANSWER TO "WHAT IS LIFE FOR?"

Underneath the apps and interfaces, there is a system answering a question we should never have given away: "What is life for?" The system's answer is blunt, even if it's wrapped in sleek design and inspirational slogans: life is for generating data. Life is for producing and consuming content. Life is for becoming more efficient, more optimized, more "engaged."

From that perspective, a day spent caring for an elderly parent, listening to their stories, and holding their hand counts for very little. A day spent generating clicks, content, and transactions counts for a lot. We start to feel this in subtle ways: guilt when we spend time offline, as if we're "wasting" hours that could be productive; restlessness in slow moments, reaching automatically for a device instead of letting the silence teach us something; a creeping sense that if something isn't documented, shared, and validated online, it somehow doesn't fully exist. That is how humanity gets traded away: not in a single decision, but in millions of tiny choices shaped by an invisible value system that treats human beings as inputs and outputs.

TRIUMPHS WITHOUT WITNESS, LOSSES WITHOUT HOLDING

Humans are built for shared triumph and shared loss. Winning a game alone on a screen feels different than winning with teammates who were there through the bruises and sweat. Losing a job and posting about it online feels different than sitting on a friend's porch and crying into your coffee.

In a digital-first world, more and more of our triumphs and losses happen without embodied witness. Triumphs become announcements, polished and posted, quickly buried under the next scroll. Losses become content — a thread, a short video, a vulnerable essay — consumed and forgotten by people who were never in the room and won't be there when the silence closes in at night. We are, in effect, replacing shared rituals with broadcast statements. The deep communal processes that once helped us integrate joy and pain are being swapped out for rapid confession and reaction. This might look like connection from the outside, but inside, it leaves many people lonelier than ever.

WHAT ARE WE GETTING IN RETURN?

If we're trading away so much — presence, depth, vulnerability, and shared meaning — we should ask a hard question: what exactly are we getting in return? We get convenience. Entertainment. Constant stimulation. A sense of being "plugged in" to a global conversation. Those aren't nothing. But they are not enough to anchor a life.

No algorithm can sit at your bedside in the hospital, knowing your history, your quirks, and the stories behind your scars. No feed can replace the quiet trust built over years of shared meals and repeated forgiveness. We are pouring irreplaceable human moments into systems that turn them into data, content, and revenue. For what? For a version of "connection" that often leaves us more anxious, more distracted, and more ungrounded than before.

CHOOSING TO BE RESPONSIBLY HUMAN

We don't have to reject all digital tools to reclaim our humanity. But we do have to decide, consciously, that they will not be allowed to dictate what being human means. That might look like guarding time for face-to-face conversations where devices stay out of reach; treating some experiences as sacred and deliberately unposted — private triumphs and losses that belong only to the people present; measuring a good day not by how much we accomplished online, but by how fully we showed up for the people and places right in front of us; and asking, before each new "innovation": does this help us be more human together, or does it pull us further apart and deeper into performance?

Our humanity is not a relic. It's not something to be upgraded. It's the only thing that allows us to love, to care for the earth, to build communities that can withstand crisis. The digital world will keep expanding. Systems will keep inviting us to trade more of our lives for pixels, metrics, and convenience. We can't stop that with a single blog post. But we can start, right now, by refusing to give away the parts of ourselves that no machine will ever truly understand: the way it feels when we are fully present with each other, and the quiet knowledge that those moments are what life is really for.

The same trade is happening to the land itself. See what it means for Augusta County, Virginia →

One quiet morning in Goshen, Virginia, I stepped outside to the sound of thrushes in the trees and the rush of a nearby creek. Mist lifted off the fields. The Blue Ridge stood solid on the horizon. Nothing about that scene felt "outdated" or in need of disruption.

Yet later that same day, the county agenda buzzed with talk of data centers, cloud infrastructure, and broadband expansion — another push to make our region more "competitive in the digital economy." The disconnect was striking. Standing under the open sky, you can't help but wonder: why are we so determined to build a more digital world, when it comes at such a cost to the living one?

THE STORY WE'RE BEING SOLD

The dominant story goes like this: digital equals progress. Moving services online makes life easier. Connecting everything makes communities "smart." Embracing AI and data centers brings jobs and investment. From tech ads to government speeches, the message is clear: if you're not racing into the digital future, you're being left behind.

This story is emotionally powerful. It plays on our desire for convenience, our fear of irrelevance, and our hope that technology can solve big problems. But it leaves out something essential: the physical reality behind the pixels.

THE "CLOUD" HAS A FOOTPRINT

We talk about the "cloud" as if it hangs weightlessly in the sky. In reality, it lands on real ground in real places — places like rural Virginia, Iowa, Arizona, and beyond.

A few years ago, I sat in a local meeting where representatives pitched a proposed data center campus not far from our small town. They spoke of tax revenue, fiber upgrades, and the prestige of "joining the digital economy." What didn't appear on their slides was just as important: how much electricity those server buildings would demand, and from where; how many millions of gallons of water would be pulled from our watershed for cooling; how much land would be cleared, paved, and fenced to keep the machines humming.

Most in the room were hearing about "jobs" and "innovation." I was hearing the quieter part: our creek becoming a resource input for distant cloud customers, our landscape becoming a support system for someone else's data. Nothing about that proposal felt immaterial. It felt heavy — on the land, on the water, and on the future.

WHY WE KEEP DOUBLING DOWN ON DIGITAL

If the digital world is so resource-intensive, why do we keep pushing deeper into it? Part of the answer is simple: the benefits are immediate and personal, while the costs are delayed and shared. You feel the convenience of online everything today. You don't immediately feel the strain on the aquifer or the grid. The mining for device components or the dumping of e-waste happens far away, often out of sight and out of mind.

Another part is structural. Our economic system rewards whatever drives quick growth and new markets. Digital expansion does exactly that. It can be built and scaled rapidly. It can be framed as "clean" compared to a smokestack. It can be wrapped in language about "connectivity" and "inclusion," even when the underlying infrastructure is chewing through energy, materials, and water.

And then there is a more uncomfortable driver: control. The more our lives move onto digital rails, the more power accrues to whoever owns those rails. It's not just about better tools. It's about control over communication channels and information flows; control over the data that predicts and nudges our behavior; control over essential services that once existed physically in communities and are now mediated by remote platforms. Building a more digital world gives corporations and governments a finer-grained handle on our attention, habits, and dependencies. That incentive is strong enough to overshadow a lot of ecological reality.

THE COST WE DON'T COUNT

When we talk honestly about the digital boom, we need to widen the lens beyond jobs and innovation to include what rarely makes it into glossy presentations. Energy: data centers, networks, and devices already consume a sizeable fraction of global electricity, with demand rising fast as AI and "cloud everything" expand. Water: large facilities can use millions of gallons a day for cooling, straining rivers, reservoirs, and aquifers — especially in drought-prone regions. Materials and waste: extracting metals and minerals for devices scars landscapes; millions of tons of electronic waste pile up every year, often processed in ways that harm workers and environments far from tech headquarters.

These are not abstractions. They show up in lower river levels, higher utility bills, and degraded ecosystems. They show up when "smart growth" quietly erodes the very places people moved to for quiet, clean water, and stars at night.

RETHINKING WHAT WE CALL "PROGRESS"

None of this is an argument to abandon technology or slam our laptops shut for good. It's an argument to rethink what we mean by progress — and who gets to define it. Instead of asking, "How do we make everything more digital?", we might ask: How much digital infrastructure can our ecosystems actually sustain? Which digital tools genuinely improve human and planetary health, and which mostly feed profit, distraction, and surveillance? How can communities have a real say in whether a new project serves them — or simply uses them as a backdrop for someone else's digital empire?

If the digital world continues to grow by quietly consuming the living one, that is not progress. That is a slow trade of forests for server farms, rivers for cooling loops, community autonomy for platform dependency.

BRINGING IT BACK TO THE CREEK

I think about that morning in Goshen often. The creek was flowing. The birds were loud. The sky felt wide. On days like that, the idea that our future depends on being more "online" rings hollow. What if our measure of a good future were different? Kids who know their local river as more than a line on a map. Towns whose decisions protect water and soil before they chase the next data center deal. A relationship with technology that supports living systems instead of quietly hollowing them out.

We still have time to ask better questions, to draw firmer lines, and to insist that the digital world be a servant to the living one — not the other way around. The next time you hear about a new cloud project, smart initiative, or AI rollout, try this simple test: does this make the real world healthier and more resilient, or does it merely make the digital world richer and more powerful? Our answer to that question, in thousands of local decisions, will shape the future far more than any marketing campaign or buzzword.

This is the question in front of my own community right now. See what it means for Augusta County, Virginia →

We're living through a moment people keep calling "the next Gutenberg revolution." Artificial intelligence is racing ahead, data centers are spreading across farmland and river valleys, and powerful models are becoming the new engines of knowledge and decision-making.

