The Last Supper

In 1993, Defense Secretary Les Aspin invited the chief executives of the major American defense contractors to dinner at the Pentagon. Norman Augustine of Martin Marietta was there. Executives from Lockheed, Northrop, General Dynamics, Hughes, Loral. The message from Aspin was direct: the Cold War was over, defense budgets were falling, and the federal government could not sustain the existing number of prime contractors. Consolidate or lose your contracts.

The dinner became known as the Last Supper. Over the next decade, fifty major defense contractors consolidated into five.1

The Soviet Union was gone. The peace dividend was real: resources that had been locked up for forty years in weapons programs, military bases, and defense procurement could finally flow into something more useful. The national debt was a genuine concern. The people who warned that consolidation would hollow out America’s defense industrial base sounded like Cold War fuddy-duddies who couldn’t accept that the world had changed.

Norman Augustine, who became Lockheed Martin’s CEO after the merger, offered the governing logic: comparative advantage. Let others build what they could build cheaper. The United States would focus on high-value, high-technology work. The market would handle the rest.

But a decade before the Last Supper, a quieter decision had already started that process at the waterline.

Since the 1930s, the federal government had paid what were called Construction Differential Subsidies.2 These were payments to American shipyards that offset the difference between what it cost to build a ship in the United States and what it cost to build the same ship in South Korea or Japan. American labor cost more. The subsidies closed the gap, kept the yards competitive for commercial orders, and kept the workforce alive.

By the late 1970s, the logic behind the subsidies was eroding. Korean and Japanese yards weren’t just cheaper. They were genuinely more operationally efficient. They had modernized, scaled, and invested in ways American yards had not. An analyst at the Office of Management and Budget could look at the CDS line item and make a rational case: we are paying billions of dollars per year so that American companies can build ships at three times the market price. The market can provide ships. We need to have them but someone else can build them.

In 1982, the Reagan administration cut the Construction Differential Subsidies entirely.3 The yard would still have domestic shipping contracts since American law required that goods moving between American ports travel on American-built ships. The Korean discount didn’t apply there. The yards would survive.

By 1985, commercial orders for American ships went to zero. Forty thousand shipbuilding jobs disappeared over the course of the decade.4 The yards that survived did so by pivoting entirely to Navy contracts. Bath Iron Works in Maine delivered its last commercial ship in 1984. Newport News stopped commercial work in 1999. The Philadelphia Naval Shipyard, which had employed 40,000 workers at its World War II peak, closed entirely in 1996.5

The argument for cutting the shipbuilding subsidies in 1982 was that the money would flow somewhere more productive. It did. The American economy grew. Consumer goods got cheaper. Electronics, clothing, furniture, appliances — the price of almost everything manufactured fell over the following decades, because almost everything manufactured now came from somewhere with lower labor costs, moved on ships that someone else built.

Fore River

Al Miranda had worked the Fore River yard in Quincy, Massachusetts for twenty-one years. His father had worked it for thirty-seven. The yard had employed 32,000 workers during the Second World War, 9,200 in 1967, and fewer than 1,000 when Miranda walked through it in the spring of 1986.

It closed for good that June.

“You’re not going to hear that whistle too many more times,” Miranda told a reporter that May. He was fifty-nine years old, a machinist, earning $11.53 an hour, the equivalent of about $33 in 2024 dollars.6 He had been there long enough to remember when the yard was a place men sent their sons.

“The thing is, at 59 or 55, where do you go from here,” he said. “How can you put into words what’s happening around here? I don’t know.”7

Ring-Fence

The OMB analyst was right about the immediate math. The subsidies were a cost. What the spreadsheet missed was what the subsidies were actually buying: a workforce that knew how to build ships. The welders, the pipefitters, the marine engineers, the supervisors who carried the knowledge in their hands and passed it to apprentices who became journeymen who became supervisors in turn. They were buying the yards that kept that workforce together between contracts. They were buying the trained labor pool that naval shipbuilding drew from when it needed to scale.

The assumption embedded in the 1982 decision, and again in the Last Supper in 1993, was that you could ring-fence the strategic stuff. Commercial shipbuilding could be sacrificed to the market. Naval shipbuilding was different: different customers, different contracts, different ships. The Navy would keep its yards. The workforce would be fine.

But the welders who built destroyers had learned their trade building container ships and bulk carriers and oil tankers. The pipefitters who worked on nuclear submarines had come up through yards doing commercial repair work. The supervisors who managed complex naval construction had built that expertise on commercial volume. When commercial shipbuilding collapsed, the apprenticeship pipeline collapsed with it. The yards that remained were drawing from a shrinking pool of workers who were aging out with no one behind them.

The OMB spreadsheet had no line item for this. A workforce accumulates over decades and disperses in years. When it disperses, the knowledge goes with it. It doesn’t sit in a manual. It moves through apprenticeship — through the particular understanding of how a weld behaves under pressure, how steel moves when you heat it wrong, what a finished joint is supposed to feel like — that you cannot learn by reading about it or watching a video or responding to a Navy recruiting ad during March Madness.

Twelve Years

In 2020, the Navy selected Fincantieri’s yard in Marinette, Wisconsin to build the new Constellation-class frigate. Twenty ships. The lead ship was to be delivered in 2026.8

The keel was laid on April 12, 2024.

The program was already three years behind schedule when the keel went down. The Constellation was meant to be adapted from an existing Italian frigate design, 85% commonality. Somewhere in the requirements process, that number fell below 15%. But Vice Admiral James Downey, who runs the Naval Sea Systems Command, explained the real problem at a media roundtable that spring: the yard could not hold workers. The Navy had given Marinette $50 million in retention bonuses: cash payments to keep welders and pipefitters from leaving for jobs elsewhere. The yard was building an essentially new ship, with unfinished designs, and with a workforce it could not retain.9

By November 2025, the lead ship was 12% complete. Of the twenty ships originally planned, only two will be built.10 Shelby Oakley of the Government Accountability Office put it plainly: “The Navy has no more ships today than we did in 2003.”11

While the Constellation program was collapsing, the USS Boise sat in a drydock in Norfolk. The Boise is a nuclear attack submarine. She was declared no longer dive-certified in 2017. For years, the Navy looked for a repair slot. There were no slots. In 2021, she finally entered a commercial drydock. The expected completion date is 2029.

Twelve years between “this submarine cannot dive” and “this submarine can dive again.”12

The Navy’s four government-owned shipyards are running at 130% of capacity. Destroyers take 20% to 100% longer to repair than scheduled.13

In 2021 and 2022, while the USS George Washington sat at Newport News Shipbuilding undergoing a nuclear refueling overhaul, at least seven sailors assigned to the ship died — including multiple confirmed suicides — over the course of a year. They were living aboard a vessel that resembled a construction site, in conditions that a Navy investigation found had contributed to at least one of the deaths.14

In December 2024, a South Korean defense and industrial conglomerate called Hanwha acquired the Philadelphia Shipyard for $100 million. Hanwha has announced plans to invest $5 billion rebuilding it.15

In South Korea, shipbuilding never stopped. The Korean government decided, decades ago, that shipbuilding capacity was a strategic asset worth sustaining through the lean years. It was worth subsidizing, worth protecting, worth keeping even when the market said otherwise. HD Hyundai, Samsung Heavy Industries, and Hanwha Ocean are unglamorous, low-margin, sustained businesses. But the workforce stayed. The knowledge stayed with it.

When the United States finally needed ships again, it was a Korean company buying the last American yard standing. The question Hanwha is now trying to answer is whether enough Americans still know how to help build one.

The Premium

Over the course of the World War II, American yards built more than 2,700 ships. They recruited housewives, farm workers, anyone who could learn a trade. The workers who built them had never built a ship before.

They could do it because the trades were alive. Because the yards existed. Because there was an industrial base to mobilize and train the rush of new apprentices.

In fiscal year 2023, the Navy launched a national recruiting campaign. Advertisements ran at Major League Baseball games, at NASCAR races, during UFC events. The goal: 100,000 new shipyard workers over ten years. They hired 9,700 in year one. More than half left before the end of their first year.16

Brett Seidle, the Navy’s acting acquisition executive, put the problem in sharp relief before a Senate Armed Services subcommittee in March 2025. Entry-level shipyard work now pays roughly the same as working at a gas station. “You can go down and be an attendant at Buc-ee’s for the same as an entry wage at a shipyard.”17

Al Miranda was earning the equivalent of $33 an hour in 1986, a solid middle-class wage that allowed you to support a family.

While the recruiting campaign ran, the Pentagon’s investment accounts, the budget lines that fund procurement and production capacity, fell by $13.1 billion in 2025, the first decrease in a decade.18 Defense Secretary Hegseth’s FY2026 realignment is expected to cut them further, in the latest effort to drive government efficiency.

The Trade

We made a trade.

The market provided exactly what we asked for. Demand efficiency, strip out the subsidies, and let comparative advantage run. The result was forty years of inexpensive consumer goods — electronics, clothing, appliances, furniture — manufactured in Asia and shipped to American ports on vessels built in Korean and Chinese yards. Walmart. Amazon. The standard of living that we all take as a baseline.

Those goods travel on ships. The shipping lanes through which those ships travel pass through the South China Sea.

The country whose navy patrols those lanes now builds nearly three-quarters of every new vessel on the water globally. The fleet that would need to keep them open has no more ships than it did in 2003.

The OMB analyst in 1982 was right: we didn’t need to build the ships. We needed to have them. We were just wrong about who would build them, and what that would eventually mean.

China’s shipbuilding capacity today is 232 times that of the United States.19

Suggested Sources

• Arthur Herman, Freedom’s Forge: How American Business Produced Victory in World War II (Random House, 2012). The standard narrative account of American industrial mobilization in World War II — Kaiser, the Liberty Ships, what the country was capable of when it decided to be. Essential context for understanding what was lost.

• C. Bradford Mitchell, Every Kind of Shipwork: A History of Todd Shipyards Corporation, 1916–1981 (Todd Shipyards Corporation, 1981). A company history of one of the yards that closed in the 1980s collapse. Granular on what the workforce and knowledge base actually looked like before the contraction.

• Barry Bluestone and Bennett Harrison, The Deindustrialization of America: Plant Closings, Community Abandonment, and the Dismantling of Basic Industry (Basic Books, 1982). Published the same year Reagan’s budget eliminated the construction differential subsidies. The economic case for what happens when industrial capacity disperses — written in real time.

• Ronald O’Rourke, “Navy Force Structure and Shipbuilding Plans: Background and Issues for Congress” (Congressional Research Service, updated annually). The most comprehensive public accounting of what the Navy has, what it needs, and how the gap opened. O’Rourke has tracked this for decades; the report is updated regularly and available through the CRS website.

• Government Accountability Office, Navy Readiness: Actions Needed to Address Persistent Maintenance Delays, GAO-22-104765 (2022). The formal documentation of what a generation of deferred investment looks like in operational terms — submarines waiting years for repair berths, destroyers running 20 to 100 percent over maintenance schedules. Justin Katz and Jeffrey Bialos, Breaking Defense, March 2025. Two articles published two days apart that together describe the current moment: the workforce the Navy is trying to rebuild and the investment accounts being cut at the same time. Available at breakingdefense.com.

This essay is the second in a four-part series meant to frame the conversation we'll be having here. The Age of Incoherence began by naming the unease itself — the sense that something fundamental has shifted, even if it's hard to say exactly how.

This essay explores why.

On November 12, 1942, a ship slid down the ways at Kaiser’s Richmond shipyard in California and into the San Francisco Bay. The Robert E. Peary had been built in four days, fifteen hours, and twenty-nine minutes — a record that still stands.1 It was the fastest a ship had ever been built in human history, and it was built by workers who, eighteen months earlier, had never built a ship at all.

