The Reason Left

Boyers, Pennsylvania. About an hour north of Pittsburgh. A small town surrounded by farmland. Underneath it, 220 feet below the surface, the Office of Personnel Management operates the federal government's Retirement Operations Center inside a former limestone mine.¹

U.S. Steel began cutting limestone there in 1902 to feed its steel mills. By the early 1950s, mining had stopped. A private company turned the empty caverns into a vault, sold as safe from the weather and the Soviets. In 1960, the government moved its retirement records underground. It needed space, and the vault had plenty. Limestone walls, 220 feet of earth, constant temperature. Underground storage made sense. The processing followed the paper down.²

The mine grew. By 2025 it held about 26,000 file cabinets and about 400 million pieces of paper. Workers drove golf carts through caverns that looked, in the words of one reporter, like a supermarket full of five-drawer file cabinets and people in business casual. Easy-listening music played at their desks. A pizza delivery person with a security clearance arrived each day. In winter, workers entered in the dark and left in the dark. No open flame was permitted. It was still a mine.³

The retirement process matched the facility. A federal employee submitted paperwork to their agency. The agency shipped it to Boyers via FedEx. The documents went into a manila envelope. Computer files were printed out and placed inside the envelope. The file moved to an adjacent cavern where workers called, emailed, faxed, and badgered other federal agencies for missing paperwork. Benefits were calculated by hand. A single missing signature could set the process back weeks.⁴

The official processing target was sixty days. In November 2025, paper claims averaged ninety-four. During heavy backlogs, retirees waited three to six months. Some longer. They received interim payments of sixty to eighty percent while they waited, but the first interim check took six to eight weeks to arrive. Federal employees who had spent thirty years serving the government filed their retirement papers and entered a queue inside a cave.⁵

The Cold War ended in 1991. The Soviet Union dissolved. The nuclear threat that justified underground record storage passed more than three decades ago.

The mine remained. The paper process inside it remained. Over thirty-plus years, at least three separate modernization initiatives were launched. More than $100 million was spent. All of them failed. The Government Accountability Office cited "big bang" approaches to modernization, no dependency tracking between project components, no process for identifying risks, and inconsistent OPM leadership. Each new administration started over rather than building on what came before. The cost per claim increased. The process stayed.⁶

The fix

In early 2025, two engineers walked into the government to move federal retirements out of the mine and onto the internet. One had spent a decade at Airbnb. The other came from startups in manufacturing and AI. They used industry-standard web technologies. Proper table indexing. Efficient database joins. Batch execution. They built retire.opm.gov in approximately six months. It went live on June 2, 2025.⁷

Months earlier, in February 2025, OPM had pushed its first retirement claim through end to end without paper. No paper. No FedEx. No manila envelope. Completed in approximately two days.⁸

In July 2025, OPM stopped accepting new paper applications, though paper kept arriving for another year. By late 2025, digital claims averaged thirty-eight to forty days. The system was not perfect. A surge of federal employees leaving government overwhelmed the platform, and the overall backlog grew even as per-claim processing times fell. But the constraint had shifted from architecture to volume. The architecture was never the obstacle.⁹

Thirty-plus years. More than $100 million. Three failed modernization attempts. Then two engineers and six months. The problem was never hard. It was never questioned.

The pattern

The same mechanism operates in every major infrastructure project on the continent.

Fifteen years

The project you could drive past. For fifteen years, you could not miss it.

Construction began in the summer of 2011 on a nineteen-kilometer light rail line across Toronto's busiest crosstown corridor. The original opening: 2020. Tall fencing and scaffolding sealed off storefronts along Eglinton Avenue. The street narrowed to one lane. Parking disappeared on the north side. Sidewalk passages shrank to one meter between shop doors and construction barricades. Business revenue along the corridor dropped roughly fifty percent.¹⁰

More than 140 businesses closed. In the neighborhood known as Little Jamaica, 300 small businesses were gone. The Business Improvement Area declared a state of emergency. Its chair told reporters that sixteen years of construction had changed the neighborhood's identity permanently.¹¹

The opening pushed to 2021. Then 2022. Then 2023. Leaked internal documents revealed the contractor consortium had "no credible plan" to finish. An 800-meter stretch of track was found to be millimeters out of alignment, enough to derail a train. The entire stretch had to be redone. The signaling and train control software required six versions before it worked. The original budget of $4.6 billion climbed past $13 billion, a figure that folds in thirty years of maintenance.¹²

The Eglinton Crosstown opened on February 8, 2026. Fifteen years. Billions over the budget it started with. Every excuse had its own name: contractor disputes, water infiltration, design approval delays, signaling issues. None of them explained why a single transit line took fifteen years to build.¹³

Ninety-eight years

The Eglinton Crosstown tested how long a city could watch construction outside its window. The Second Avenue Subway tested whether anyone remembered why the project existed.