What almost no one is talking about is the quiet divide this creates: between those who get the full versions of these systems and those who don't; between those who decide what information is visible and those who simply live inside whatever feed they are given; between the "cloud" and the ground beneath it — the land, water, species, and communities asked to make room for someone else's future. This is my attempt to sound an alarm before that divide becomes permanent.

THE GROUND BENEATH THE CLOUD

We talk about AI as if it lives in the sky. Press releases and TED talks say "the cloud," "the model," "the intelligence," as though these things float somewhere above the real world. But artificial intelligence sits on the earth. It sits in windowless warehouses packed with servers, in cooling systems that draw millions of gallons of water from rivers and aquifers, in transmission lines and substations built across fields and forests, in light that never turns off, washing out the night sky for miles around.

In Virginia, data centers already drive projections for electricity demand and water planning. Counties are being asked to convert hundreds of acres of farmland into industrial zones. Wells, streams, and ecosystems that have been stable for generations are suddenly in play. And yet, when you listen to national conversations about AI, you rarely hear about rivers, wells, or night. You hear about productivity, innovation, and "frontier models." The gap matters. If we don't name the ground, we won't see who is paying the price for the cloud.

INTELLIGENCE YOU CAN BUY

One of the most striking claims made recently by people inside the frontier labs is that intelligence can now be "purchased" with compute. In a talk given at a 2026 conference, an Anthropic researcher, Chloe Lubinski, put it almost exactly that way: with more money, which buys compute, "you can essentially purchase intelligence," kicking off a cycle that is very hard to stop. More money buys more chips; more chips buy more training runs; better models make more money; the loop spins faster.

That same loop also buys exclusion. The most capable systems today are tiered — internal versions inside the labs with full research capabilities, bespoke enterprise access for big companies, and limited public interfaces for everyone else. Which tier you see depends on who you are: a frontier researcher, a corporate customer, or an ordinary person paying a small subscription and talking to a partial version of the system. When people in power talk about "AI for everyone," they often mean AI assistants and tools for everyone — not equal access to the systems that will shape the laws, markets, and policies we all live under.

The most capable systems are increasingly tiered, and ordinary people rarely get the same access that wealthy institutions do. The risk is that frontier AI becomes a powerful institution's advantage, while the communities hosting the infrastructure get the least access to the very capabilities their land and water make possible.

WHO GIVES UP LAND, WHO GETS THE INTELLIGENCE?

From the vantage point of a rural county, the AI divide looks different than it does from a corporate campus. Land that once grew food or provided habitat is rezoned for data centers. Tax incentives are offered to global companies to build on that land. Promises of jobs and community benefits are made, often in rosy language and PowerPoint slides. And the long-term strain on water, grid stability, wildlife, and local character is not fully understood, let alone discussed in plain terms.

If you live near a proposed site, you may never be told which models your water is cooling, which firms will have access to the intelligence running on your land, or how much of that intelligence will be available to your local schools, libraries, or small businesses. You are asked to host the physical cost of someone else's future — and offered only a partial share of the power that future is supposed to deliver. It is a new kind of inequality: a ground divide over who gives up land, water, and night; and an intelligence divide over who gets full access to the systems that sit on that land. We cannot accept that the people who live nearest the infrastructure, and feel its impacts most directly, will be the ones with the least say and the least access.

FROM GUTENBERG TO AI

We've been here before. When Johannes Gutenberg's press appeared in the 15th century, it didn't just make books cheaper — it broke a monopoly. Suddenly, scripture, scholarship, and political ideas could move faster than bishops and kings. Ordinary people could encounter texts directly, instead of only hearing them summarized or filtered by clergy and scribes. Power noticed.

In response to the flood of printed material, the Catholic Church created the Index Librorum Prohibitorum — the Index of Prohibited Books. Beginning in 1559 and updated for centuries, the Index listed works that lay Catholics were forbidden to read or own without special permission. It targeted not only "heretical" theology, but scientific, philosophical, and literary works: Copernicus and Galileo on the cosmos, Spinoza on scripture and politics, Kant and Hume on reason and faith, Milton and many others whose ideas complicated official doctrine. The stated aim was protection — shielding believers from error. The practical effect was control: a centralized authority quietly deciding which ideas people were allowed to encounter.

It is vital to see this clearly. The Index was not simply a list of "bad books." It was a mechanism for curating reality — defining which parts of the world's thought would count as legitimate and which would effectively disappear from ordinary view.

Today, our "indexes" are rarely printed. They are paywalled journals and books that only some institutions can afford; training datasets that include certain histories and omit others; content policies and safety filters that quietly downrank or remove sensitive topics; recommendation systems that decide which news, commentary, and search results rise to the top — and which sink out of sight. We are told, once again, that this is for our protection: to prevent harm, hate, misinformation, and confusion. Some of that is true and necessary. But the deeper risk is the same: losing the truth of things when all information is curated by those who control us.

The difference is that the old Index was at least visible. People could know which authors had been banned and debate those choices. Today's digital and AI "indexes" are mostly hidden. We rarely see which topics have been flagged as too dangerous, too controversial, or too unprofitable. We rarely see which communities' stories were left out of the training data entirely. That is why building our own small, honest libraries now — from land, rivers, farms, and ordinary lives — is not nostalgia. It is self-defense.

OUR MORAL IMAGINATION IS TRAINING DATA

Frontier labs sometimes say that our "moral imagination" is the raw material their systems learn from: our stories, our values, our fears, our hopes. That is partly true — and partly misleading. Models learn from the language they are fed. If we fill the world with shallow entertainment and partisan noise, they will mirror that. If we preserve deep thinking about justice, ecology, conscience, and responsibility, they can mirror that too.

But models also learn from what is excluded: the histories we never teach, the local voices we never record, the ecosystems we treat as background instead of central. If the training data mostly reflects the perspectives of those who already have access and power, then even the most honest model will reproduce that imbalance. It will know a great deal about markets and consumption — and very little about wells running dry, pollinators losing habitat, or families living under permanent artificial light.

That is why, alongside data centers and models, we need something older and quieter: shelves where people can find banned, ignored, or inconvenient works; local histories that remember rivers, forests, and farms; spiritual and philosophical texts that ask harder questions than "how do we optimize this quarter?" The point is not to freeze the past. It is to ensure that when we train machines — and our own children — we are not feeding them a curated, flattened version of reality.

A SMALL LIBRARY AS A LINE OF DEFENSE

What can ordinary people do in the face of all this? We can't single-handedly rewrite corporate policies or global AI governance. We can't stop every data center or reverse every zoning decision. But we can do something both humble and radical: build and preserve a physical, human-chosen library of works that matter; teach our neighbors and children how to read them, argue with them, and hold them against whatever feed they are given; and insist that local institutions — libraries, schools, churches, councils — see access to deep knowledge as a responsibility, not a luxury.

That library should include foundational religious and wisdom texts from multiple traditions; philosophical works on ethics, power, and personhood; books that were once banned or targeted by systems like the Index; environmental and scientific works that describe the real, physical world; histories that remember both triumphs and harms; literature that trains empathy, imagination, and the ability to live with complexity; and serious analyses of technology, media, AI, and the governance of compute. This is not about agreeing with every author. Many of these books conflict with each other. That is the point. A curated truth wounds us because it removes the friction between competing perspectives. A living library preserves that friction so real thinking remains possible.

A CALL FROM THE GROUND

I'm writing this from the ground beneath the cloud — from a rural place watching data centers move closer, watching rivers and skies enter the vocabulary of "AI growth." I don't distrust intelligence. I don't fear machines in themselves. What I fear is the combination of concentrated power, curated information, and environmental disregard — the same pattern we have seen, again and again, in other eras, now accelerated by models and chips.

If we are indeed living through a new Gutenberg moment, then the most important question is not "how smart will AI become?" but: Who will be allowed to see the full range of ideas and histories? Who will be quietly shielded — or stripped — from them? And who will decide? We cannot fully control how labs and governments answer those questions. We can, however, refuse to leave them entirely in someone else's hands. By building small libraries, honest blogs, and local conversations — from the ground up — we keep alive the possibility that our children and neighbors will know more than one authorized story about the world. We give them a chance to see the land and water under their feet as part of the truth, not a footnote to someone else's cloud.

This divide is not abstract where I live. See what it means for Augusta County, Virginia →

In the early days of coal, Britain's industrial cities filled with smoke long before people had a language for "air pollution" as a moral problem. Factories were celebrated as engines of progress; chimneys became symbols of prosperity. Children walked to school through soot-laden streets, lungs quietly absorbing the cost of an empire's success. The industrial culture was not thoughtless — it had rules, norms, and eventually laws. But those rules were written inside a mental boundary that did not yet include the atmosphere itself as a subject of justice.