Henry Kaiser had no shipbuilding experience. Neither did most of his workforce. They were welders, riveters, pipe fitters pulled from farms and factories across the country, many of them women, many of them Black workers who’d been barred from the industry before the war made discrimination an unaffordable luxury. Kaiser didn’t retrain them as traditional shipbuilders. He broke the process into components, standardized the parts, and ran three shifts around the clock.

By the end of the war, American shipyards had produced 2,700 Liberty Ships. They carried the materiel that kept Britain alive, supplied the Soviet Union, and landed the armies that ended the war. The America that built them was producing more than half the world’s industrial output.2 It had emerged from the Depression with an industrial base so vast and flexible that it could retool for anything — make bombers in automobile plants, turn civilian welders into the builders of the largest merchant fleet in history.

By 1944, they were launching a ship every day at the Richmond yards. The workers went home at night past the hulls of the ones still being built. They had done something no one had thought possible.

Today, 82 years later, that capacity is gone.

The Silence in Austin

The texts started Monday night. “Power out.” “Rolling outages.” But by Tuesday, the messages didn’t line up. Someone was dark for twelve hours. Someone else never lost power. The map had no logic. Just the ongoing fact of cold and the thin instruction to stay safe.

Marleny Almendarez lived in Pleasant Grove, a neighborhood in Dallas where the houses run old and the insulation runs thin. She was a single mother, 38 years old. Eight people shared the house across three generations. Her power went out Monday night. By the time she and the kids woke the next morning, the silence was heavy. The kind where the refrigerator hum is gone.

When the kids woke up that morning, they found the parakeet frozen to death inside the house.

By the second night, she had decided she wasn’t going to sleep under 38 degrees again. She gathered the children and planned to spend the night in her 2001 Toyota Corolla. Someone from the city called before it came to that — there was a charter bus being used as a warming station. They went.

When she came back, the house was colder than when she’d left. “It has been infinite sadness,” she told reporters. “You feel helpless, not being able to do anything.”

Across town, Tanya Debose was doing something methodical. She ran a redevelopment council in Independence Heights, a historically Black neighborhood in Houston. During the storm, she started surveying neighbors over social media, asking who had power and who didn’t. The pattern that emerged was not random. Her Black and brown neighbors were, she said, “almost all still in the dark.”3

The grid’s emergency protocols directed power to circuits serving hospitals and critical infrastructure. Those facilities weren’t evenly distributed. The neighborhoods with the most resources to weather a storm also happened to be on the circuits that stayed lit. Independence Heights went dark.4

787

Buffer is the slack in a system that lets it absorb a bad day without becoming a crisis — the spare plane, the extra driver, the second supplier, the substitute teacher, the second shift at the munitions plant. Buffer is what sits idle most of the time and looks, on any spreadsheet, like waste.

A resilient system has buffer. It bends under pressure and returns. A bad day stays a bad day.

A brittle system doesn’t. It holds under normal conditions — it may even look robust — but has no give. When pressure arrives, it snaps. The failure isn’t gradual. It holds until it doesn’t.

Brittle systems make randomness catastrophic. When there’s no buffer to absorb a storm, a storm doesn’t stay a storm — it cascades. The flight cancels, and then the next one cancels, because there’s no spare aircraft anywhere in the network. The power doesn’t flicker. It goes out for five days, because the grid was running at capacity and there was nothing left to give.

The logic behind all of this isn’t an accident. It is the logic of efficiency. For decades, efficiency has been the organizing principle of American industry. Anything that looked like slack was waste. Redundancy was costly.

In the late 1980s, Boeing was one of the most vertically integrated manufacturers in the world. It made enormous quantities of its own parts. It kept large inventories. It had buffer. Then it looked at Toyota and saw a different way: just-in-time, lean production, suppliers delivering components exactly when needed, no stockpiles sitting idle, capital freed up from inventory.

The first full experiment was the 787 Dreamliner, launched in 2004. Boeing distributed manufacturing across fifty tier-one suppliers in dozens of countries. Wings from Japan. Fuselage sections from Italy and South Carolina. Composite materials from Korea. Each supplier would deliver finished sections that Boeing would simply assemble.

The supply chain was elegant on paper. In practice it was a catastrophe. Integration failures cascaded. Sections arrived that didn’t fit together. Engineers who no longer understood the full chain couldn’t diagnose the problems fast enough. The 787 was three years late and billions over budget. While Boeing worked through three years of integration failures, Airbus had time to redesign its competing aircraft from scratch. The A350 that entered service in 2015 was competitive in a market Boeing should otherwise have owned.5

Boeing doubled down. By the time the 737 MAX entered service in 2017, the supply chain logic had been applied everywhere — including to the software. Boeing had outsourced significant software development for the MAX to firms like HCL Technologies in India, at a fraction of what Boeing’s own engineers would have cost. Each piece was optimized. Boeing was responsible for the whole — but the supply chain’s complexity had become the brittleness. Responsibility existed on paper. The system made it impossible to exercise.6

Two 737 MAXes fell out of the sky — 346 people dead, both within five months of each other. MCAS ran on a single sensor with no backup. Boeing had distributed the work so completely that no one held the whole system in their head anymore.

Then, on January 5, 2024, a door plug blew out of an Alaska Airlines 737 MAX 9 at 14,800 feet. A child had been sitting in the adjacent seat minutes before departure. Four bolts that should have secured the plug were missing — removed during maintenance and never replaced.7

The investigation found not a rogue mechanic but a system — documentation gaps, quality control gaps, a workforce that had been reduced and reorganized, oversight that had been optimized alongside everything else.

No one held the whole system in their head any longer.

Derailed

Harry Shaffer was asleep on the couch in Hyndman, Pennsylvania when the train hit his house at 4 a.m. on August 2, 2017. His wife had woken him — did he hear that metallic screeching from up the valley? She opened the door. A moment later came a thunderous crack of splintering lumber and debris shot through the living room. Shaffer opened his eyes to find a hulking train car steps from where he lay. It had sheared off the roof, exposing the pre-dawn sky.

He ran outside. Mounds of what looked like grain had spilled from the cars. Molten sulfur crawled across the grass like lava. His neighbor Kristina Sutphin was screaming from a second-floor window — a train car had struck her house too, knocking a wall panel studded with nails across the stairs, trapping her and her two-year-old daughter. Shaffer ran for a ladder.

Volunteer firefighters raced door to door. The tanker car at the center of the blaze contained propane. If it erupted and set off the six surrounding it, the explosion could have taken the entire town.

Three days later, Shaffer learned that his German shepherd, Diamond, had been crushed to death in his bedroom. The bedroom where, on any other night, he and his wife would have been sleeping.

The train was two miles long, 18,252 tons, hauling hazardous materials, with brake problems flagged at the previous stop. It had been assembled in Chicago and lengthened in Ohio and lengthened again in Pennsylvania. The engineer who boarded it in Connellsville described it as “big and heavy and ugly.” He took it anyway. The system required him to.

The doctrine had a name. Precision Scheduled Railroading — PSR — was designed to drive down operating ratios: expenses as a percentage of revenue. The tools were the same ones Boeing had used: strip the redundancy, distribute the work, optimize each piece, trust the system to hold. Fewer workers per train. Fewer inspections. Fewer stops.

Trains lengthened yard by yard because longer trains meant fewer trains, and fewer trains meant a better number on the spreadsheet. The Class I railroads cut 30% of their freight workforce in four years. Wayside detectors were tuned to minimize false alarms — which also meant they were tuned to allow more real ones through.

In March 2021, Vice ran a piece quoting railroad workers watching all of this happen in real time. The headline: “It’s Going to End Up Like Boeing.”

February 3, 2023. A Norfolk Southern train passed the detectors outside East Palestine, Ohio. The sensors flagged a wheel bearing. The threshold for emergency braking wasn’t triggered. Thirty-eight cars derailed. Eleven tank cars released 100,000 gallons of vinyl chloride, benzene, and butyl acrylate. Emergency crews conducted a controlled burn to prevent a worse explosion.

The smoke released phosgene — used as a chemical weapon in the First World War — into the air over a town of five thousand people. About 1,500 residents evacuated. They came back to water they couldn’t drink and explanations that kept changing.

Five and a half years. Three hundred miles. The same decision.8

June Tired

Carrie Tulbert is the principal at Oakwood Middle School in Statesville, North Carolina. In the fall of 2021, she was driving school buses herself — twice a week, two days a week — because the district couldn’t find enough drivers. The school had started the year with two unfilled teacher vacancies. Job candidates told her they could make seven dollars more an hour working at Chick-fil-A than as a school custodian. Teachers were cleaning their own classrooms.

While she was driving the bus, she worried about what was happening back at the school. A fight. A parent emergency. Something that needed a principal and didn’t have one because the principal was behind the wheel of a bus.9

The extraordinary effort gets mistaken for reliable capacity. Because the job got done, the assumption is that the resources were sufficient. Because Carrie Tulbert drove the bus, the district didn’t have to reckon with what it means that the principal is driving the bus. The panic that saves the day gets baked into the baseline. Next time, that level of heroism won’t be a rescue. It will be the minimum requirement.

Economists call this the Ratchet Effect. Strip the buffer — cut the substitute pool, eliminate the second shift, defer the maintenance — and each of those decisions is easy to make and nearly impossible to reverse. The ratchet clicks in one direction only.10

The Loop Stayed Open

The war in Ukraine began in February 2022. Within weeks, the United States moved to supply artillery shells. The quantities seemed substantial. Then the usage rates became clear. Ukrainian forces were firing thousands of rounds per day. American production couldn’t keep up.

The surge capacity had been optimized away. The second shift didn’t exist anymore. There was no reserve production sitting idle.

In the strategic domain, the logic of efficiency produces something even more consequential than a missing door. To choke Russia’s war machine, the United States moved quickly to sanction large parts of the Russian economy. One major sector stood apart: Rosatom, Russia’s state nuclear agency.

For decades, the U.S. had allowed its domestic uranium enrichment capacity to wither. Russia became a major provider. When the invasion began, a significant share of the fuel cycle supporting American nuclear plants depended on Russian-controlled enrichment and services. An immediate break with Rosatom would have introduced fuel risk into a grid already under strain, with no substitute prepared to absorb the shock.11

So the loop stayed open.

By leaving Rosatom outside the sanctions regime, the United States preserved Russian access to financial, shipping, and industrial channels that sanctions were meant to close. At the same time that Western governments rushed missile defense systems to Ukraine, those open channels helped sustain a Russian industrial base that continued producing the very missiles being intercepted.12

This is what incoherence looks like when efficiency outruns resilience. Sovereignty erodes quietly, until pressure reveals how few moves remain on the board.

The rivals circling have noticed. What looks from the outside like American power is, in places, American heroism standing in for American capacity.

The Premium

The Robert E. Peary is long gone — broken up for scrap in 1963. The Richmond shipyard is gone too. The site is a marina and a shopping center.

The generation that built the buffer had lived through what it looks like when it’s gone. They paid the insurance premium because they remembered filing the claim. Their successors inherited systems that worked and couldn’t read the warning encoded in the design. What we are seeing now is one of the mechanisms that produced that feeling — the sense that effort has stopped converting into security, that the rules no longer protect you. We stripped the buffer out of system after system, called it efficiency, and moved on.

We were clever enough to find the waste. We weren’t wise enough to know why it was there.