First studied in 1919 as part of New York's subway expansion plan. The Board of Transportation approved a route in September 1929 at an estimated cost of $99 million, with completion projected for 1941. The Depression stopped it. A new plan surfaced in 1939. World War II stopped that. Between 1940 and 1942, the city demolished the Second Avenue Elevated, the rail line it was supposed to replace. Workers pulled down the steel structure above the street. Riders looked up and saw open sky where their train used to run. The underground replacement was not built. They got nothing back.¹⁴

Voters approved a bond measure in 1951. Most of the money went elsewhere. Tunnel construction finally began in October 1972. In 1975, all work stopped. New York was bankrupt. Three short concrete tubes sat unfinished beneath Second Avenue for thirty-two years. No trains. No platforms. Just empty tunnels, sealed off, while the city above them lost people and then won more than a million back.¹⁵

Construction resumed in 2007, starting from those same 1970s tunnels. Phase 1 opened January 1, 2017. Three stations over about 2.9 kilometers. Cost: $4.5 billion. Roughly $1.5 billion per kilometer. The most expensive subway construction in the world per kilometer. Eight to twelve times more expensive than comparable projects in Paris, Seoul, and Berlin. Phase 2, three more stations, is budgeted at $7 billion with a target of September 2032.¹⁶

From first proposal to first completed phase: ninety-eight years. Three stations.

Excuse: underground construction in Manhattan is uniquely difficult.

Zero kilometers

The Eglinton Crosstown and the Second Avenue Subway spent decades building something. California High-Speed Rail has spent decades preparing to.

Voters approved Proposition 1A in November 2008. Original budget: $33 billion. Los Angeles to San Francisco by 2020. Trip time: two hours and forty minutes. The cost as of the February 2026 business plan: $126.2 billion. Zero operating kilometers. After seventeen years and more than $18 billion in public funds spent, no track has been laid. As of mid-2026, the first rail was projected to be installed by year's end.¹⁷

The active construction is in California's Central Valley. Flat farmland between Merced and Bakersfield. Almond orchards and cotton fields. The easiest terrain on the entire route. Eighty miles of elevated guideway structure are complete. If you drive along Route 99, you can see them: concrete pillars rising out of the farmland, carrying nothing. No track on top. No wire above. Just empty structures standing in the sun, waiting. The mountain crossings to reach the Bay Area and Los Angeles, the segments that would present actual engineering challenges, have not begun. Spain built Madrid to Seville, 471 kilometers of high-speed rail, in about three years. Japan built the Tokaido Shinkansen, 515 kilometers, in five and a half years, in 1964. California has spent seventeen years on flat farmland and has not laid track.¹⁸

In July 2025, the federal government terminated approximately $4 billion in grants, finding that the authority could not meet its obligations and had no "viable path" to complete even the Central Valley segment by 2033.¹⁹

Excuse: environmental review. Land acquisition. Funding gaps.

Twenty-six tons

The other three projects cost time and money. The Big Dig cost a life.

On July 10, 2006, at approximately eleven p.m., Milena Del Valle was riding in the passenger seat of a car in a connector tunnel beneath Boston. A forty-foot section of concrete ceiling separated from the roof above her. The concrete and its hardware weighed twenty-six tons. It fell. She was thirty-eight years old. Her husband, who was driving, escaped with minor injuries.²⁰

The National Transportation Safety Board determined the cause. A fast-set epoxy adhesive had been substituted for the specified standard-set product during construction. The fast-set epoxy had poor creep resistance. Over the years since installation, it had slowly pulled away from the ceiling anchors. A material substitution that no one caught. Anchors that slipped in 1999 and that no one kept watching. A ceiling that no one inspected until it was on the ground.²¹

The tunnel was the Central Artery/Tunnel Project, known as the Big Dig. It replaced Boston's elevated highway with an underground expressway. Original budget: $2.8 billion, completion targeted for 1998. Final construction cost: $14.8 billion. With interest on the debt, roughly $24 billion. Payments continue through 2038. Construction ran from December 1991 to December 2007. Sixteen years. Five times the original estimate.²²

The Big Dig was supposed to solve Boston's traffic. The elevated highway carried 200,000 vehicles a day on infrastructure designed for 75,000. The tunnel added capacity. Boston now ranks fifth worst in the United States for traffic congestion. Eighty-three hours per driver per year. The new capacity filled. Induced demand replaced the cars the tunnel was designed to move. Twenty-four billion dollars to move the congestion underground and then fill it back up.²³

Excuse: most complex urban highway project ever attempted.