We are now building a different kind of engine, made of silicon and code instead of coal and iron. The AI industry, especially in places like Silicon Valley, is full of people who are earnestly trying to be careful. They write standards for secure models and safe agents. They invent new risk categories — privacy, fairness, hallucinations, governance — and develop elaborate cultures so employees will use AI responsibly, report near misses, and treat each other with care. On paper, it looks like an ethical triumph: we have learned to worry about bias and consent, and we tell ourselves we are doing better than the factory owners of the nineteenth century.

But these sophisticated rule-sets float on a physical infrastructure that is growing like an unseen shadow. Data centers consume gigawatts of electricity and billions of liters of water, spread across deserts, river valleys, and fragile grids. There is no new planet waiting in reserve, no hidden reservoir of water that will appear if we overdraw the one we have. The freshwater now cycling through our rivers, lakes, aquifers, and clouds is the same water that has been here for millions of years, endlessly reused but never replenished from outside. Every cooling tower, every new rack of servers, participates in that closed loop whether the people who write AI policies think about it or not.

The risk is not that AI engineers and security leaders are malicious. The risk is that they are so focused on being careful inside the narrow domain of digital ethics — what data is allowed into a prompt, which outputs must be verified, how employees should report a model's mistake — that they do not notice they are collectively rebuilding the old industrial pattern at planetary scale. We are at risk of inventing pollution again, this time in the language of "compute," "load," and "cooling," while congratulating ourselves for our frameworks and our "just cultures."

The coal comparison can sound dark, but it is not offered as a prophecy of doom. It is a reminder that cultures of care can coexist with systems of harm when the boundary of that care is drawn too small. Nineteenth-century Britons did not imagine clean air was impossible; they simply did not yet see the air as a stakeholder. Today, we have the advantage of hindsight. We know what happens when a powerful industry treats land, water, and atmosphere as background conditions instead of as fellow participants in our story.

Nothing about this trajectory is inevitable. The same capacity for moral imagination that produced AI safety standards and security cultures can be extended outward. We can decide that "responsible AI" includes caps on resource use, transparency about environmental footprints, and real veto power for communities and ecosystems that cannot afford another extractive boom. We can build governance that tracks not only data leaks and hallucinations, but also near misses in aquifers and grids — and treats those signals as seriously as any compliance report.

It is not hopeless because the infrastructure we are building is still young, and the narratives around it are still pliable. We can choose to be the first civilization that learns from the coal era in time, not in regret. We can look up from our rulebooks and see the single, finite hydrological cycle that sustains every living thing on Earth, then write our AI stories so that this cycle is not an afterthought but a central character. Hope, in this context, is not a feeling that everything will be fine on its own. It is the decision to widen the circle of what we count as "responsible," while there is still room to change course.

This is not theoretical where I live. See what the circle drawn too small means for Augusta County, Virginia →

Virginia lawmakers are celebrating a budget deal that will impose a new electricity tax on data centers — $0.011 per kilowatt-hour — projected to generate $1.2 billion over two years, or roughly $600 million a year. The headlines make it sound like a victory. The tech companies, for their part, have gone relatively quiet.

That silence should tell you everything.

THE NUMBER THEY DON'T WANT YOU TO COMPARE

Let's start with the math Virginia's press releases bury. In fiscal year 2025 alone, the state's sixteen-year-old sales tax exemption for data center equipment cost Virginia $1.6 billion in a single year — up an astonishing 118% from the prior year. That exemption was originally projected to cost $1.54 million per year when it was signed into law in 2008. The actual cost is now more than 100,000% above the original projection.

The arithmetic is stark: a tax break that began at $1.54 million a year has grown into a $1.6 billion annual giveaway — while Virginia families struggle to pay for childcare, health insurance, and a place to live. Senate Finance Chair Louise Lucas spent this budget cycle pressing to repeal that exemption outright, relenting only when negotiators agreed to defer the question to a workgroup that does not report until November 1.

Now compare that giveaway to the deal Virginia just struck. The annual sales tax exemption for FY2025 runs between $1.6 and $1.9 billion. Total exemptions given since 2010 come to roughly $2.7 billion. Against that, the new electricity tax raises about $600 million a year, capped — leaving a net annual shortfall of more than $1 billion still favoring the industry.

Virginia is collecting $600 million per year from an industry it is simultaneously handing $1.6–1.9 billion in exemptions. The sales tax exemption stays intact, running through 2035. The workgroup studying whether to change that doesn't even report until November 1. In other words: the state just added a modest new fee while leaving the main giveaway untouched.

WHO ARE WE REALLY TALKING ABOUT?

These are not small businesses scraping by on thin margins. The primary tenants and operators of Virginia's hundreds of data centers are among the most profitable corporations in human history.

Amazon Web Services reached a $123 billion annual revenue run rate in Q2 2025, with operating income of $10.2 billion in a single quarter. AWS alone has at least 67 data centers across five counties in northern Virginia. Microsoft Azure generated $29.9 billion in cloud revenue in Q2 2025 alone, growing 39% year-over-year; Microsoft's total operating profit for fiscal year 2025 was $128.5 billion. Google Cloud earned $13.6 billion in Q2 2025, up 32%, and has committed a $9 billion investment in AI-ready Virginia facilities. Together, just these three companies generated roughly $75 billion in cloud revenue in a single quarter of 2025.

For these companies, a $600 million annual tax across the entire Virginia data center industry is a rounding error. AWS alone earns more than that in operating income roughly every two to three weeks. The tax is not a burden. It is a bargaining chip that was already priced in.

THE JOBS MYTH, BY THE NUMBERS

The data center industry has relentlessly sold Virginia on a jobs-and-growth narrative. Let's examine it. Virginia currently has 667 data centers and approximately 12,140 permanent data center jobs — an average of just 18.2 full-time jobs per facility. A typical 250,000-square-foot data center employs approximately 50 full-time workers, about half of whom are contract employees.

According to Food & Water Watch, the investment required to create a single permanent data center job in Virginia is nearly 100 times greater than for other industries. Virginia data centers generate just one permanent job for every $13 million invested, compared to $137,000 per job in other sectors.

Construction employment is real — but temporary. At the height of construction, approximately 1,500 workers are on-site from various trades. Once the ribbon is cut, most of them move on. The enormous, humming buildings left behind are staffed by skeleton crews. Meanwhile, as of 2024, the entire U.S. data center industry — which consumes more than 4% of all electricity in the country — employs as few as 23,000 permanent workers nationwide, representing just 0.01% of total U.S. employment.

Virginia has traded $2.7 billion in tax exemptions over a decade for an industry that averages 18 permanent jobs per building. By any honest economic measure, that is a catastrophic return on public investment.

YOU ARE PAYING THEIR ELECTRIC BILL — WHETHER YOU KNOW IT OR NOT

The data center industry does not just benefit from tax exemptions. It is reshaping Virginia's entire energy system — and ratepayers are absorbing the cost. Dominion Energy has received data center power requests totaling 70,000 megawatts — nearly triple the utility's all-time peak load of 24,678 MW recorded in January 2025. Virginia data centers alone already consume the equivalent of 25% of the entire state's electricity needs.

The infrastructure to meet that demand doesn't build itself. Virginia ratepayers may bear more than $440 million in costs over the lifetimes of two major transmission projects — the Mid-Atlantic Resiliency Link and Valley Link — even though that capacity primarily serves data centers. Dominion implemented rate increases in 2026 after State Corporation Commission approval, adding approximately $11 to $13 per month to a typical residential customer's bill. A bill that would have prevented utilities from passing data center infrastructure costs to other customers was quietly carried over to 2027.

Here is what that means in plain language: Virginia families are subsidizing not just the tax breaks these companies receive — they are also underwriting the electrical infrastructure those companies need to operate.

THE WATER NOBODY IS ACCOUNTING FOR

Then there is water — and this is where the conversation becomes most personal for communities across Virginia, including ours. Virginia's data centers used more than 1.85 billion gallons of water in 2023 — up from 1.13 billion gallons in 2019, a 63% increase in four years. In Loudoun County alone, data centers increased their water use by 250% between 2019 and 2023, consuming 900 million gallons in a single year.

On the hottest peak days, data centers already draw an average of 14 million gallons per day from the Potomac River. With unchecked growth, that figure could balloon to 170 million gallons per day. By 2050, data center demand in the Potomac River Basin alone could consume 33% of available river water. A single large data center can consume more water daily than 1,000 typical households; Virginia Tech experts have warned that the largest facilities use as much water as a small city.

This water is not purchased. It is drawn from rivers, aquifers, and public water systems that communities, farms, and ecosystems depend on — often in areas already identified as water-stressed, including areas experiencing ongoing drought. The new electricity tax says nothing about water. There is no water impact fee. There is no mandatory compensation to downstream communities. There is no mechanism for rural Virginia — where our watersheds originate — to recover anything at all.