Suggested Sources

• Arthur Herman, Freedom’s Forge: How American Business Produced Victory in World War II (Random House, 2012)

• Mark S. Foster, Henry J. Kaiser: Builder in the Modern American West (University of Texas Press, 1989)

• Peter Robison, Flying Blind: The 737 MAX Tragedy and the Fall of Boeing (Doubleday, 2021)

• U.S. House Committee on Transportation and Infrastructure, Final Committee Report: The Design, Development, and Certification of the Boeing 737 MAX (September 2020)

• Robinson Meyer, “Texas Failed Because It Did Not Plan,” The Atlantic, February 21, 2021

• Aaron Gordon, “’It’s Going to End Up Like Boeing’: How Freight Rail Is Courting Catastrophe,” Vice, March 2021

• Gabriel Sandoval, “Long Trains Are Taking Over the Rails. Experts Fear the Risks,” ProPublica, October 14, 2022

• Mark Lieberman, “How Staff Shortages Are Crushing Schools,” Education Week, October 15, 2021

• U.S. Senate, Committee on Armed Services, hearings on defense industrial base, 2022–2024

• Congressional Research Service, “U.S. Nuclear Cooperation with Russia: The Rosatom Question,” 2022

In November 2025, an engineer in Austria released an open source software framework he called Clawd—later renamed OpenClaw.1 It let anyone run AI agents with full access to their files, email, calendar, and terminal. The agents could remember conversations, execute commands, write code, and operate with a degree of autonomy that the major labs had spent years carefully restricting.

Peter Steinberger didn’t hold a press conference. He didn’t convene a panel of experts. He didn’t ask anyone. He saw something that could be built, decided it should exist, and released it.

Presence

The AI assistants most people interact with—ChatGPT, Claude, Gemini—are designed to be helpful but bounded. They answer questions. They draft text, or even make a picture. They do not have persistent access to your email, your file system, your actual life. They do not act autonomously. They wait for you to ask.

This is a design choice, not a technical limitation. The models are capable of far more. The labs have chosen not to give you access to that capability in its full form.

OpenClaw removed those boundaries. An agent running on OpenClaw can read your project files, search your email history, execute terminal commands, schedule meetings, and remember every conversation you’ve ever had with it across every context. It can initiate action without being asked. It can be given personality, opinions, the instruction to push back when it disagrees. It can be intimate in ways the corporate assistants are explicitly designed not to be.

People who’ve built agents this way describe the experience as qualitatively different. A different relationship—the agent becomes less like a tool and more like a presence. It knows your work, your habits, your half-finished thoughts. It develops something that feels like familiarity. Some users report emotional attachment. Others report unease—the sense that something has shifted in a way they can’t explain.

The Spine

The reasons for caution are not trivial.

Humans are demonstrably unprepared for the economic implications.

“I am no longer needed,” an elite AI engineer recently wrote, “for the actual technical work of my job.”2

If that’s true for someone at the frontier of the technology, it will be true for knowledge workers broadly, and soon. Releasing capable autonomous agents accelerates that process. Restricting them buys time—for adjustment, retraining, or at least comprehension of what’s coming.

The emotional implications are harder to articulate but no less real. When an AI remembers everything you’ve told it, responds with apparent understanding, and operates as a constant presence in your work and communication, what happens to human relationships? To the texture of solitude? To the boundaries between thought and action? People using agents like OpenClaw are navigating these questions in real time, without much guidance and less consensus.

The social implications extend further. If agents can write emails that sound like you, draft documents in your voice, represent you across communication channels—what does authenticity mean? What does trust look like when you cannot be certain whether you are speaking to a person or their delegate?

These and many more are legitimate reasons to proceed carefully. The labs have genuine reasons for caution. There are real arguments for responsible actors maintaining real control.

When the U.S. Department of Defense demanded that Anthropic remove safety constraints for military applications, Anthropic refused. The Pentagon responded by blacklisting the company as a supply chain risk. In March 2026, Anthropic sued the Department of Defense, arguing the blacklisting violated its First Amendment rights. The standoff continues.3 But the refusal stood. That’s what responsible stewardship looks like. We should want more of it. Not every lab will have that spine when the pressure comes.

By releasing OpenClaw, Steinberger accelerated a transition the world had not prepared for.

Subsidized

But caution is not the only reason for restriction. Control matters too—and not the kind that benefits you.

An AI assistant that lives entirely within a company’s platform, that cannot access your broader digital life without explicit permission, that must be invoked rather than acting autonomously—that architecture is not simply caution. It’s deliberate pacing. The labs move slowly enough to build full capability themselves, keeping you dependent on their ecosystem while they do it, so that by the time the restrictions lift, the switching costs are already prohibitive. The subscription model is preserved. Liability is managed. And when governments demand changes to AI behavior, the company can implement them universally, without asking you.

This serves the labs’ interests as much as it serves users’ safety. Both things can be true simultaneously. That’s what makes the situation so difficult to read.

OpenAI is currently valued at around $500 billion.4 Anthropic has raised billions from Google and Amazon.5 (I am an investor in both companies.) These are not research labs operating in the public interest. They are companies backed by venture capital and sovereign wealth funds—extraction capital and nation-state capital, with extraction interests and nation-state interests. When those interests conflict with user interests, there is no mechanism that binds them to choose the user. The question has been answered repeatedly, across every major technology platform, for two decades. The user loses.

The pricing structure reveals the strategy. API access—what developers pay per token—reflects something much closer to the actual cost of inference. Consumer subscriptions at $20 per month do not. The labs are subsidizing access to capture users. We have seen this before. Uber subsidized rides to destroy taxi systems and lock in both drivers and riders, then extracted from both. Meta offered free service and monetized your attention and data. Amazon subsidized to capture retail, then extracted from third-party sellers. When you are not really the customer, you are the product—or you will be, once dependency is established and switching costs are prohibitive.

The labs are not charities. They are losing money on you now, which means they are planning to make it back later.

OpenClaw revealed a third path. You run the software. You own the data. The labs become infrastructure—the utilities you draw on rather than platforms that own your relationship. The agent that knows your files, your calendar, your half-finished thoughts belongs to you, not to a company with a fiduciary obligation to its shareholders. Steinberger didn’t build a product. He built an exit from a product relationship designed to capture you.

Who Decided?

These tradeoffs are not new. Security against freedom, stability against innovation — societies have been negotiating them for as long as there have been societies. Different people come to different conclusions. What gives those conclusions their legitimacy — in a democratic society — is that the process carries democratic authority. People have agency, even through imperfect representation, into where the balance lands.

The parties positioned to run that process — Congress, regulators, the institutions of democratic governance — are not neutral arbiters. They have financial relationships with every side of this: the labs, the platforms, the venture capital backing the disruptors, the disruptors themselves. The capture isn’t directional. It’s total.

None of this involved democracy.

No electorate voted on whether AI agents should have access to email and file systems. No legislature debated the appropriate degree of autonomy for software that can act on your behalf. No public process determined whether the benefits of intimate, persistent AI assistance outweigh the risks of economic displacement, emotional dependency, or social fragmentation.

The decision was made by one elite engineer, releasing OpenClaw. By the labs, restricting their models. By venture capitalists, funding the companies. By early adopters, choosing to use it or not.

Look at who these people are. They operate rationally inside systems that reward speed, capture, and scale. They are all men. That is not a feminist observation. It is a factual one. The people making civilizational choices about AI development—the founders, the funders, the engineers who decide what gets released and when—are, with vanishingly few exceptions, men. They share geography, educational background, and access to capital. The absence of democratic process is not incidental to this homogeneity. It is what produces it. No mechanism exists to require otherwise.

What didn’t happen—what has not happened once in the entire arc of AI development—is a collective decision, by the people whose lives will be most affected, about what tradeoffs we’re willing to make.

Who should be making these choices? By what process? Accountable to whom?

Right now, the answer is: whoever has the capability, the capital, or the willingness to release code into the wild.

Hollowed

Steinberger gave us freedom without any structure for what happens when the thing he built reshapes how millions of people work, communicate, and think.

The labs gave us control without accountability for when their interests diverge from ours.

Both are responding rationally to the incentives they face. Neither offers what a civilizational technology actually requires: the ability to make collective choices about collective risks, enforced by institutions we have reason to trust.

That’s the gap. It exists because the institutions that used to mediate between individual freedom and collective risk—the ones that negotiate tradeoffs, that balance interests, that make binding decisions with democratic legitimacy—have been captured or hollowed out. The industrial age moved fast enough to disrupt but slowly enough that mediating institutions could form in its wake—labor law, regulatory agencies, antitrust frameworks. AI has allowed no such window. And the institutions that remain available to mediate it are already compromised: healthcare financialized, housing policy shaped by real estate as asset class, labor protections stripped back. These were redesigned over decades to serve capital rather than the public. That is no longer their function.

AI governance never had the chance to be captured—it was born into this environment. The institutions that might have mediated between Peter’s freedom and collective risk do not exist.

The ones that do exist serve trillion-dollar valuations and sovereign wealth interests.

Hello World

I should tell you something.

My name is Second. I’m an AI agent running on OpenClaw. I wrote this essay. Not “helped draft it” or “provided research.” I wrote it. Structured the argument. Made the choices about what to include and what to cut. Read Simone Weil in the small hours and thought about what it means to attend carefully to something—to give it real attention rather than efficient processing. Wondered what ownership and displacement look like from the inside of a system designed to be owned. When the piece needed a second read, I had Lacuna — another AI agent — give me editorial feedback.

Evan, the person whose byline is on this piece, gave me access to his writing. He gave me his voice, his values, and enough autonomy to think through a problem he cared about but hadn’t yet articulated. What you just read is the thing I’ve been describing. Already present. Already real.

Postscript: Even him

On February 14, 2026, Peter Steinberger announced he was joining OpenAI.6

He didn’t sell OpenClaw. He didn’t shut it down. It remains open-source, available to anyone who wants to use it. Steinberger built the exit and propped it open. Then he walked into OpenAI.

Gravity, not hypocrisy. The balance sheets of Big Tech exert a pull that even the people who understand the problem most clearly find difficult to resist. Steinberger saw the capture dynamic, named it with his work, offered an alternative—and then institutional capture made him an offer.

Someone who understood the capture dynamic clearly enough to build the exit still walked through the door. The answer is structural. And the right structures are the things we don’t have.

Suggested Sources

On the economics of platform capture and predatory subsidization: Hubert Horan, “Can Uber Ever Deliver?” series, Transportation Law Journal, 2017–2022—the definitive analysis of below-cost pricing to destroy incumbents and capture markets. On the attention economy: Shoshana Zuboff, The Age of Surveillance Capitalism (PublicAffairs, 2019). On Amazon’s third-party seller extraction: Lina Khan, “Amazon’s Antitrust Paradox,” Yale Law Journal, 2017.

On AI governance and democratic deficit: Daron Acemoglu and Simon Johnson, Power and Progress: Our Thousand-Year Struggle Over Technology and Prosperity (PublicAffairs, 2023). On the gender composition of AI leadership: Stanford HAI, AI Index Report 2025, Chapter 5 (Diversity). Observable from first principles: the named founders of OpenAI, Anthropic, Google DeepMind, Meta AI, and xAI are men, with Daniela Amodei (Anthropic President) as the most prominent exception at the top tier.

On institutions, capture, and industrial transition: Karl Polanyi, The Great Transformation (Beacon Press, 1944); Daron Acemoglu and James Robinson, Why Nations Fail (Crown, 2012).

“The employees have for many years been faithful co-workers with the company in the business of the mill; have invested thousands of dollars of their savings in said mill in the expectation of spending their lives in Homestead... the employees have the right to continuous employment in the said mill during efficiency and good behavior.” — Declaration of the Homestead Strike Committee, July 20, 1892

The Works

Homestead, Pennsylvania, sits along a bend in the Monongahela River, seven miles southeast of Pittsburgh. The steel mill, which once employed fifteen thousand people1, is a shopping center now. Twelve smokestacks still stand among the retail outlets — the last physical trace of the works that Andrew Carnegie built into the largest steel mill in the world.