Four projects. Four cities. Four decades. The excuses are different every time. Contractor disputes. Geology. Environmental review. Complexity. If the cause were the excuse, each project would fail for its own unique reason and the pattern would not repeat. But the pattern repeats. Across cities, decades, political parties, and engineering challenges that share nothing in common except the outcome. The excuses change. The outcome does not. The excuse is the costume. The mechanism underneath is the same one that kept retirement records in a cave.

The mechanism

Process accumulates because the person who adds a step never bears the cost of that step.

The OPM administrator who maintained the paper retirement system did not wait months for a pension check. The retiree did. The Metrolinx supervisory board that approved each Eglinton delay did not own a hair salon on Eglinton Avenue. The 300 business owners in Little Jamaica did. The legislator who designed California's environmental review process does not commute on a train that does not exist. The commuter does. The contractor who substituted a fast-set epoxy in a Boston tunnel ceiling did not drive under it at eleven p.m. on July 10, 2006. Milena Del Valle did.

In every case, the person who added or maintained the process captured a benefit. Political cover. Risk reduction. Career protection. Compliance. In every case, the person who bore the cost had no mechanism to remove it. The retiree could not redesign the system. The business owner could not overrule the board. The commuter could not shorten the review. The woman in the car could not inspect the ceiling.

The feedback loop is open. The architect of the process and the victim of the process never meet.

Environmental review exists because developers once dumped chemicals in rivers. Safety inspections exist because buildings once collapsed. The process was there for a reason then, and some of it is there for a reason now. The question is not whether process should exist. The question is whether anyone checks which layers still serve the reason they were created for, and which survive on the sentence alone.

Nobody checks because nobody has to. The person who could remove the process bears no cost for keeping it. The person who bears the cost has no authority to remove it. That is why process only accumulates. There is no structural force pushing in the other direction. Each layer was rational when it was added. Each layer served a purpose for the person who added it. And each layer remained after the purpose passed because the person who could remove it had no reason to, and the person who needed it removed had no power to.

It is the same separation that appears in every system where decisions and consequences live in different rooms. At Boeing, the engineers who knew were separated from the executives who decided. In the mine, the retirees who waited were separated from the administrators who could have changed the process at any point in three decades. On Eglinton Avenue, the businesses that closed were separated from the consortium that had "no credible plan" to finish. The rooms change. The separation does not.

Process decays on its own in exactly one scenario: when the person who maintains it also pays for it. Remove that condition and process is permanent. It does not erode. It does not expire. It does not get audited. It accumulates, year after year, because "the process is there for a reason" is a sentence that nobody is allowed to finish. There for a reason. But nobody checks whether the reason is still there.

The comparison

Two airports

Both wealthy democracies. Both airports. Both decided in the same era. One opened on time. One opened nine years late.

South Korea decided to build Incheon International Airport around 1990. The country's GDP per capita was roughly $6,000 to $8,000. Gimpo Airport, boxed in by urban sprawl, could not handle the demand. Construction began in November 1992 on a site that did not yet exist. Between Yeongjong and Yongyu Islands in the Yellow Sea, there was nothing but shallow tidal flats. Engineers built 13.4 kilometers of seawall to hold back the ocean, dredged millions of tons of material, and stabilized soft marine soil before a single runway could be laid. They built the airport on water.²⁴

Incheon opened on March 29, 2001. On time. On budget. Phase 1 cost approximately 5.6 trillion won, about $4.4 billion. Phase 2 opened in 2008. Phase 3 opened in January 2018, in time for the PyeongChang Winter Olympics. Phase 4 completed in December 2024. All four phases delivered on schedule. Incheon handled 74 million passengers in 2025, an all-time record. It was named the world's best airport by the Airports Council International for twelve consecutive years.²⁵

Germany decided to build Berlin Brandenburg Airport in the early 1990s, shortly after reunification. The country's GDP per capita was roughly $25,000 to $30,000. Construction began in September 2006 on solid ground at the edge of the existing Schoenefeld airfield. The airport was supposed to open in October 2011.