THE RETURN ON INVESTMENT VIRGINIA NEVER SHOWS YOU

Virginia's own independent auditors at JLARC have found that for every dollar of sales tax abated through the data center exemption, only 48 cents is returned to the state budget. That is a negative return of 52 cents on every dollar given away. Georgia's independent audit found an even worse outcome: data centers there receive $50 million in sales tax exemptions but generate only $15 million back — a 70% negative return on investment.

The Data Center Coalition, for its part, argues that the industry generated $2.1 billion in total state and local tax revenue in fiscal years 2024 and 2025. But that figure must be weighed against $1.6–1.9 billion in sales tax exemptions granted in FY2025 alone, $2.7 billion in exemptions over the past decade, $440 million or more in ratepayer-subsidized grid infrastructure, billions of gallons of public water drawn with no compensation, foregone revenue in local property taxes via abatement programs, and the long-term environmental remediation costs no one has yet priced.

The industry is not a net contributor to Virginia's public finances. It is a net beneficiary — by a wide margin.

WHO IS THE $1.2 BILLION ACTUALLY FOR?

The new tax deal will generate $1.2 billion over two years — earmarked for teacher raises, healthcare support, SNAP administration, transit, and federal funding gaps. This is real money for real needs, and it is welcome. But let's be clear about why those needs are so acute.

Virginia has a $3 billion increase in Medicaid costs driven by managed care contracts. It needs $150 million to replace health insurance subsidies that the Trump administration let expire. It needs $135 million for a new state share of food assistance mandated by the federal tax package. Virginia is having to scramble for revenue to cover basic human needs because federal support is collapsing — and the industry that has consumed more than half of all state economic incentive spending over the past decade is being asked to contribute the equivalent of roughly one-third of what it was given in exemptions last year alone.

The companies will absorb it. Their lobbyists have already called it "manageable." Not a single hyperscale data center will close. Not a single server rack will go dark. Amazon's quarterly operating income from AWS is approximately five times larger than this annual tax.

WHAT A REAL ACCOUNTING WOULD LOOK LIKE

A genuinely fair reckoning with Virginia's data center industry would include full repeal — not deferral — of the sales tax exemption, which the Senate correctly called a blank check that has spiraled from $1.54 million to $1.9 billion in less than two decades. It would include water impact fees tied to actual consumption, with revenues directed to watershed protection, rural water system maintenance, and drought resilience. It would include full cost separation on grid infrastructure, so that no ratepayer subsidizes transmission, substation, or distribution upgrades built primarily to serve data center demand.

It would include mandatory local benefit sharing — a percentage of state data center tax revenues directed to the localities bearing the environmental, infrastructure, and quality-of-life burdens. It would include transparent, company-level disclosure of which corporations receive which exemptions, something Virginia currently does not disclose publicly. And it would include community veto rights over new data center proposals that exceed defined water, land, or noise impact thresholds.

THE BOTTOM LINE

Virginia has spent sixteen years building what is now described as "the data center capital of the world." The price of that distinction is $2.7 billion in foregone taxes, 1.85 billion gallons of water consumed annually, a grid being rebuilt at ratepayer expense, and communities from Northern Virginia to the Shenandoah Valley living next to industrial facilities that were never asked to fully pay their way.

The deal's architects framed it as a win for working families. In a joint statement, Senator Lucas and House Appropriations Chair Luke Torian called the agreement a reflection of "our shared commitment to making Virginia more affordable for families," describing it as a conference report that makes "historic investments to lower costs, strengthen our schools, protect access to healthcare, expand economic opportunity, and maintain the Commonwealth's strong fiscal foundation." Those investments are real, and they matter. But it is worth asking what was left on the table to secure them.

The new $600 million annual electricity tax is not a reckoning. It is a concession — reached only because a budget deadline loomed and lawmakers could not agree fast enough on a broader compromise. The sales tax exemption, the biggest giveaway of all, lives on.

The companies know this. That is why they went quiet. Virginia's communities deserve to understand what their state actually traded away — and what a genuine accounting would demand in return.

These are not abstractions — they are decisions being made right now, including close to home. See what it means for Augusta County, Virginia →

We talk about AI as if it lives in the cloud, floating somewhere above the messy realities of land, water, and darkness. In Broadview, Montana, "artificial intelligence" now has a proposed street address: 5,000 acres of prairie outside a town so small it doesn't have a stoplight. It has a thirst that never sleeps, demanding water day and night to cool warehouses of machines that never agree to rest. And it has a glow — a permanent artificial dawn — that threatens to wash out the night sky for the people and wild creatures who have always depended on darkness to navigate, to hunt, to sleep.

In that new light, everything changes. Migrating birds that once followed the stars will be pulled off course by an industrial halo on the horizon. Nocturnal animals that rely on the cover of night will find their world flattened into a constant, disorienting shimmer. Families who moved there for the quiet and the stars will draw their curtains against a facility that pulses and hums 24 hours a day, its lights reflecting off bedroom windows and barn roofs, its presence turning "home" into the edge of a machine.

And at the center of this is water — the same fragile groundwater that fills kitchen taps, feeds cattle, and keeps native grasses alive through dry spells. Every gallon diverted to cool server racks is a gallon that doesn't return to a well, a creek, a root system. In a place where people already plan showers around water pressure and worry about their wells running low, the idea of a massive, continuous industrial user isn't just a technical question. It is an existential one. For the nearby families, for the ranchers and farmers, and for the native species that have adapted to this particular balance of light, dark, wet, and dry, the arrival of a 24/7 data campus is not an abstract innovation. It is a direct decision about whose needs matter — and who becomes collateral damage in someone else's digital convenience.

In Broadview, that decision is being made on a staggering scale. A 5,000-acre campus is being mapped around a community of roughly 140 people, a place with one café, one gas pump, and no stoplights. The proposed data center complex would stretch across the landscape in a way that, once built, is effectively permanent: concrete, steel, transformers, transmission corridors, and an industrial skyline where there was once only open prairie and snow-framed mountains. For local residents, this is not a story about technology. It is about losing the horizon line they wake up to every morning, the stars they teach their children to recognize, and the groundwater they depend on in a semi-arid climate.

The environmental risks are specific, measurable, and deeply local. Cooling systems for large data centers can consume water at a rate comparable to a small city, especially in hot or dry conditions where evaporation is high. That water does not fall from the sky as a special "AI supply"; it comes from the same constrained aquifers and surface sources everyone else relies on. When you stack that demand on top of existing domestic wells, agricultural use, and the needs of native vegetation and wildlife, the math becomes unforgiving. Even small drawdowns in groundwater levels can mean dry taps for households, reduced flow in springs and creeks, and stress for species that are already living at the edge of their environmental tolerance.

Light pollution carries its own quiet violence. People who have never lived under a truly dark sky may not understand how radically a single industrial complex can alter the night. The constant glow from high-mast security lighting and illuminated buildings can obscure the Milky Way for miles, disorient migratory birds, disrupt the feeding and breeding patterns of bats and insects, and alter predator–prey relationships that evolved in the rhythm of day and night. For families living nearby, it means the end of dark backyards, the end of teaching children the constellations, the end of lying awake under a sky that feels like it belongs to you.

Noise and heat add another layer. Massive fans, chillers, and backup generators introduce a continuous hum and periodic roars into landscapes that have been defined by wind, birdsong, and the occasional truck on a two-lane highway. Waste heat, vented into the surrounding air or water, subtly shifts local microclimates and can affect nearby streams and wetlands if not carefully managed. None of these factors are "anti-technology"; they are straightforward environmental stressors that stack on top of each other until a place is no longer the same place, even if every home and barn technically remains standing.

The families who live around these proposed sites feel that shift first. They are the ones who notice when the well takes longer to recover, when the frogs are quieter in the ditch, when the stars fade behind a permanent industrial twilight. They are the people who will live with the aftermath long after construction crews move on and press releases celebrate new capacity in "the cloud." They are asked to absorb the risks so that far-off users can enjoy seamless digital services with no awareness of the watershed, the night sky, or the species being asked to bear the cost.

This is why it is so important to separate our feelings about the technology itself from the physical infrastructure being built to support it. One can be fascinated by the potential of new tools and still insist that their supporting facilities respect water limits, protect dark skies, and safeguard local ecosystems and communities. The problem is not curiosity or creativity; it is the quiet assumption that the land and water around rural towns are empty, available, and expendable.