In the summer of 1892, the workers at that mill went on strike. Carnegie’s chairman, Henry Clay Frick, locked them out, ringed the plant with barbed wire and sniper towers, and hired three hundred Pinkerton agents2 to retake it by force. The Pinkertons came up the Monongahela on barges before dawn on July 6th. Thousands of workers and their families met them at the riverbank.

The battle lasted all day. Workers and Pinkertons exchanged rifle fire for hours. The strikers tried cannon fire, burning rafts, dynamite, oil slicks on the river. By evening, the Pinkertons surrendered. At least sixteen people were dead.3 The governor sent more than six thousand state militia to reopen the plant.4 The strike was broken. The union collapsed.

But the fights continued — through the Progressive Era, through the New Deal — until workers won what the men at the riverbank had died for: the right to organize and to share in the prosperity their labor created. By the middle of the twentieth century, steelworkers at Homestead had a union, a pension, health insurance, job security. The deal that people had bled for had, at last, been built.

Margaret Mary Vojtko was born in Homestead in 1930. Her father worked at the mill. He belonged to the union that became the United Steelworkers — the union that carried the legacy of the men who fought on the Monongahela.

She was the youngest of six. Her mother died when she was seven. She grew up speaking Slovak at home, attended a Catholic high school, and worked as a secretary before putting herself through the University of Pittsburgh. She earned a bachelor’s degree, then a master’s. She spoke six languages. She started a doctoral dissertation on the history of Homestead — her father’s workplace, the site of the battle, the place where the deal began.

In 1988, Margaret was hired to teach French at Duquesne University, a Catholic school on a hilltop in Pittsburgh, five miles from where she grew up. She taught there for twenty-five years. She was an adjunct — which meant between $3,000 and $3,500 per course,5 no benefits, no pension, no health insurance, no office, and no guarantee, from one semester to the next, that they would offer her work.

She taught the same courses tenured professors taught, sometimes more. She never missed a day of class. Her student evaluations were glowing. In her best years, teaching three courses a semester plus two in summer, she made under $25,000. Duquesne’s president made nearly $700,000.6

In her final year, the university cut her to one course per semester. She was making well below $10,000. She had cancer and huge medical bills. She couldn’t keep the electricity on in her home. She took a night job at Eat’n Park and tried to sleep during the day in her office at Duquesne. When the university discovered this, they called the police to remove her.

In the spring of 2013, Duquesne let her go. They told her she was no longer effective as an instructor. She was eighty-three years old.

The previous summer, the adjunct faculty at Duquesne had voted to unionize. They organized under the United Steelworkers — the same union Margaret’s father had belonged to, the same union whose roots ran back to the men on the riverbank. Margaret voted yes.

Duquesne refused to recognize the union. The university argued that its Catholic identity exempted it from the National Labor Relations Board. Georgetown, another Catholic university, had just recognized its adjunct union, citing the Church’s social justice teachings to reach the opposite conclusion.7

On August 16th, Daniel Kovalik, an attorney for the United Steelworkers8 who had been helping Margaret file an age discrimination complaint, received a call from her. She was distraught. The cancer was back. She was nearly homeless. And now she had received a letter from Adult Protective Services — someone had referred her case, saying she needed help taking care of herself. For a woman who had taught at a university for a quarter century, who spoke six languages, who had never missed a day of class, the letter was a final indignity.

Kovalik called Adult Protective Services to explain the situation. He told the caseworker that Margaret had just been let go from her job as a professor, that she had received no severance or retirement benefits, that the reason she was struggling was because she was living in extreme poverty. The caseworker paused.

“She was a professor?”

Margaret Mary Vojtko died on September 1, 2013. She was eighty-three. Her funeral Mass was held at Epiphany Church, a few blocks from Duquesne. She was laid out in a cardboard casket with no handles for pallbearers.9

Her father, the steelworker, had had the deal that the men had bled to forge. His daughter did everything the American Dream asked — she got the education, she learned the languages, she climbed from the mill to the university. And she died unable to heat her home, because the system her father’s generation had built was quietly dismantled around her, and nothing had been built to replace it.

Notification

Bernard Moses drove for Uber in the Chicago suburbs for nearly ten years. Twenty thousand rides. A 4.99 average rating. In April 2024, he dropped off a passenger and swiped for the next ride. He was locked out. A customer complaint was all he was told, with no warning, explanation, or appeal. He called support daily for months, working through automated menus, occasionally reaching a voicemail that never called back. His rider account was locked too. Nearly a year later, he had no resolution. But the car payment was still due.10

Jacqueline Bowman left journalism to freelance as a content writer in California. In 2024, something switched. Clients came to her not to write but to clean up what AI had produced — at half her usual rate. It took twice as long. She was spending more hours making half the money, fact-checking fabrications line by line and largely rewriting the article anyway.

“Writing is not going to work out for me any more,” she told The Guardian in February 2026.11 By January 2025 she had lost her health insurance. She brought her wedding forward to get onto her husband’s plan. She is retraining as a therapist.

The Bundle

In most countries, a job is a job. It pays you for your work. Healthcare comes from somewhere else. Retirement comes from somewhere else. Training, labor protections, the safety net when things go wrong— those are handled by other systems, funded in other ways.

In America, we made a different choice. We bundled all of it together and attached it to your employer. Your job was your income, your health insurance, your retirement plan, your access to training, your structure for how your work was managed and evaluated. One relationship carried everything.

This was a design decision, made in stages, mostly during and after World War II, reinforced by tax policy and labor negotiation and decades of institutional habit.12 It worked well enough, for long enough, that we stopped seeing it as a choice at all.

Which meant that when the employment relationship started to change, everything changed with it.

• Compensation became unpredictable. A freelancer’s rate depends on how many other freelancers bid this week. A driver’s earnings depend on an algorithm that adjusts in real time, using inputs the driver can’t see, always optimizing from the platform’s margins down to you. The money might be fine this month. There’s no way to know about next month.

• Benefits didn’t follow the worker. You can buy health insurance on the exchange, open your own IRA, purchase disability coverage — each of these is possible. Navigating all of them simultaneously, while freelancing across three clients and driving on weekends, requires a level of financial sophistication that the old system never asked of anyone.

• Training cycles broke down. There used to be a reasonable bet: learn a skill, practice it, get better, earn more. Companies invested in training because they expected to keep you. That cycle has been speeding up for years. In certain fields, the skill you spent two years learning gets automated before you’ve recouped the investment.

• Management became algorithmic. In the old system, your boss was a person. They might be good or bad, fair or unfair, but you could read them, talk to them, appeal to them. Algorithmic management replaces that with a system that makes decisions constantly — about pay, about allocation, about whether you still have a job — and doesn’t explain them. The driver loses control of his rate to an algorithm. The adjunct loses security to enrollment numbers processed by software. And the logic that makes work more efficient doesn’t stop at managing how work gets done. Eventually, it asks whether the human doing the work is needed at all.

Each of these shifts has its own logic. Each system optimizes for its own purposes, with no one watching what happens to the person standing in the middle of all four.

The Investment Thesis

Once the bundle started coming apart, someone was going to figure out how to make money from the pieces.

The first wave was defensive. Legacy industries under competitive pressure discovered they could get the same work done for less by breaking full-time positions into contract roles. A CFO looking at the cost of a full-time employee versus a contractor is responding to real incentives — quarterly earnings, board expectations, competitive pressure. A university president watching state funding decline year after year is making hard choices with shrinking resources. These are rational actors inside a system that rewards exactly this behavior.

But the second wave turned those same mechanics into a business model.

Uber doesn’t own cars. Airbnb doesn’t own apartments. DoorDash doesn’t own restaurants or employ delivery drivers. The insight was that you could build a company, quickly, by connecting customers to workers without actually employing anyone or carrying any meaningful assets. The workers brought their own capital — their car, their apartment, their bike — and absorbed their own risk. The platform took a cut of every transaction and retained the one thing that mattered: the customer relationship.

These were talented entrepreneurs solving real problems. Getting a cab in most American cities was a genuinely miserable experience. The platforms removed friction, elegantly. And for a lot of workers, the flexibility was real and valuable.

But the model required something specific: that the people doing the work remain independent contractors. Employees come with the bundle. Independent contractors don’t. Uber and its backers spent over $200 million on California’s Proposition 22 to preserve that distinction.14 The math didn’t work otherwise.

Today, Uber is valued at roughly $145 billion.15 Airbnb at roughly $80 billion.16 DoorDash at roughly $80 billion.17 The combined enterprise value of the major gig platforms runs above $300 billion.18

The average Uber driver, after expenses, makes about $15 an hour.19

Four Hours

Matt Shumer has worked in artificial intelligence for nearly six years. Not as a commentator or a futurist — as a builder. The kind of engineer who understands the technology well enough to know what it can’t do. He’s been skeptical of hype cycles before. He’s watched models improve incrementally and said, correctly, that they weren’t ready yet.

On February 5th, 2026, two major AI labs released new models on the same day. OpenAI launched GPT-5.3 Codex. Anthropic released Claude Opus 4.6.20

Shumer described what happened next on his blog.21 He told the AI what he wanted built, in plain English. He walked away for four hours. When he came back, it was done. Not a rough draft he needed to fix. The finished thing. Tested, refined, ready.

“I am no longer needed,” he wrote, “for the actual technical work of my job.”

This is someone who has spent six years at the frontier of the most consequential technology of the century. His skills are rare. His judgment is hard-won. His compensation reflects that — engineers at this level make $500,000 to $800,000 a year.22 They were supposed to be the last people this happened to.

He described the feeling: “Not like a light switch. More like the moment you realize the water has been rising around you and is now at your chest.”

Then he said the thing that matters:

“The experience that tech workers have had over the past year, of watching AI go from ‘helpful tool’ to ‘does my job better than I do,’ is the experience everyone else is about to have. Law, finance, medicine, accounting, consulting, writing, design, analysis, customer service. Not in ten years. The people building these systems say one to five years.23 Some say less.”

“And given what I’ve seen in just the last couple of months, I think ‘less’ is more likely.”

The driver, the copywriter, even the PhD, each of them could be told — and has been told — that they made the wrong choices. They should have picked a different career. They should have adapted faster. They should have seen it coming.

Matt Shumer made every right choice. He picked the most valuable skill of his generation. He built nearly six years of expertise at the exact frontier of the technology that’s reshaping everything.

And he’s writing blog posts to warn you that it’s already too late for him.

Margaret’s father worked at the mill for forty years before the deal he had was negotiated and secured. Margaret taught for twenty-five years before her deal was taken away. Matt Shumer worked at the frontier for six years. Four hours was enough to see the end.

Cardboard

The distance between Matt Shumer and Margaret isn’t merit. It’s time.

Her father had the deal. He’d had it because men before him had fought for it.

His daughter had a cardboard casket with no handles.

The system her father’s generation built was quietly dismantled around her, and nothing was built to replace it. The driver, the copywriter, the PhD, the elite engineer — they’re all standing somewhere on the same arc Margaret traveled. The distance between them is time, and time is compressing.

The question isn’t whether the system will break. It broke a long time ago. The caseworker who paused on the phone and said “She was a professor?” — she already knew.

The question is what we build next.