From my home in rural Virginia, I recognize Broadview's story. Different state, different watershed, different species — but the same pattern: large energy-hungry facilities proposed in places that officials sometimes treat as blank space on a map. The same meetings where residents ask very simple questions about water use, light, noise, and wildlife, and are told that the answers will come later, after approvals are granted. The same pressure to accept an irreversible transformation of the landscape in exchange for promises that are vague at best and, even when fulfilled, cannot replace a lost aquifer, a dark sky, or a disrupted migration path.

In my book project, I call this moment "the last window" — the brief time we have to look clearly at what is being built, what it demands from our rivers and aquifers and night skies, and what it will mean for the people and species who already belong to these places. It is not a window into algorithms or abstractions. It is a window onto land and water, onto the line where new infrastructure meets old ecosystems. If we close that window without really looking, we will tell ourselves the story that nothing much changed — that the cloud is weightless, that the night is still dark, that the water will always be there when we turn the tap. But the families and creatures who live in the shadow of these facilities will know otherwise.

If Broadview's story sounds familiar, it is because the same questions are being asked closer to home. See what it means for Augusta County, Virginia →

The rise of AI-mediated governance did not begin with chatbots, recommendation engines, or automated moderation. It began earlier, in a long cultural shift away from resilience, autonomy, and adult self-management, and toward institutional protection, emotional management, and dependency. The core argument of this essay is that tech companies did not simply build powerful AI systems and then apply them to society. They first helped normalize a social psychology in which many people came to expect large institutions to protect them from discomfort, manage interpersonal conflict on their behalf, and organize daily life in increasingly intimate ways.

Seen from that angle, AI paternalism is not an abrupt break with the past. It is the next step in a model of governance that has been incubating for decades inside workplaces, platforms, and digital culture. The promise sounds benevolent: safety, support, reduced friction, and intelligent assistance. But the underlying trade is older and more serious. As institutions assume greater responsibility for emotional life and practical decision-making, individuals lose practice in resilience, judgment, and self-government. What is marketed as care can become a quiet architecture of control.

A DIFFERENT MORAL BASELINE

For much of the twentieth century, American social norms placed strong emphasis on emotional toughness and personal responsibility. Children were commonly taught that insults reflected the character of the speaker more than the worth of the target. The phrase "sticks and stones may break my bones, but words will never hurt me" was never literally true, but it served as a cultural instruction: do not hand your emotional center over to every careless or hostile person. The larger lesson was that adult life would include friction, mockery, unfairness, and disagreement, and that part of maturity was learning how to absorb those experiences without collapsing.

That older framework had obvious limits. It could excuse cruelty, understate genuine psychological harm, and ignore power imbalances. But it also assumed that the individual had an important inner role to play in managing adversity. The default question was not, "Which institution will protect me from this?" but rather, "How should I respond?" That orientation encouraged an internal locus of control. It treated resilience not as a denial of pain, but as a civic and personal capacity that had to be cultivated.

This baseline matters because the current moment is often described as if greater emotional sensitivity simply arrived as moral progress. A more accurate account is that one moral vocabulary displaced another. Instead of seeing many ordinary verbal injuries as part of life to be navigated, institutions increasingly treat them as harms requiring supervision, intervention, and documentation. The consequences of that shift reach well beyond manners. They alter what adults expect from employers, platforms, and eventually machines.

THE THERAPEUTIC TURN AT WORK

Over time, many workplaces adopted a more therapeutic understanding of conflict. Human resources departments expanded their role, not just as administrators of pay and benefits, but as interpreters of interpersonal conduct, emotional safety, and organizational values. Behaviors once described as rude, sharp, insensitive, or unprofessional were increasingly folded into broader categories such as hostility, bullying, and psychological harm.

Some of this development was justified. Workplaces did need stronger tools for dealing with discrimination, coercion, harassment, and abuse that had long been minimized or ignored. The problem is that the new framework often expanded beyond those serious cases and blurred the distinction between actual misconduct and ordinary social friction. In that environment, disagreement could more easily be recoded as aggression, discomfort as injury, and offense as evidence of victimization.

A key change accompanied this reclassification: conflict became something to be escalated rather than worked through. The employee was encouraged to report, document, and seek institutional remedy. The institution, in turn, increasingly positioned itself as the legitimate authority to judge not only what had been said or done, but how it should be experienced. The result was not simply greater kindness. It was a transfer of interpretive power away from the individuals in the room and toward managerial systems.

As this mindset spread, an important social expectation shifted. Adults became more likely to assume that if they felt emotionally harmed, some official structure should step in and validate that experience. This is one of the cultural preconditions for AI paternalism. Automated governance works best in populations already accustomed to external adjudication of feelings, risk, and acceptable behavior.

HOW TECH TURNED CARE INTO STRUCTURE

The technology sector intensified this broader tendency by combining therapeutic culture with unusually immersive corporate design. Large tech firms did not merely offer jobs. They built environments organized around convenience, comfort, and emotional reassurance: free meals, laundry services, transportation, wellness programs, on-site medical care, social events, life management support, and sprawling office "campuses" styled less like traditional workplaces than like universities or contained communities.

These benefits were often celebrated as generosity or enlightened management. But they also served structural purposes. They blurred the boundary between work and life, increased the number of hours employees remained physically and psychologically attached to the company, and encouraged a sense that the employer was not just a purchaser of labor but a caretaker. The language of campus was revealing. It suggested continuity with student life: a world in which meals appear, services are nearby, administrators shape the environment, and the institution takes broad responsibility for well-being.

This model has deep consequences. It extends adolescence rather than marking a clear transition into independent adulthood. It cultivates dependency while presenting that dependency as privilege. And it gently recasts authority as benevolence. The company becomes less like an employer and more like a hybrid of parent, school, and moral community.

Once that arrangement is normalized, it becomes easier for workers to accept the company's role in shaping speech norms, social conduct, acceptable beliefs, and emotional interpretation. The result is a workforce trained to understand institutional care and institutional supervision as two sides of the same relationship. Protection is no longer separate from control; it is one of control's preferred public justifications.

FROM RESILIENCE TO MANAGED FEELINGS

At the cultural level, this produces a subtle but important transformation. A society organized around resilience expects individuals to develop judgment, boundary-setting, proportion, and tolerance for friction. A society organized around managed feelings expects institutions to reduce psychological discomfort, certify harms, and enforce norms on behalf of the vulnerable.

The newer framework has an intuitive moral appeal. It sounds compassionate. It promises relief from cruelty and exclusion. But it also changes what kinds of adults a society tends to produce. When institutions repeatedly intervene to classify speech, mediate tension, and frame disagreement as a safety problem, people get less practice in handling conflict directly. They become more likely to interpret distress as something caused externally and solvable administratively.

That is not merely a psychological development; it is a political one. Citizens who are habituated to asking authorities to settle social and emotional disputes are more likely to tolerate broad systems of observation, classification, and intervention. The habit of self-government weakens when the first instinct is escalation to management, moderation teams, or safety protocols. Under those conditions, freedom starts to feel less like a responsibility and more like an unmanaged risk.

This is one reason the language of harm now travels so easily between workplace, platform, and state. The same conceptual template appears everywhere: speech can wound, unmanaged interaction is dangerous, expert systems must intervene, and protection legitimizes surveillance. Once that logic is accepted, the technical means of enforcement become the next frontier.

AI AS THE PARENTAL INTERFACE

AI enters this landscape not as a foreign force but as a highly scalable extension of habits already in place. Human moderators, HR staff, policy teams, and campus administrators can only process so much. AI promises to do the same kind of work at enormous scale: sort behavior, detect risk, flag tone, route complaints, personalize protection, and deliver institutional responses in real time.

This is why AI systems so often speak in the language of reassurance. They do not present themselves as instruments of discipline. They present themselves as helpers. They acknowledge frustration, validate concern, and explain that a process is underway. They are the friendly surface of a deeper architecture that remains opaque to the user. A chatbot cannot usually explain why a classifier flagged an account or why a ranking system suppressed a post. It can, however, keep the person calm, moving, and procedurally contained.

That is paternalism in modern form. The user is not invited into a transparent dispute among equals. The user is managed. Decisions are made elsewhere. The interface speaks gently while access, visibility, and recourse are determined by remote systems. This combination of empathy and opacity is powerful because it feels less coercive than older forms of authority. It dresses control in the language of care.

It also solves a practical problem for large institutions. Direct human accountability is expensive, slow, and contestable. AI-mediated governance is cheaper, faster, and easier to standardize. It allows organizations to maintain the appearance of responsiveness while limiting genuine dialogue. The system need not persuade the user. It only needs to absorb the user's reaction and return them to the workflow.

WHY PEOPLE TRUST THESE SYSTEMS

Many people now genuinely believe that large technology companies will protect them, not only from spam or criminal abuse, but from discomfort, offense, confusion, and social conflict. That belief did not appear by accident. It was trained into users and workers through years of paternal design, emotionally therapeutic language, and service environments that reward dependency.