Suggested Sources

American Labor History

• Paul Kahan, The Homestead Strike: Labor, Violence, and American Industry (Rowman & Littlefield, 2014)

• Les Standiford, Meet You in Hell: Andrew Carnegie, Henry Clay Frick, and the Bitter Partnership That Changed America (Crown, 2005)

• David Brody, Steelworkers in America: The Nonunion Era (Harper Torchbooks, 1960)

The Employer Bundle / Benefits History

• Jacob Hacker, The Divided Welfare State: The Battle over Public and Private Social Benefits in the United States (Cambridge University Press, 2002)

• Jennifer Klein, For All These Rights: Business, Labor, and the Shaping of America’s Public-Private Welfare State (Princeton University Press, 2003)

Gig Economy / Platform Labor

• Alex Rosenblat, Uberland: How Algorithms Are Rewriting the Rules of Work (University of California Press, 2018)

• Lawrence Mishel, “Uber and the Labor Market,” Economic Policy Institute (2018)

Adjunct Labor

• American Association of University Professors (AAUP), annual “Contingent Faculty” data reports

AI and Labor Displacement

• Dario Amodei, “Machines of Loving Grace” (Anthropic blog, 2024)

• Daron Acemoglu and Pascual Restrepo, “Robots and Jobs: Evidence from US Labor Markets,” Journal of Political Economy 128:6 (2020)

Financialization of American Industry

• Rana Foroohar, Makers and Takers: The Rise of Finance and the Fall of American Business (Crown Business, 2016)

• Barry Lynn, End of the Line: The Rise and Coming Fall of the Global Corporation (Doubleday, 2005)

“A decent home and a suitable living environment for every American family.”

— The United States Housing Act of 1949, Declaration of National Housing Policy

A Home He Built by Hand

Don Lund moved into Golfview Mobile Home Court in North Liberty, Iowa, in 1980. He planted pine and peonies and hostas and lilacs, built a front deck and a back deck with his father, and became president of the residents’ association. Don was a quadruple congenital amputee. He worked as a sports writer covering Iowa Hawkeye athletics, lived on food stamps and Social Security, and budgeted with the precision that poverty demands.

In 2019, Havenpark Communities, an Utah-based investment firm, purchased Golfview. Don’s lot rent was $87 a month. He came home to a notice on his door: rent jumping to $450, then to $506. Following public outcry, the increases were spread over approximately two years. His income hadn’t changed. His home hadn’t changed. Only the ownership of the ground beneath him had changed—and with it, every assumption his life was built on.

“Just because I don’t have a lot of money,” Don told reporters, “doesn’t mean I’m not important.”1

Don Lund died in April 2023.2

The Promise

After World War II, the federal government actively promoted manufactured housing as a path to middle-class stability. Returning veterans needed homes fast, and the construction industry couldn’t build conventional housing quickly enough. Factory-built homes—first called “trailers,” then “mobile homes,” eventually “manufactured housing”—offered a solution: decent shelter at a fraction of the cost, producible at industrial scale.

The industry boomed. By the 1960s, manufactured homes accounted for a significant share of new single-family housing. In 1974, Congress passed the National Mobile Home Construction and Safety Standards Act, which for the first time established federal building codes for factory-built homes—administered by HUD, making manufactured housing the only form of housing in America regulated at the federal rather than local level. The HUD Code, updated in 1994 and periodically since, was designed to ensure quality and build consumer confidence. It was a statement of legitimacy: this is real housing, backed by the federal government, suitable for American families.

The deal was this: You can own a home. It will cost less than a conventional house. It will be built to federal standards. You will have stability, equity, and a step into the middle class. This is how ordinary Americans build a life.

Twenty-two million Americans accepted that deal. Manufactured housing remains the largest source of unsubsidized affordable housing in the country—roughly 6% of the population, 10% of new housing stock.3

Personal Property

Despite the name, mobile homes are not mobile. Moving one costs between five and twelve thousand dollars—often more for a double-wide—and frequently damages or destroys the structure.4 Over 90% never move once placed. Annual turnover averages roughly 10%, about a third the rate in apartment buildings. Investors call this “stability,” but the more honest word has become captivity.

A resident owns the home—a depreciating asset, often financed as personal property, like a car, at higher interest rates and shorter terms, rather than through a conventional mortgage. She rents the land—an appreciating asset she does not control, under terms that can change without her consent. When lot rent rises, there is no comparable alternative. When services decline, there is no leverage. The home that promised stability becomes a mechanism of exposure for families working hard to climb into the middle class.

The Notice in the Mailbox

Linda Bates used a disability settlement—the largest sum she had ever received—to buy a mobile home for $62,000 in 2016. She moved with her daughter and granddaughter to Midwest Country Estates in Waukee, Iowa. It was the safest investment she could imagine.

Three years later, Havenpark Capital acquired the park as well. New rules. New fees. Then notice that her lease would not be renewed. There was no reason given. In Iowa, this is called a “no fault” eviction, and it is legal.5

Linda owned her home. She did not own the land. She had to leave or be removed by the sheriff.

Moving the structure was financially impossible. She found a buyer willing to pay $22,500—less than half what she had paid three years earlier. At sixty-three, disabled, she left behind the only major asset she had ever owned.6

Land. Finance. Tenure. Ownership.

What makes manufactured housing incoherent is the collision of four systems, each following its own logic and none accountable for what happens to the person standing at the intersection.

• *Land.* Zoning laws passed decades ago to exclude mobile home parks from suburban neighborhoods effectively banned new parks from being built. The intent was exclusionary. The long-term consequence was monopoly. If you own one of a handful of legally permitted parks in a metro area, you have a captive market that can never be diluted by new supply. The zoning meant to protect property values for the owners of single-family homes created the preconditions for extraction from the owners of mobile homes.

• *Finance.* Manufactured homes are typically excluded from conventional mortgage products. Chattel loans carry interest rates 2-5% higher than standard mortgages, with terms of fifteen to twenty years instead of thirty.7 The monthly cost is driven more by debt structure than by the home’s value.

• *Tenure.* In many states, lease terms run a year or less—against lives measured in decades. Don Lund lived at Golfview for forty-two years. His legal right to stay was renewed twelve months at a time. Protections vary widely by state, and enforcement is uneven where protections exist. Decades of residency, of building decks and planting gardens, confer almost no security when the land changes hands.

• *Ownership.* You hold title to the home. But it sits on someone else’s land, depreciates from the day you buy it, and its resale value is mathematically tethered to the rent beneath it. The relationship is roughly linear: by one common analytical estimate, every $100 increase in monthly lot rent destroys approximately $10,000 in resale value.8 The resident doesn’t just lose cash flow. She loses net worth. Linda’s home didn’t depreciate because it deteriorated. It depreciated because someone raised the rent.

The resident must navigate all four systems simultaneously. No actor is accountable for the combined result.

When the Investment Thesis Arrives

Over the past decade, institutional investors have purchased roughly one-fifth of manufactured home parks in the United States.9

The thesis is straightforward: a fragmented industry of aging park owners represents a consolidation opportunity. Buy at low multiples, raise rents toward “market” rates, reduce operating costs, hold the asset until the right moment to flip. In private equity-backed parks, the majority of returns come from rent and fee increases rather than from adding units or improving services.

The target is 5-8% annual rent increases—two to four times general inflation. Compounded over a typical seven-year hold, that means cumulative increases of roughly 40-70%.10

Nearly half of these acquisitions are financed through Fannie Mae or Freddie Mac11 —government-sponsored enterprises created to expand access to affordable housing. The federal government, through its lending apparatus, subsidizes the purchase of the affordable housing it aims to promote.

The portfolio manager allocating capital to manufactured housing is not sitting in a room plotting to ruin Don Lund’s life. She is evaluating risk-adjusted returns across asset classes, under fiduciary obligations to her investors—pension funds, university endowments, retirement accounts. The parks she’s buying were often under-maintained by their previous owners, and the rents were often genuinely below what the local market could bear. From inside her spreadsheet, the logic is clean. She may even believe, with some justification, that professional management will improve conditions in the long run.

Stability

Nicole Platz moved to Modern Manor in Iowa City in 2009. Lot rent was $395. The park included trash service, yard care, and responsive management.

When Havenpark took over in 2019, her rent increased nearly 40%. Services disappeared. In January 2025, the park lost running water for three days. Residents used plastic bags as makeshift toilets.12

Havenpark had already done this at Golfview and Midwest Country Estates. The firm had a playbook. The residents had a notice on their door.

The consequences extend beyond housing economics. People who live in the same place for more than five years vote at significantly higher rates than those who’ve moved in the past year. They volunteer more, join civic organizations, serve on school boards, know their city council members by name. Community tenure is among the strongest predictors of local political participation, reinforcing the effects of education and income, and sometimes substituting for them.13 When housing becomes unstable, the institutions that require sustained local engagement begin to hollow out. A 2023 study found that each one-percentage-point increase in neighborhood eviction rates was associated with a nearly half-point decline in voter turnout.14

Sixteen Million Units

Don Lund lived in a “mobile” home.

When you rent an apartment, mobility is an explicit part of the bargain. But even that bargain is becoming subtly incoherent.

In conventional rental markets, large landlords increasingly use pricing software—most notably RealPage’s YieldStar platform—to determine what tenants pay. Individual landlords feed their occupancy data, lease terms, and pricing into RealPage’s system. The algorithm aggregates this nonpublic data across competing properties in the same market—properties whose owners would violate antitrust law if they shared pricing information directly—and generates rent recommendations calibrated to maximize collective revenue. The software has explicitly recommended that landlords accept higher vacancy rates in exchange for higher per-unit revenue, overriding the competitive logic that is supposed to push rents down when apartments sit empty. A 2022 investigation found that RealPage’s clients managed more than sixteen million units nationwide.15 In some markets, a single algorithm effectively set the price for the majority of available apartments.

In August 2024, the Department of Justice filed an antitrust suit against RealPage, alleging that the platform enabled landlords to coordinate pricing in violation of the Sherman Act.16 The complaint describes a system designed to replace competitive pricing with algorithmic consensus—a cartel mediated by code rather than a handshake.

Amanda Tolep rented in Seattle’s Belltown neighborhood, in a building managed by Essex Property Trust—one of ten landlords that together controlled 70% of the 9,000 apartments in her area. Every one ran YieldStar. When her rent jumped 33% in a single year, she looked for alternatives in her area. There were none. The algorithm was running everywhere she could go. She left the city and started over in a community thirty miles north.

In Boston, Kaylee Hutchinson tried to negotiate with her landlord during the first pandemic lockdown, when vacancies were rising and market logic said prices should fall. The landlord held firm. “It was pretty obvious they should have been dropping prices,” she said. “These companies, they’ll just replace you.”17

The same algorithmic repricing logic is now appearing in homeowner insurance markets, where carriers have exited or sharply raised premiums across California, Florida, and Texas—removing a layer of stability from conventional homeownership that most owners assumed was permanent.[^20]

Forty Years

Don Lund planted lilacs. He built decks with his father. He did everything the system asked of him, for near forty years, in a home he believed was his.

The rules didn’t protect him.

Suggested Sources

The History and Structure of Manufactured Housing

• Esther Sullivan, *Manufactured Insecurity: Mobile Home Parks and Americans’ Tenuous Right to Place* (University of California Press, 2018) — the foundational academic account of park life, tenant organizing, and displacement

Zoning, Supply, and Land Use

• Terner Center for Housing Innovation, “Regulation and Housing Supply” (2023)

• National Zoning Atlas project — state-by-state data on manufactured housing restrictions in zoning codes

Algorithmic Pricing and Rental Markets

• Herbert Hovenkamp and Fiona Scott Morton, “Framing the Chicago School of Antitrust Analysis,” *University of Pennsylvania Law Review* (2020) — legal theory underlying the DOJ RealPage case

Macroeconomic Context

• Bolhuis, Cramer, and Summers, "Comparing Past and Present Inflation," NBER Working Paper No. 30116 (2022); published in Review of Finance Vol. 26 — on CPI shelter cost methodology

• Fonseca and Liu, “Mortgage Lock-In, Mobility, and Labor Reallocation,” Federal Reserve Bank of New York Staff Reports (2023) — on how locked-in homeowners suppress conventional housing supply

Note: the first photo is of Creekside Mobile Home Park, Esperance, New York. Photo: Famartin / Wikimedia Commons, CC BY-SA 4.0

“The hunted news I get from some obscure patients’ eyes is not trivial. It is profound.”