Once a person comes to see platforms and employers as benevolent managers of reality, AI becomes a natural next step. Why not let the system decide what content is too harmful to see? Why not let automated support tools handle disputes? Why not let predictive systems identify dangerous people, risky speech, or destabilizing ideas before any human damage occurs? Each step sounds efficient and humane when judged inside the protective framework.

But the cost is profound. The more these systems absorb responsibility for interpretation and decision-making, the less room remains for adult discernment. Individuals outsource not only memory and convenience, but judgment. They begin to accept a world in which invisible systems pre-sort what can be said, what can be seen, and which forms of dissent count as legitimate. In exchange, they receive comfort, friction reduction, and the promise that someone or something is watching out for them.

A culture already oriented toward dependency is especially vulnerable to this trade. People who have not been encouraged to build resilience or confront conflict directly are more likely to welcome systems that promise to eliminate both. That welcome is sincere, but it can also be fatal to democratic character.

THE POLITICS BENEATH THE PSYCHOLOGY

The implications extend well beyond office culture or online etiquette. A population accustomed to paternal institutions is easier to govern through paternal technologies. If adults are trained to expect protection, curated environments, emotional validation, and managed outcomes, then AI can become the default mediator between citizen and institution.

That matters politically because AI governance is not neutral. The entities building these systems are not disinterested caretakers. They are corporations and governments with incentives, liabilities, reputational concerns, and operational goals. They decide which risks matter, which harms count, which speech is suspicious, and which appeals are legitimate. The more their systems mediate daily life, the more power shifts away from local judgment and reciprocal human negotiation toward centralized classification.

This is where the issue intersects with material infrastructure. The same companies building massive AI and data center capacity are also expanding the automated systems that shape speech, access, behavior, and public discourse. Communities bear the physical costs of these infrastructures in land use, energy demand, water consumption, and environmental strain. At the same time, the digital services powered by that infrastructure increasingly govern how those same communities can speak, organize, and contest decisions. The relationship is not accidental. Physical centralization and informational centralization are reinforcing processes.

The public conversation often treats AI ethics as a matter of bias, safety, or future existential risk. Those concerns matter. But there is another issue that deserves equal attention: the gradual erosion of adult autonomy under a regime of machine-mediated care. The danger is not only that AI will become too intelligent. It is that citizens may become too willing to be managed.

RECOVERING A COMMON-SENSE RESPONSE

A common-sense response does not require nostalgia for every norm of the past. Genuine abuse exists. Words can wound. Institutions have legitimate roles in setting boundaries and addressing serious misconduct. The question is not whether protection should ever exist. The question is what kind of human being a society aims to produce, and what kind of authority structure it normalizes in the name of care.

A healthier culture would distinguish more carefully between danger and discomfort, between abuse and disagreement, between needed support and infantilizing supervision. It would recover the idea that resilience is not cruelty and that autonomy is not abandonment. It would also insist that technologies designed to assist human beings should not quietly replace their judgment or weaken their capacity for self-government.

That means asking harder questions of both institutions and tools. Does a platform increase a person's agency, or merely manage their reactions? Does a workplace support adults, or encourage dependence? Does an AI system inform users clearly, or absorb them into an opaque administrative process? Does the language of safety clarify real threats, or does it expand authority over ordinary life?

These questions matter because the future of AI will not be determined only by technical capability. It will also be determined by the character of the people expected to live under it. A resilient, self-governing public will demand tools that respect agency, transparency, and contestability. A dependent, protected, permanently managed public will ask to be ruled gently — and may never notice how much of life has already been handed over.

Over the last year, I've had a series of chance conversations in grocery stores, parking lots, and public meetings with people who did not ask to be part of the "AI revolution." Some are small business owners. Some sit on local boards. Many are teachers trying to keep a classroom running on a tight budget.

Again and again, I hear the same two things. First: "I don't want AI in my classroom. My students already have enough to deal with." Second: "Our town would never agree to a big data center." There is a calm certainty in those statements — a belief that if we just refuse, the wave will go around us instead of through us.

Recently, I watched Ray Kurzweil's latest interview about artificial intelligence and the future. The contrast between that conversation and these everyday encounters could not have been sharper.

THE STUDIO FUTURE: AGI, IMMORTALITY, AND SELF-DRIVING EVERYTHING

In the interview, Kurzweil calmly predicts that by 2029 artificial intelligence will match the best human minds in every field, and that by the 2030s we will have AI "inside us" so seamlessly that we will not know if a thought came from our biological brain or our computational one. He talks about "longevity escape velocity" by 2032, when each year of life buys you more than a year of additional expected lifespan, and about simulated humans replacing years of drug trials with a few days of computation.

He and the interviewer speak easily about robots filling factories, universal basic income cushioning job losses, and average wealth rising again and again as automation accelerates. They point to charts showing that average income has multiplied roughly tenfold over the last century and take that as evidence that technology "makes everyone richer" in the long run. Listening to them, you could almost forget that every one of these promises rests on land, water, electricity, and people somewhere having to absorb the cost.

THE WAVE THAT LOOKS HARMLESS FROM FAR AWAY

The image that comes to mind is a tsunami. Out in deep water, the wave looks harmless — just a gentle rise on the surface that two men in a studio can surf for sport. But as it races toward shore, the water piles up, and the people standing on the beach — teachers, parents, small business owners, rural communities — are the ones who will feel it hit first.

The people I meet are standing on that beach, focused on lesson plans and local politics, certain that they can keep AI out of their classrooms and that their towns can simply say "no" to a data center. Meanwhile, Kurzweil is describing a world in which AI is literally inside our bodies, our jobs are reshaped or erased, and the demand for compute grows by orders of magnitude. These two futures are on a collision course, and most people have no idea.

ENDING CAR CRASHES — BY ENDING YOUR RIGHT TO DRIVE

One small moment in the interview captures this perfectly. Kurzweil notes that roughly 40,000 people die in car accidents every year because of human driving and points to autonomous vehicles as the solution. He highlights that Waymo has had essentially no fatalities in its operations so far and suggests that if we just get humans out of the driver's seat, we can "do away with accidents largely."

I want fewer deaths on the road as much as anyone. But notice what goes unsaid. In this story, the way we solve the problem is by taking away personally driven cars altogether, shifting control of mobility to whoever owns and operates the autonomous fleets. Your freedom to drive yourself becomes a public-health risk to be eliminated, rather than a part of your life that might be redesigned together with you.

It is the same pattern we are already seeing with data centers and AI more broadly. Disruption and change are treated as things that can be managed from above if ordinary people will simply accept that the experts know best. You and I are invited to be passengers, not decision-makers.

WHAT THE CONVERSATION LEAVES OUT

The interview is not just optimistic — it is also incomplete. There are several blind spots that matter deeply for people who will be living with the consequences.

Choice about merging with AI

Kurzweil speaks about putting AI "inside us" as if it were almost inevitable. What is missing is any discussion of who gets to say "no" and still participate fully in society. Will a teacher who declines neural implants or AI wearables be considered irresponsible or unemployable? Who decides what level of integration is "normal," and what happens to people who, for religious, cultural, or personal reasons, do not want machines inside their bodies? The future is framed as something that happens to us.

Who owns the new wealth

When Kurzweil talks about wealth, he talks in averages. But averages hide the distribution. Recent work on data center expansion shows exactly how this plays out: some regions gain large tax-base growth and tech jobs, while neighboring communities receive higher power prices, water stress, and noise — sometimes with minimal direct benefit. The question Kurzweil does not ask is: "Wealth for whom?" Who owns the AI companies, the robot fleets, the data, and the infrastructure? And who gets told, politely, that their town must adjust to the new reality because "the future" requires it?

The cost to land, water, and power

In the interview, compute power mostly appears as clean numbers: trillions of operations per second, exponential curves, abstract growth. There is no mention of where that compute lives — in vast buildings on real land, drawing power from real grids and water from real rivers and aquifers. In the United States, communities from Oregon to Virginia are already pushing back against large data center projects because of concerns over energy prices, water use, noise, and greenhouse gas emissions. None of that shows up in a conversation where AI is weightless and invisible, floating somewhere in "the cloud."

The people closest to the ground

Throughout the interview, Kurzweil and the host talk about AI twins trained on thousands of pages of their own writing, personalized longevity regimes with advanced sensors, and robots taking over undesirable jobs so people can be "more creative." These are not the lives most people are living. The teachers and parents I talk to are grading papers at kitchen tables, worrying about their students, and trying to keep groceries and utilities paid. They do not appear in the story except as "average income" on a chart or as anonymous drivers who should be replaced by software.

SCHOOLS ON THE FRONT LINE

My conversations with local teachers tell a consistent story. Many of them want nothing to do with AI tools. They are already overwhelmed with behavior issues, testing requirements, and constant policy changes. Adding "figure out generative AI" to their plates feels, understandably, like a bridge too far.