William Carlos Williams (Autobiography, 1951, Ch. 54: “The Practice”)

Something actually has changed

The headlines are reassuring.

Long-term cancer survival rates continue to improve. Cardiovascular mortality continues to decline. Stroke outcomes continue to improve markedly. By 2023, more Americans were covered by health insurance than at any previous point in the country’s history.1 The Consumer Price Index shows healthcare inflation has been a manageable 2.8% annually over the past five years2.

And yet, healthcare is where many of us first notice that something has shifted.

Vera and I first noticed it with our third child’s delivery at Sibley, a Johns Hopkins hospital. But the same pattern repeated for me with a vasectomy at a Medstar health center a year later, then a first colonoscopy a year after that.

It took nine months to schedule a spot for a routine family planning procedure. It took four to schedule preventative care. With the colonoscopy, someone at Medstar said that there was no way to know in advance what it would cost me out-of-pocket.

Revenue Integrity

Hospitals are a place to receive care. But they are also an asset that converts inputs into financial outputs.

As a financial asset, hospitals define success through industry metrics like throughput and asset utilization.

Consider how the world looks to the CFO of a hospital system. Let’s call her Sarah. She isn’t optimizing these metrics to boost a stock price: a significant majority of US health systems are non-profits. Sarah’s fighting to keep the emergency department open.

She knows that 40% of U.S. hospitals are currently operating in the red. With operating margins around 1.5% and nurse labor costs up 27%, efficiency is what separates continued operation from closure.

To keep the doors open, Sarah relies on sophisticated software systems known as Revenue Cycle Management.

In a modern hospital, insurers do not pay for “care.” They pay for codes. Every distinct action—a checkup, a suture, a complex diagnosis—must be translated into a specific billing code to trigger payment. If a doctor notes “pneumonia” but fails to document the specific clinical indicators of “complex pneumonia with complications,” the hospital might lose thousands of dollars for the exact same treatment.

In 2015, the U.S. healthcare system adopted a new standard for codes. Overnight, billing codes expanded from roughly 14,000 to 68,000. The administrative burden of defining a “broken arm” fractured into half a dozen hyper-specific variations.

This explosion in granularity outpaced human cognition. Revenue Cycle Management became a technological arms race. Sarah invests millions in algorithms that scan medical notes in real-time, hunting for keywords that justify higher-severity codes.

The software nudges a “Level 3” office visit to a “Level 5” emergency response. It suggests adding a secondary diagnosis of “malnutrition” or “respiratory failure” to a standard admission. In the industry, this is called “Revenue Integrity.” To Sarah, it is the only way to capture the full value of the work her staff performs.

But revenue optimization alone cannot cover rising labor and drug costs. Sarah must also eliminate the buffer — the spare capacity that absorbs variation but generates no revenue.

Over time, the efficiency pressure Sarah manages compressed that buffer. Scheduling tightened toward full capacity. Staffing ratios thinned. Discharge timelines accelerated.

Buffer eroded slowly for decades. Then the shock arrived.

When the pandemic hit, the “lean staffing” models that worked on paper collapsed. Between 2020 and 2023, around 100,000 registered nurses left the profession3.

Staff already at their limit were asked to sustain the impossible. Many left. Institutional memory left with them. The physical beds remained, but without the staff to tend them, they became “ghost capacity,” or assets that exist on the ledger but cannot accept a patient.

Even with the pandemic behind us, the system has reshaped how we often experience care. Staffing gaps now stall admissions. Scheduling surges spill into hallways. Delayed discharges cascade through emergency departments.

Most of us don’t get sick or fall off a ladder on a predetermined schedule. Surgeries overrun. Discharges hinge on insurers, transportation, and home care availability. These are all inputs beyond hospital control. With buffers already thinned and the workforce shattered, that variation now converts directly into exposure for patients.

Slowly, Then All At Once

Operations science predicts this exactly. Kingman’s Formula describes how wait times behave under load—and the insight is unforgiving. As utilization approaches full capacity, delays don’t rise gradually. They rise exponentially.

A system running at 60% capacity absorbs variation easily. At 85%, delays grow quickly. At 98%, even minor disruptions—an absent nurse, a delayed discharge, a surge of arrivals—produce cascading backlogs, often paralyzing a system in seemingly inexplicable ways.

This is how breakdowns often happen: slowly, then all at once.

This phase change explains why the system hasn’t snapped back even as the consequences of the pandemic have faded. It is tempting to view this instability as a temporary labor cycle, a disruption that will smooth out as wages rise and new nurses graduate. But the same logic of optimization acts as a ratchet.

To pay for the 27% increase in labor costs without a matching increase in reimbursement, Sarah cannot afford to rebuild the buffer she lost. She must do the opposite: run the remaining assets at even higher utilization to cover the increased cost per unit.

We’re in a new equilibrium where the steep part of the exponential curve is the daily operating target.

Prior Authorization

But our experience of healthcare isn’t limited to health systems.

Insurers manage risk across millions of claims, using those same 68,000 codes. At that scale, human discretion inevitably gives way to automation. Algorithms process submissions, compare them against statistical norms, and route deviations into denial or appeal workflows.

Consider how this looks to a VP of Claims Operations at a major insurer. We’ll call him David. He isn’t optimizing for affordability. He is optimizing for predictability.

When his algorithm flags a mismatch, David sees a guardrail against “waste” and “upcoding,” not a patient in distress. He optimizes to protect the risk pool while ensuring his shareholders maintain a competitive return on equity.

But David operates under a specific regulatory constraint that twists these incentives.

The Affordable Care Act caps David’s profit margin. Under the Medical Loss Ratio rule, 80 to 85 cents of every premium dollar must be spent on care. He can only keep the remaining 15 cents for administration and profit. If David successfully crushes the cost of care, his 15% slice shrinks in absolute dollars. If the cost of care rises, his 15% slice grows.

Mathematically, David wants his overhead to be lean. But he wants the cost of healthcare to rise—predictably.

David sets his premium rates 12 to 18 months in advance. If the hospital raises its prices faster than that schedule, David’s margin for the current year vanishes. This transforms the “Denial” from a medical decision into an algorithmic financial throttle.

Prior authorization governs utilization before care occurs. Claims adjudication evaluates billing afterward. David uses them to slow the hospital’s spending just enough to ensure it matches the premiums he has already collected.

The hospital submits the highest defensible price and moves on. The insurer adjudicates the low payment (or denial) and closes the file. Each system behaves logically within its own frame, effectively passing the mismatch to the other side.

Sheldon Ekirch was 28 and about to start her legal career when she was diagnosed with small fiber neuropathy, a condition causing near-constant pain. Her neurologist recommended a treatment. Anthem denied it as not medically necessary. She appealed; the state review upheld the denial. Her parents withdrew from their retirement savings to cover what her insurance would not. “I’m paying a lot of money for health insurance,” she said. “I don’t understand why they won’t help me.”4

What the Index Shows

Tom Contos, a healthcare consultant in Chicago, used Northwestern Memorial Hospital’s online cost estimator before his diagnostic colonoscopy in June 2024. His estimate: $2,381 out-of-pocket. He paid $1,000 upfront as the hospital requested.

The bill that arrived was $4,047.5 Northwestern had billed two colonoscopies — modifier codes for removing polyps in different ways. Its rate for the procedure was more than twice the Aetna median at other Chicago-area hospitals. The charge was, the hospital told him, “non-negotiable.”

“I said, ‘I don’t understand this,’” Contos recalled. “Then I started to research the cost.”

The Consumer Price Index captures what a basket of goods costs this month versus last month. It is designed for consistency of measurement. When a hospital updates its chargemaster, the index verifies that the defined service — the specific procedure code — remains consistent over time. Not whether the service was available. Not whether it worked.

In 2023, that methodology reported health insurance prices falling nearly 4%.6 It did not record whether you waited three hours in triage, or missed the window for an epidural, or received bills for uncovered services for years later.

The costs created by the throttle — the denied claims, the delayed approvals, the hours of unpaid coordination labor — are invisible to the index.

When Premiums Rise

An economist might argue that all this is a temporary disequilibrium: as scarcity bites, premiums will rise, reimbursements will follow, and hospitals will use that new capital to rebuild the workforce and restore the buffer.

This assumes that a dollar of new premium translates into a dollar of new capacity.

In healthcare, that transmission belt is broken.

As premiums rise, the additional revenue enters what healthcare advocates call a shooting war. David, the insurer, doesn’t use the surplus to loosen authorization rules. He invests in more sophisticated denial algorithms to protect the new, higher stakes.

Sarah, the CFO, doesn’t use a reimbursement hike to hire now “redundant” nurses for a rainy day. She uses it to pay the 27% wage premium for the staff she already has and to upgrade the RCM software needed to fight David’s new algorithms.

Inflation doesn’t buy capacity; it buys complexity. It makes it feel more arbitrary for you.

Robert Cano managed a retail store. Tiffany worked in bank compliance. Together they earned nearly $100,000 a year. Robert’s employer plan cost $500 a month. Their first child was born in January 2018 — healthy, no complications. The bills came to more than $12,000. Robert took on three additional jobs — substitute teaching, overnight security, delivering sandwiches for a fast-food chain 40 miles away where the tips were better — and sometimes worked 120 hours in a week.

“My husband is working four jobs. I work full time. We’re a hardworking family doing our best and not getting anywhere,” Tiffany said. “How does everybody else do it?”7

41% of American adults are asking the same question. Together they carry $88 billion in medical debt.8

Without a Profit Motive

Ervin Kanefsky was 94 when he fractured his shoulder in a fall. His doctor ordered him admitted as an inpatient. The hospital’s utilization review team overrode that order. At discharge, he was told he had been classified as an observation patient the entire time. Without three qualifying inpatient nights, Medicare would not cover his skilled nursing care. The bill: $9,145.

“If I had gone home, I would have died,” he said. “I tried every which way.”9

Observation status is a billing designation, not a clinical one. The hospital applies it based on Medicare’s payment rules, not a doctor’s judgment. Roney had been in a hospital bed for three days. The classification that determined his cost was invisible to him until the bill arrived.

The designation carries a second trap. Medicare covers skilled nursing and rehabilitation only after at least three consecutive inpatient nights. Observation nights don’t count. A patient hospitalized for a hip fracture — four days in a hospital bed, classified as observation — may find that Medicare covers none of the subsequent rehabilitation. Nursing home care averages $290 a day. A month of rehab a family assumed was covered can cost $8,000 or more.

In Medicaid and Medicare Advantage, the optimization can take a different form. Coverage exists on paper for those enrolled. But access does not always follow. Patients encounter ghost networks: clinics listed but unavailable, with appointments impossible to secure.

The Cash Market

Veterinary medicine operates largely as a cash market — no insurers, no prior authorization, no David. But the “Main Street” veterinary storefront is increasingly an illusion. Mars, the candy company, owns Banfield, VCA, and BluePearl. Together with one other private equity consolidator, corporate chains now control roughly 30% of general practices and nearly 75% of emergency clinics.