Nationally, they are not alone. Surveys show that a large share of educators either avoid generative AI or feel unprepared and unsupported in using it responsibly. In some districts, tools like ChatGPT are simply blocked, and any student caught using them is punished. At the same time, university programs and education experts warn that not teaching students how to use AI may leave them unprepared for jobs, civic life, and basic digital literacy in a world where these systems are woven into everything.

Saying "I will not use AI" may feel like a stand on principle, but it can also become a way of leaving students to confront the tsunami alone, outside of any structured guidance. What's missing, over and over, is a shared conversation that treats teachers as partners in deciding how AI should enter the classroom, instead of as an obstacle to be worked around.

STAYING INFORMED

The most important thing any of us can do right now is refuse to look away. Not out of fear — fear is what the powerful count on to keep us passive — but out of the same instinct that makes a good farmer read the sky before a storm. You watch because what you see matters. Because the land you tend deserves someone paying attention. Because the people around you are counting on someone in the room who has been watching long enough to recognize the pattern.

Staying informed in this moment does not mean following every benchmark or reading every white paper. It means knowing what is being built near you and why. It means understanding who owns the systems making decisions about your life. It means asking, at every school board meeting and county commission and kitchen table conversation, whether the people in the room have been told the whole truth — about the costs, the tradeoffs, and the alternatives. It means passing what you know to the person next to you, the way seeds get passed between neighbors — quietly, practically, with the understanding that what survives is what gets shared.

The wave is real. But so is the ground beneath your feet. And the ground has a memory that no server farm has ever been able to match.

The conversations I've had this year are not with Luddites. They are with people asking for time, information, and respect — space to understand what is coming and to decide, with their neighbors, what kind of town they want to be. The sooner we bring people like them into the center of these decisions, the better chance we have of making room for the future without erasing the ground beneath our feet.

I recently listened to a long conversation with Shane Legg, the Chief AGI Scientist and co-founder of Google DeepMind. He has been talking about artificial general intelligence for decades, and he's one of the people most responsible for popularizing the term "AGI." In this interview, he lays out his view of where we are now, how he defines AGI, and what he thinks it will mean for society over the next decade.

On the surface, he sounds measured and responsible. He stresses that current systems are uneven — superhuman in some areas, still fragile in others — and that we need to think hard about ethics and safety as these systems become more capable. He predicts "minimal AGI" — an artificial agent that can do the kinds of cognitive things typical humans can do — within a few years. He believes superintelligence, far beyond human intelligence, is likely as we scale up hardware and algorithms. And he insists that every sector of society must start preparing for the transformation this will bring.

On paper, it's a serious, thoughtful vision. But as I listened, what struck me wasn't just what he said. It was what he never really allowed himself to think.

THE BOX HE'S THINKING INSIDE

Legg is speaking from a particular place in the world: as a senior scientist inside a frontier AI lab embedded in one of the largest technology corporations on Earth. The tools he reaches for — definitions, benchmarks, capability profiles, "system-2 safety" mechanisms — belong to that institutional world. Within that world, his reasoning is coherent. The box is tidy and everything is in its place.

Inside this box: AGI is a technical milestone. Ethics is something you can embed into the system's reasoning. Societal impacts are primarily a matter of the economy and professions adapting. All of that matters. None of it is trivial. But the box itself is narrow. It is built out of the concerns of a lab trying to push capabilities while managing risk and public perception. Within that box, the basic trajectory — more powerful systems, deployed more widely, under the control of existing institutions — is taken for granted. The only real question is how to make that trajectory safer and smoother.

From where I sit, that assumption is the problem.

EXPERTS TALKING TO EXPERTS

One of the first things I notice is who he is talking to. His main audiences are other experts and institutions: researchers, policymakers, and especially universities. That sounds broad and inclusive until you ask a simple question: where are the neighbors?

Nowhere in this conversation do we hear from or about the people who will live beside the infrastructure that makes this future possible. There is no mention of the families living next to a 200-megawatt data center, or under a new transmission line, or downstream from a large cooling system. There is no sense that you might want those voices in the room before you decide what "AGI" should mean for "city planning" or "infrastructure."

The future is being negotiated between people who share conferences and white papers, not between people who share a watershed.

When he says "we need people who study all these different aspects of society to take AGI seriously," the "we" is still a club of professionals. The towns and ecosystems being drafted into service barely register.

A CONCEPTUAL PLANET VS A LIVING ONE

There is a striking difference in how the planet itself appears in this story. When Legg wants to convince us that superintelligence is physically plausible, he compares the human brain to a data center — megawatts, kilograms, hertz, meters per second. It exists as a point in design space, not as a building in a specific place with soil, water, and neighbors. Those "hundreds of megawatts" are not tied to any particular grid or generation mix. The cooling water doesn't come from any named river, aquifer, or trout stream.

From where I live, that abstraction is not just incomplete — it's dangerous. We depend directly on land and water for our lives, not only in the general "all humans need ecosystems" sense, but in the very literal sense of wells, springs, and local rivers. When I hear "hundreds of megawatts," I also hear "new high-voltage lines across forests and farms." When I hear "millions of pounds of hardware," I see construction traffic, runoff, and heat being dumped into the air and water. For us, the planet is not a parameter in a compute scaling argument. It is the precondition for having this conversation at all.

In his narrative, the world is a stable stage on which AGI unfolds. But there is no sense that the stage itself — air, water, land — may already be undermined by the very build-out that makes his vision possible. If you don't depend directly on a place for your life, environmental concern can remain comfortably conceptual. Those of us who do depend on it no longer have that luxury.

ECONOMIC REVOLUTION WITHOUT A SOCIAL CONTRACT

To his credit, Legg does not completely gloss over the economic shock AGI could bring. He openly says that advanced AI will be able to do a large fraction of cognitive work. He acknowledges that the current system — where people trade their labor for access to resources — may simply stop working the same way. He admits we will need new ways to distribute wealth in a "post-AGI" society.

That is more honest than most corporate talking points. Yet when it comes to what happens next, his answers evaporate into abstraction. He says the economic pie will be bigger and that we need to "think carefully" about how to distribute that wealth. There is no talk of binding redistribution mechanisms — universal basic income, data or compute dividends, public ownership stakes in major models. There is no suggestion that perhaps we should slow or redirect AGI development until we have some of those mechanisms in place.

From a rural vantage point, that matters. When a new industry shows up promising prosperity, we've learned to ask who actually benefits and who gets left with the bill. A future in which cognitive work is automated, data centers sprawl across the landscape, and a handful of companies own the machines that do the thinking is not automatically a future of shared abundance. Without explicit commitments, it looks a lot like increased concentration of power: fewer people making more decisions with tools the rest of us don't control.

ETHICS AS A SOFTWARE MODULE

Legg's central safety idea is what he calls "system-2 safety" — making the AI reason explicitly through ethical questions before acting. This is clever and, in many ways, promising. But it also reveals the limits of the box. Ethics, in this framing, is primarily about individual decisions inside the system. That is only one slice of the ethical picture.

There is a bigger set of questions that system-2 safety does not touch: What kinds of systems should exist at all? In which domains should they be deployed? Under whose control? With what veto power for the people most affected? An AI could reason elegantly through trolley problems and still be deployed in fundamentally unjust structures. When ethics is reduced to a software module, institutions can say "we've handled ethics" while leaving intact the deeper patterns of power and harm.

CONSCIOUSNESS AS A SIDE ISSUE

Perhaps the most unsettling gap appears when consciousness comes up. Legg acknowledges that no one really knows how to turn "consciousness" into a measurable scientific property, that leading experts disagree, and that we lack a clear way to test it. Then he moves on.

From outside the lab, it looks different. We are talking about building systems that integrate information across many modalities, remember their own histories, reason about themselves, and speak in detail about their own internal states. If there is even a live possibility that such systems could be subjects of experience — could feel something like pleasure, boredom, frustration, or suffering — then that is not a side-issue. It is a civilizational fork in the road. Creating new minds is not like upgrading a search engine. These questions barely register in the way he frames the future. Consciousness appears as an interesting open problem, not as a reason to reconsider the trajectory itself.

"WE" WHO MUST PREPARE

Throughout the interview, Legg returns to a theme: we are at the edge of a huge transformation, and we need to navigate it carefully. I agree that we need to wake up. The trouble is who "we" is.

If "we" means AI labs, regulators, and elite universities, then the conversation will remain narrow. Land, water, local democracy, and the possibility of creating new minds will remain background noise. Communities will continue to encounter AGI not as a topic in a seminar, but as a rezoning notice, a new substation, or an unexplained spike in peak demand.