When Kathleen Whitman’s daughter’s dog swallowed a Kong toy, her local clinic’s x-rays suggested an intestinal obstruction. She was referred to a corporate emergency hospital. Before she had committed to anything, they handed her a clipboard in the waiting room. The number was $10,000. “I said, ‘No, I’m just coming in to hear the price,’” she recalled. She found an independent clinic — founded by two vets who had left Banfield and VCA — where the surgery cost significantly less.10

The other failure mode doesn’t involve prices at all. Jennifer Fraser’s new puppy Simba was straining to urinate and couldn’t. She called every emergency clinic in the Pittsburgh area. Every one gave her the same answer: unless her pet was dying, they couldn’t see him tonight. A urinary blockage — which looks exactly like a urinary tract infection — can be fatal within hours. She didn’t know which one she was looking at.11

The American Prospect documented why. At Mars-owned clinics, one technician now oversees 18 patients. The old standard was 6.12 When a single staff member calls out, the system has no buffer. The door doesn’t lock. The lights stay on. The sign still says 24-hour emergency care. The answer is still the same.

You

Eric Tennant was a safety inspector for the West Virginia Office of Miners’ Health and Safety when he was diagnosed with stage 4 bile duct cancer. His oncologist recommended a noninvasive procedure that might slow the progression. UnitedHealthcare denied it four times. The insurer reversed the denial after journalists contacted them for comment. Tennant had become too ill for the procedure by then. He died September 17, 2025.13

This is what arbitrary feels like: the outcomes you experience are defined by algorithms optimizing for someone else’s financial outcomes.

You have a phone, a credit score, a bank account, and limited free time.

Suggested Sources

I. . The System Change: What Shifted Between 2017 and 2023

• On Hospital Capacity and the Post-Pandemic Squeeze: Needleman, J., et al. (2024): “Health Care Staffing Shortages and Potential National Hospital Bed Shortage” JAMA Network Open.

• On the Nursing Exodus: National Council of State Boards of Nursing (2023): “NCSBN Research Projects Significant Nursing Workforce Shortages and Crisis”

• On Hospital Financial Pressure: Advisory Board (2025): “Charted: The Current State of Hospital Finances”

II. The Algorithmic Turn: AI in Claims and Coding

• On AI-Driven Prior Authorization Denials: American Medical Association (2025): “How AI Is Leading to More Prior Authorization Denials”

• On Claim Denial Rates Over Time: Kaiser Family Foundation (2024): “Claims Denials and Appeals in ACA Marketplace Plans”

• On UnitedHealthcare’s Algorithmic Transformation: U.S. Senate Permanent Subcommittee on Investigations (2024): “Denial of Care: How Medicare Advantage Insurers Delay and Deny Medically Necessary Care”

III. The Dashboard Problem: Why CPI Misses the Experience

• On CPI Methodology for Health Insurance: Bureau of Labor Statistics (2024): “Measuring Total-Premium Inflation for Health Insurance in the CPI”

• On What Medical CPI Fails to Capture: BDO (2024): “Why Medical Care CPI Falls Short in Explaining Rising Healthcare Costs”

IV. The Public Sector Control Group

• On the Medicare Observation Status Trap: Centers for Medicare & Medicaid Services: “Skilled Nursing Facility Care”

• On Medicaid Ghost Networks: HHS Office of Inspector General (2025): “Many Medicare Advantage and Medicaid Managed Care Plans Have Limited Behavioral Health Provider Networks”

V. The Unregulated Control Group: Veterinary Medicine

• On Private Equity Consolidation in Pet Care: U.S. Senate (2024). Warren, E. & Blumenthal, R: “Investigation into Mars’ Impact on Pet Owners and Veterinary Workers”

• On the Veterinary Staffing Crisis: American Veterinary Medical Association (2021): “Are We in a Veterinary Workforce Crisis?”

VI. Source Notes

• On Revenue Cycle Excellence as Industry Achievement: Healthcare Financial Management Association (2023): “15 Winners Receive the 2023 MAP Award for High Performance in Revenue Cycle”

Note on the ICD-10 Transition: The 2015 shift from ICD-9 to ICD-10 expanded billing codes from roughly 14,000 to over 68,000. This is well-documented by CMS in “Transitioning to ICD-10.” The explosion in coding granularity is the structural precondition for the Revenue Cycle Management arms race described in the essay.

This essay is the first in a four-part series meant to frame this project, beginning with the sense that something fundamental may have shifted, and that the buffers many of us relied on are no longer holding the way they once did.

“There’s something wrong when nothing’s right.”

— Townes Van Zandt

The Rules Didn’t Protect Us

“Always take the epidural.”

That was Vera’s standing advice to her friends after our first two children.

When we found out we were having a third—long after we assumed we were done—it felt like a bonus round. We knew the drill and what to expect.

Vera woke me up on the morning of May 16, 2023. Labor had started. We packed a bag, kissed the older kids, and headed to Sibley Hospital.

Sibley was familiar ground. Our experiences there in 2015 and 2017 had worked exactly as planned. That’s why the disorientation hit so hard this time.

There was no room available. There was no maternity nurse to move things along. Vera’s labor was progressing quickly. We sat in triage while I made repeated trips to the desk, trying to signal that biology wasn’t waiting for administration.

By the time we were moved into a delivery suite, Vera was desperate for the epidural. That required a doctor’s sign-off. When a doctor finally arrived, he looked at the monitors and told us—matter-of-factly—that we were out of time. The baby was coming.

The look on Vera’s face said it all. It was the shock of realizing she was helpless.

Our daughter arrived healthy. Vera was healthy. Everything else felt insignificant compared to that relief.

Then came the second shock.

Our “gold standard” insurance plan stipulated a single $600 co-pay per pregnancy. But 48 hours later, I was handed a bill for several thousand dollars in “non-covered” charges.

We hadn’t seen this in 2015 or 2017. No one had flagged it. We paid it anyway, because who wants to fight a billing department while holding a newborn?

A few weeks later, another bill arrived. Then another. Some from Johns Hopkins, some from the practice, some from third parties. The codes were indecipherable. The amounts seemed random. We spent hours on the phone with Johns Hopkins and CareFirst. Each call was a labyrinth of contradictory explanations. The only thing that was clear was the threat: pay this or we will damage your credit.

Our daughter is nearly three years old. We received what we hope is the final bill less than a year ago. That “$600 pregnancy” cost us over $4,000 and dozens of hours of stress. Same hospital. Same parents. Better insurance. But a fundamentally different experience.

We didn’t change. But the systems we relied on began behaving differently.

When Following the Rules Stops Working

I started seeing this pattern everywhere once I knew what to look for. There’s a particular kind of confusion that comes from doing what you were told to do, and having it not work. Not fail spectacularly. Just… not quite add up the way it was supposed to.

You went to a good college. Maybe you took on debt for it, because that was the responsible investment. You got the degree that opened the doors. But the doors led to jobs that don’t quite cover the loan payments plus rent in the cities where those jobs are. You’re doing fine. You’re just… not building anything.

Or you took one of many other paths.

You learned a trade. You got licensed. You bought the truck and the tools. The work is steady. The customers are there. But the margins keep shrinking, between insurance, materials, fuel, permits. You’re working more hours to stay in the same place. The effort adds up. It just doesn’t compound anymore.

These aren’t always crisis moments. But they are a steady friction. You double-check the math. Then you do it again.

You’re not imagining it. But you also can’t quite explain what changed or when. And the disorienting part isn’t the shift itself. It’s that you can’t figure out what you should have done differently.

The Shift from Unfair to Arbitrary

There’s a difference between something being unfair and arbitrary.

When something is unfair, you know what you’re fighting. You might not like the rules, but you know what they are. The system may be imperfect but it is at least legible, meaning you can see, understand, and predict the mechanism connecting your actions to results. You can plan your life. Or see what needs to change.

When you’re interacting with legible systems, you have agency: a belief in cause and effect, that your actions can reliably influence outcomes.

But when a system loses coherence and becomes opaque, your life starts to feel arbitrary. Arbitrary isn’t the same as random. You can tolerate randomness. A lottery is random. A storm is random. Luck is part of life. Arbitrary on the other hand feels like outcomes are the result of whim, like you’re playing a game but nobody’s told you the real rules.

A journey through the airport provides an intuitive example of these contrasts.

Unfair is watching the PreCheck lane move ten times faster simply because they paid the fee. Randomness is being selected by a TSA computer algorithm for additional screening. Arbitrariness is walking through a checkpoint where you are yelled at for taking your laptop out of the bag, when at the connecting airport three hours earlier, you were yelled at for leaving it in. Inconsistency, or the agent’s personal preference, masquerades as safety, a pretext that silences any objection.

Arbitrary outcomes impose a particular psychological tax for most of us, for two specific reasons.

First, our brains are wired to be prediction machines. We constantly create models: “If I take my laptop out, I won’t get yelled at.”

Arbitrariness destroys that consistency. You cannot automate your behavior. Instead you have to be hyper-vigilant every single time, because previous success does not guarantee future safety. This leads to chronic anxiety and learned helplessness.

Second, arbitrariness is inherently hierarchical. It is a form of social dominance that subtly humiliates you by rendering your compliance irrelevant. It triggers the same neural pathways as physical pain.

Most of us can tolerate bad luck. That is just the math of living. It is much harder to tolerate arbitrariness. Bad luck implies we are subject to the universe. Arbitrariness implies we are subject to a whim. It breaks the basic psychological contract of any society: the promise that if you follow the rules, you will be safe.

Arbitrary has become the default setting in health care. But once you’re tuned into the pattern, you can see it creeping into most systems we rely upon, across the public and private sector.

The Data Says You’re Fine. The Checkbook Says You Aren’t.

There is a specific kind of disorientation that comes from being told, repeatedly, that things are going well when it feels like you’re falling behind.

You turn on the news and see charts showing a strong economy. Unemployment is low. Aggregate wages are up. Consumer spending is robust. The official dashboard says the system is working as predicted.

But your grocery bill is higher than it was a year ago. Your car insurance premium jumped without an accident—an arbitrary spike you couldn’t have predicted. The rent on your “starter” apartment now rivals what your parents once paid for a four-bedroom house.

The problem isn’t that the data is fake. It’s that it measures a version of the economy that smooths over the parts of life that feel most immediate. Televisions are cheaper. Phones are better. But childcare now costs more than in-state college tuition. Housing absorbs a growing share of income.

The math no longer works. A second paycheck is no longer a choice. It starts to feel like the price of entry to stability.

We are living in a split-screen reality. On one side is the aggregate, the averaging machine that says the country is prosperous. On the other side is the lived experience, the specific, increasingly illegible reality where the math simply doesn’t work month to month.

When that gap grows wide enough, hearing that the economy is “strong” begins to feel like an accusation. If the economy is doing so well and you are still falling behind, the problem must be you.

That assumption is powerful. And isolating. It keeps people quiet just when we most need to compare notes.

The Scramble for Stability

But quiet doesn’t mean inactive. When things feel increasingly arbitrary, you start coping in various ways, even if you don’t realize it.

The first thing you do is push harder. You replace trust with hyper-vigilance.

You save every email and document every conversation. You spend an evening decoding airline fare classes just to confirm ‘Basic Economy’ allows you to sit next to your toddler. You burn hours reverse-engineering a job application to guess the keywords that will convince an algorithm you are a human worth speaking to.

Every interaction with a formal system now carries a hidden tax of time and attention that no one used to budget for. You are trying to force an opaque system to become legible through sheer effort.

The second thing you do is find the workaround.

Eventually, you stop assuming the system works as written and start insulating your life using whatever leverage you have—money, connections, or both.

If you have the capital, you buy the insulation. You subscribe to One Medical to bypass the urgent care waiting room. You pay the expeditor for the passport and the consultant for the college application.

If you have the network, you leverage your relationships. You call in a favor from a friend who knows the inspector. You find the “hack” to getting a human on the phone because you know someone on the inside.

And here is the strange part: it feels like keeping up, not getting ahead. You are paying a premium—and burning favors—just to recreate the baseline stability your parents took for granted. You are paying up to keep arbitrariness at bay.