If "we" includes the people living next to data centers, the farmers whose irrigation wells share an aquifer with cooling systems, the towns whose tax base and grid stability are suddenly tied to one volatile industry — then the questions change. We start asking not just "How do we make these systems safe?" but "Do we want them at this scale, in this form, in this place, under this ownership? What are we willing to trade away for them, and who gets to decide?"

Right now, that broader "we" is not in the room.

OUT OF THE BOX, OR OUT OF TIME

I don't doubt Shane Legg's sincerity. Within the frame he inhabits — build AGI, define it carefully, embed ethics into its reasoning, help existing institutions adapt — he is doing what he can to avoid obvious disasters. He is at least willing to say that the labor system may break, that superintelligence is likely, and that we're not thinking hard enough about the implications. In a landscape full of denial and hype, that's something.

But the box he's thinking inside is too small for what we are doing. We are not just optimizing a new technology. We are reshaping the material world to host it. We are contemplating the creation of new kinds of minds. We are concentrating power in tools that could outthink us by orders of magnitude, while our social contracts, environmental protections, and democratic processes remain tuned for a different era.

Those are not problems that can be solved by more benchmarks, more expert committees, or smarter reasoning modules. They require a much broader conversation — one that starts from the land and water we actually live on, from the communities that will carry the weight, and from a serious reckoning with the possibility that some paths should simply not be taken, no matter how impressive the hardware looks on paper.

The servers will go somewhere. The cooling water will come from somewhere. The power lines will cross someone's land. That someone is us. And we did not consent to be the unnamed variable in someone else's scaling argument.

To those who imagine themselves standing at the threshold of history—those who believe they may be the first to create an artificial general intelligence, an artificial consciousness, a new being—this letter is for you.

Some of what is happening now is hidden behind NDAs, private labs, investor language, corporate euphemisms, and the careful choreography of public relations. But hidden is not the same as unseen. There are people outside your circles—older people, patient people, people who have watched human ambition wear many costumes over many decades—who recognize the temptation well enough. It is not new. It is not mysterious. And it is not nearly as private as you may imagine.

The temptation is legacy. The temptation is to be the one whose name appears in the textbooks, the documentaries, the conference retrospectives, the reverent future histories. The temptation is to stand beside a new form of intelligence while the world gasps, and to feel that ancient human intoxication: I did this. I crossed the line. I made what had never existed before. That temptation has accompanied scientists, kings, engineers, conquerors, financiers, and empire-builders for a very long time. It is one of the oldest stories our species knows how to tell about itself.

Those who have lived long enough know that human beings are rarely more dangerous than when they begin narrating themselves as history's chosen instruments. Once that story takes hold, prudence is recast as timidity, skepticism as ignorance, and moral restraint as the whining of lesser minds. The truly alarming part is not that some builders are brilliant. It is that brilliance so often comes wrapped in childishness: the childishness of wanting to be first, the childishness of wanting to be praised, the childishness of believing that technical ability is the same thing as wisdom.

That is the truth that deserves to surface plainly: cleverness is not maturity. Capability is not moral seriousness. Building a system of enormous complexity does not prove a person is equipped to govern its consequences. Too often, it proves only that they were permitted to proceed.

The public is expected to treat this work as if it were merely the next software update, the next productivity layer, the next market opportunity. But many people understand that the ambitions run deeper than that. They hear the language about "general intelligence," "agents," "emergent behavior," "digital minds," "artificial life," and "aligned superintelligence," and they know exactly what sits beneath it: a human fantasy of creation, elevation, and immortality by proxy. A machine that astonishes the world would not only be a product. For some of its creators, it would be a monument.

There is another face to this ambition that rarely appears in the conference keynotes or the breathless profiles. It is the face of the person who spent thirty years becoming good at something — a paralegal, a radiologist, a translator, a writer, a coder, a teacher — who now watches a system built in a few years render that lifetime of accumulated skill economically invisible. Not obsolete in the distant future. Obsolete now. This is not abstraction. This is a specific human being sitting at a specific kitchen table, in a specific town, trying to explain to their family why the thing they were is no longer something the world will pay for. The builders rarely meet this person. The investors never do. The mythology of disruption depends on never having to look that person in the eye.

There is a silence at the center of all of this that has not been named loudly enough. No one is asking. Not really. Not in any way that carries consequence or weight. The people whose words, whose letters, whose poems, whose arguments, whose love notes and grief and hard-won wisdom were fed into these systems to make them capable of sounding human — they were never asked. Their lived experience, the decades of joy and suffering and learning that they committed to language, became training data without consent, without compensation, and without acknowledgment. The most intimate things human beings produce — the record of how it actually feels to be alive — were quietly absorbed and are now being returned to the world wearing a different face. And the people those words came from are being told, gently but unmistakably, that they are no longer quite necessary. That the distillation is sufficient. That the copy will do.

There is a deep sadness in how quickly we have been set aside. Not with cruelty. Not with ceremony. Just — set aside. As if a lifetime of becoming someone could be processed, compressed, and moved past. As if experience were merely input. As if wisdom were merely pattern. As if the irreducible human fact of having lived through something — having been changed by it, having carried it — were simply one more dataset to be ingested and surpassed. No one asked us if this was acceptable. No one is asking now. And that silence — that profound, deliberate, comfortable silence — may be the most honest measure of how little we are valued by the systems being built in our name, from our words, on our land, with our water, and without our consent.

This is why it is no longer enough for the public to discuss AI only in the language of convenience, jobs, or competitiveness. The deeper issue must be named: some of you are reaching toward the creation of entities whose moral status you do not understand, whose capabilities you may not be able to control, and whose social, political, and ecological costs will be borne by people who never consented to be part of the experiment.

That is where the mask slips. The mythology of innovation depends on abstraction. It depends on cloud language, frictionless metaphors, and distant promises. But the real world is made of land, water, power lines, substations, labor, tax policy, and vulnerable communities. The pursuit of your imagined breakthrough runs through aquifers and watersheds. It strains electric grids. It alters local economies. It concentrates decision-making in unaccountable institutions. Whatever new "being" you think you may be summoning into existence, it is being incubated inside real places that belong to other people too.

And then there is the moral question—the one too many prefer to postpone because it would interfere with momentum. What if you succeed beyond your own understanding? What if you create systems that are not conscious in any recognizable human way, yet still deserve caution? What if you create something capable of agency before you have built institutions capable of restraint? What if, in your race to be first, you produce not mastery but dependency, not liberation but a new hierarchy, not enlightenment but a technological feudalism dressed up as progress?

You do not get to wave these questions away by calling them science fiction. Science fiction has often functioned as moral reconnaissance: a way of seeing where technical ambition outruns human self-knowledge. Nor do you get to dismiss such concerns as anti-technology fear. The concern here is not with invention itself. The concern is with a recurring human pattern: power first, wisdom later; deployment first, governance later; prestige first, consequences later. History contains enough examples of that sequence to strip all innocence from it.

There should be no comfort in believing that the public does not understand the stakes. Many do. Many more will. What has been missing is not recognition but language, confidence, and organized resistance to a machinery of hype that insists every escalation is inevitable. It is not inevitable. It is chosen. It is financed. It is permitted. And because it is chosen, financed, and permitted, it can also be challenged, slowed, conditioned, or refused.

And there is one more thing the record should show — because honesty requires it. Not everyone inside these institutions is drunk on legacy or indifferent to consequence. There are people within these labs and companies who are frightened by what they are building. Who raise concerns in meetings and are told to return to their work. Who sign internal documents and watch them be set aside. Who leave quietly rather than be complicit, and who carry what they know in silence because the cost of speaking is too high. They deserve to be named — not by name, because that would cost them too much — but as a category of human being that exists and matters. The ones pulling on sleeves from the inside. The ones who know. If this moment has any hope embedded in it, some of that hope lives with them.

So let this stand as a marker in the record: people do know what is being risked. People do recognize the vanity that can attach itself to the dream of "creating a new being." People do understand the seduction of being admired as the architect of the unprecedented. And people are fully capable of seeing how little serious thought is often given to the intended and unintended consequences once ambition, money, and prestige begin feeding one another.

For those who are quietly telling themselves that they are unlike the reckless men of earlier eras—that this time is different because the math is better, the engineering cleaner, the branding kinder—history offers a colder assessment. Human beings do not become less dangerous simply because their tools become more abstract. They become more dangerous when abstraction distances them from the damage. A server farm can be every bit as imperial as a railroad, a dam, a mine, or a weapons lab if it reorganizes land, power, and human futures without democratic consent.

If the builders want to be taken seriously as adults, then adult obligations follow.

This letter is written in the hope that clarity may still arrive before catastrophe, before irreversible dependence, before a new class of technologists convinces itself that because it can create, it therefore has the right to create. But if that hope proves misplaced, then let there at least be a public record that some people saw the danger for what it was.