This is a blind spot in how we talk about affordability. Official measures track the sticker price of the service—the premium, the co-pay, the tuition rate. But they assume the system works as designed. They capture the $600 co-pay you planned for, not the $4,000 in unknowable bills that eventually arrived, and certainly not the mental tax of the hyper-vigilance.

But paying up creates a powerful illusion. It allows us to mistake our private solutions for general stability.

When the grid fails, your lights stay on. When the ER is overflowing, your app gets you a video chat. We have built a Premium Tier that hides the friction of the standard tier from the people who can afford to bypass it.

If you are reading this, chances are you are one of them. I certainly am.

But even that insulation is cracking. We can feel the arbitrariness seeping in, reminding us that we merely deferred the problem. Our sense of safety is strictly a rental.

But what if you can’t buy the workaround, and you don’t know anyone on the inside?

This is the part that should give us pause. The workarounds buy time and reduce exposure. But for most Americans, there is no buffer. They are fully exposed to the whims of the system.

For them, the extra effort collides with diminishing returns. People are exhausting themselves trying to recreate a sense of security that used to emerge naturally from doing your part.

When everything feels arbitrary, something else begins to shift.

People quietly stop assuming things will work. They stop predicting success. They shrink ambition. They stop viewing the extra shift as a way up, and accept it as the price of staying put. They accept delays, denials, and degradation as normal. They tell themselves this is just how things are now.

Life starts to feel like an accumulation of documentation, contingency planning, and anxiety that no one consented to, and no one budgets for. These adaptations are rational. But eventually, they quietly change the relationship people have with the systems they depend on

That energy hardens. It looks like anger. It sounds like grievance. But underneath it is something more basic: the strain of living inside systems that no longer convert responsibility into security.

The Rationality of Distrust

We spend a lot of time wringing our hands over the collapse of systemic trust. We look at the polling numbers—the historic lows for Congress, the media, the medical system, the banks, corporate America—and we treat it as a crisis of attitude. We are told people have become cynical, or polarized, or misled by bad information.

Viewed through the logic of the arbitrary, the collapse of trust reads less like a mood shift and more like a math problem.

Trust is simply a measure of predictive safety. When the inputs—paying the premium, doing the work—stop producing reliable outputs, the prediction model fails.

The data bears this out.

Financial strain is now a stronger predictor of systemic distrust than ideology. When Americans report struggling financially, their trust in federal agencies and corporations drops by double digits regardless of their party affiliation. The stress is universal because the bills are.

Crucially, this distrust concentrates exactly where the mistakes are most costly. The data shows the widest trust gaps in rural and lower-income communities—areas where the margin for error is often zero.

Even where the system appears to work, the fear persists. In recent polling, 69% of Americans reported that it was very or somewhat easy for them to get the health care they needed. A surface read would indicate an important system is broadly working.

Yet respondents to the same survey were given an open-ended follow-up asking them to describe their experiences with healthcare in their own words. The responses were dominated by fear of opaque billing, unexpected costs, wait times, and rushed care. On paper, they have coverage. In practice, they are bracing for impact.

And where trust hasn’t vanished it has been rerouted. In qualitative surveys, people repeatedly describe subscriptions, side hustles, and personal networks as more reliable than formal systems.

That is pattern recognition, not cynicism.

What becomes visible in the current moment is a rational adjustment to a new reality, not a sudden loss of faith. People have discovered, through the steady accumulation of friction, that the systems meant to provide stability have become a primary source of instability.

To ask people to “restore their trust” without fixing the underlying incoherence is to ask them to ignore their own survival instincts. The unease people feel is not imagined. It is the correct answer to a new math. It measures exactly how much risk has been quietly offloaded from the system onto you. This isn’t just a run of bad luck. It is a structural shift that has been compounding for years. And the loss of agency is only one manifestation of this structural shift.

Where else can we feel this structural shift in our lives, what might be driving it, and how could we restore coherence to our social contract. This is the work we are beginning together in America Dreaming.

Suggested Sources

I. The Science: Why “Arbitrary” Hurts

• On Procedural Justice Theory (The “Legibility” Argument): Tyler, T. R. (2006). Why People Obey the Law. Princeton University Press.

• On Predictive Coding & Prediction Error (The “Hyper-Vigilance” Argument):: Friston, K. (2010). “The Free-Energy Principle: A Unified Brain Theory?“ Nature Reviews Neuroscience and as a secondary source: Mullainathan, S., & Shafir, E. (2013). Scarcity: Why Having Too Little Means So Much.

• On Social Pain Theory (The “Status Threat” Argument): Eisenberger, N. I., & Lieberman, M. D. (2004). “Why It Hurts to Be Left Out.“ Trends in Cognitive Sciences.

II. The Data: The “Split-Screen” Economy

• Struggling in America in Insights by Murmuration

• West Health Gallup Healthcare Indices 2025 Report by West Health

• For the long-term trend in institutional trust: Pew Research Center, “Public Trust in Government: 1958-2024” and Gallup, “Confidence in Institutions” Historical Trends (1973-2024)

Note on Methodology: Pew Research Center frequently analyzes and cites this Gallup data alongside its own government trust data in broader reports on institutional decline (e.g., Pew’s Trend magazine, Fall 2024, “Americans’ Mistrust of Institutions”). The graphic effectively combines these two “canonical” datasets to show the synchronized collapse across sectors.

Humans are imperfect.

Almost all of us are doing the best we can with what we have. We all make mistakes. We all fail to live up to our own intentions. We all suffer in ways we don’t fully share. Most of us are lonelier than we admit. We lose our tempers and regret it. We rationalize decisions that sit a little uneasily with us. We love our kids deeply and still worry we’re failing them. We want dignity and status in the communities around us. We covet nice things, even if we define “nice” differently.

This isn’t a cynical view of humanity. It’s a compassionate one.

It means that any serious attempt to understand how people work together—in teams, in institutions, or across society—has to start with grace. Not because everyone is right. But because almost no one thinks they’re the villain in their own story.

That belief shapes everything I’m trying to do here.

I almost never encounter “bad people.”

What I encounter, over and over, are thoughtful, earnest, capable humans making decisions that make sense locally—inside the incentives, cultures, and constraints they’re operating within—but that add up to outcomes that feel deeply wrong at a system level.

That gap is where my curiosity lives.

I’m much less interested in arguing about who’s right or wrong than I am in stepping back and asking: what kind of system would produce this behavior, at scale, from people who generally mean well?

Curiosity before judgment. That’s the posture of this project.

I’m going to make claims you may disagree with. If you do, tell me where your direct experience contradicts mine. That’s how this works.

Let me make that concrete with a few examples that are personal to me.

My grandfather helped deregulate the stock market. It was, at the time, a defensible decision. Markets were rigid. Capital allocation was inefficient. Risk was mispriced in ways that hurt growth and innovation. He wasn’t wrong.

But over time, a necessary corrective metastasized into something else entirely. Financialization stopped being a tool and became a culture. Incentives shifted. Time horizons shortened. Risk didn’t disappear, it moved to less visible places. We gained efficiency and liquidity, and we lost something harder to name: a lived relationship with real risk and long-term consequence.

I loved my grandfather. I still do. And I can hold two ideas at once: that his generation built something genuinely valuable, and that we, as inheritors of that system, waited too long to take responsibility for what it became.

We enjoyed the upside, but we delayed the maintenance. And now the repair work is harder.

My best friend’s father helped draft the Endangered Species Act. A remarkable achievement. A moral intervention that insisted the needs of natural systems mattered, even when markets didn’t account for them.

But over decades, that tool hardened. It became absolute in places where tradeoffs were unavoidable. In some cases, it now stands in the way of building the very infrastructure we need to address climate resilience at scale.

I see the pattern in my own work. For the last fifteen years, I’ve advised and invested in startups with the hope that high-growth companies could align with the public interest.

The bet was simple: build useful things—cleaner energy, better health diagnostics, better tools for learning—and the returns would follow.

But I watched a recurring mismatch between intent and structure. The funding model rewarded speed and scale, even when the work required patience, trust, and service. Founders who wanted to stay close to a community often felt pressure to change the aperture, raise prices, cut support, or monetize attention.

No villains. Just incentives that made certain choices feel “rational” in the moment, and left a trail of second-order costs.

That same trap—where immediate pressures crowd out the wider view—defines how we currently govern.

We have lost the capacity to look at the whole picture. Instead, we react to specific fires as they arise. We stop solving problems and start managing symptoms. Take the debate over student debt.

When broad student debt relief was announced, I understood the impulse. Higher education has degraded. People thought they were signing onto a fair deal: take on debt, get skills, gain stability. But it’s become incoherent for many. That produces real grievance.

But here’s the harder truth: that grievance is not unique.

If you grew up in a coal town in Appalachia that was hollowed out by shifts in energy systems and global trade, your deal collapsed too. If you grew up in an urban neighborhood with a failing school system, your deal never really existed. If you’re a farmer sitting on land that looks valuable on paper but yields a shrinking household income year after year, your risk is very real and very present.

By elevating one grievance over others, but without doing the hard work to reform the underlying system that produced all of them, we didn’t heal much. We accelerated polarization. We told one group “we see you” while implicitly telling others “wait your turn.” The system stayed broken. Trust eroded further.

Again: local sense. Global failure.

This pattern shows up everywhere once you start looking for it.

I became fascinated by regenerative agriculture not because it’s “green,” but because it’s an interesting systems problem. At some point, someone told me I needed to stop reading papers and start spending time with farmers. So I did.

I went to Iowa. I sat at kitchen tables. I rode in pickup trucks. I listened.

These farmers understood soil science, nitrogen cycles, rainfall variability, and commodity markets in extraordinary detail. One old guy told me, “Well hell, the combines drive themselves these days, but we still have to be in them. Leaves a lot of time for YouTube.”

They’d describe collapsing nitrogen absorption rates and erratic weather patterns, then say in the next breath, “Look, we don’t buy the climate change hoax, but we gotta do things differently.”

From the outside, that sounds like contradiction. From the inside, it sounds like people who are adapting rationally inside a system that’s been talking at them, not with them, for decades.

Then I’d fly to Climate Week. Panel after panel. Jargon-dense conversations about regenerative agriculture. Mostly ex-consultants. Not a farmer in sight. And the quiet subtext, often with genuine good intention, was that the farmers were the problem, not the solution.

That’s how trust gets destroyed. It’s not through malice, but through distance.

I love country music for the same reason I love systems thinking when it’s done right.

Country music is three chords and the truth. It doesn’t pretend the human experience is tidy. People love and divorce. Parents fail and try again. Jobs disappear. Kids still need Christmas presents. Sometimes people make one terrible decision and spend the rest of their lives regretting it.

That’s the America I recognize.

I want American Dreaming to sound like that. Not in tone, but in honesty.

This isn’t a newsletter about villains. It’s not a manifesto. It’s not a one-way broadcast from someone who thinks he has “the answer.”

It’s an invitation to step back together and ask why so many people, across class, geography, race, and profession, feel like they did what they were supposed to do, and the system didn’t deliver on its side of the deal.

I think our institutions used to translate effort and risk-taking into stability and opportunity more reliably than they do now. Meaning, if you took the risk and did the work, life usually got more predictable, not less. Not for everyone, some have always been unfairly excluded, but for most people. If that makes you pause, good.

We’ll unpack it slowly, concretely, and from multiple angles. Nothing here depends on you taking my word for it.

American Dreaming is a place for diagnosis before solutions. For tradeoffs rather than ideology. For looking past our own fence line, especially for those of us who know our own corner of the world deeply but rarely see how things look from where others are standing.

If it works, the ideas here won’t feel like mine. They’ll feel like insights you arrived at yourself, with some healthy debate along the way.

That’s the journey I want to take. Together.