It opened on October 31, 2020. During a pandemic. Nine years late. Seven opening dates were announced and missed. The original budget of approximately EUR 2 billion reached EUR 6 to 7 billion. Three times over.²⁶

After the canceled 2012 opening, inspectors found tens of thousands of defects. The smoke extraction system, the single largest cause of delay, did not work. It was designed to pull smoke down and out of the terminal during a fire, and in tests it could not. The person responsible for the design held the position for eight years. He was not a qualified engineer. He was a draftsman whose business card incorrectly listed him with an engineering credential. Nobody checked his qualifications for eight years. Kilometers of cable had been laid wrongly. Some lights could not turn on. Others could not turn off.²⁷

A departmental construction manager received EUR 150,000 in cash at a highway rest stop on the A24 near Wittstock. In exchange, he approved more than EUR 60 million in unverified payments to a contractor whose work he never inspected. He was convicted and sentenced to three and a half years. The contractor, Imtech Deutschland, filed for insolvency. A senior engineer who had blown the whistle collapsed at work in 2015. An expert report later said he may have been poisoned through his coffee.²⁸

South Korea was poorer. The engineering challenge was harder. They built on the ocean. Germany built on solid ground. The airport that was built on reclaimed tidal flats between two islands opened on schedule and became the best in the world. The airport built on solid ground in one of the most capable engineering nations on earth came in nine years late with tens of thousands of defects, a fire safety system designed by someone who was not an engineer, and a construction manager who accepted a bribe at a highway rest stop.

This is not a North American problem. Germany is not Canada. Berlin is not Sacramento. The mechanism is the same. Process accumulated. Accountability was distributed across a three-way ownership structure of the federal government and two state governments. Politicians chaired the supervisory board. Decision-making fractured. The feedback loop between the person who approved each delay and the passengers who would not fly for another nine years was open.

After they open

The airport comparison closes one escape route. The next one closes another. The argument that infrastructure projects go wrong during construction is familiar. What is less familiar is what happens after they open.

Kansai International Airport sits on an artificial island in Osaka Bay, five kilometers offshore. It opened in September 1994. It handles approximately ten million pieces of luggage per year. In more than thirty years of operation, it has never lost a single one. Zero. Tsuyoshi Habuta, a chief baggage handler who has worked there for seventeen years, explained it simply. Luggage should not go missing, he said, because it is precious to passengers.²⁹

The handling is visible. At the carousel, every suitcase arrives with its handle facing outward toward the passenger. Workers dry off bags caught in the rain and clean them before they come through. Bags from arriving flights reach passengers within fifteen minutes of the plane touching down. No scanning algorithm achieves this. No conveyor system enforces it. It is a standard, maintained by a person, applied to each bag individually, ten million times a year.³⁰

Haneda Airport has been named the world's cleanest airport for ten consecutive years. It is ranked third in the world overall. In fiscal 2023, the Tokaido Shinkansen averaged a delay of 1.6 minutes per train, counting delays from natural disasters. In 2018, a JR West train left Notogawa station on the Biwako Line twenty-five seconds ahead of schedule. The operator issued a formal public apology for the "truly inexcusable" early departure.³¹

When a Shinkansen arrives at Tokyo Station, a cleaning crew in bright red uniforms lines up on the platform and bows to the arriving passengers. Then they board. One person per car, each car holding roughly 100 seats. They have seven minutes before the train departs again. One and a half minutes to collect trash. Thirty seconds to rotate every seat to face forward. Four minutes to sweep and wipe every surface. One minute for a final check. They do this 120 times per day. CNN called it the seven-minute miracle. It is not a miracle. It is a process that someone audits.³²

This closes the escape route of "well, at least things work fine once they are built." The same precision that builds the infrastructure runs the infrastructure after it opens. Kansai has never lost a suitcase because the standard does not stop at the blueprint. It reaches the person who wipes down your bag before you see it.

Three minutes

The mine processed government services in a cave. Estonia processes them in three minutes.

In 1991, Estonia regained independence from the Soviet Union. The country had almost nothing. Population: 1.3 million. No functioning digital infrastructure. No legacy systems worth preserving, because Soviet systems were not worth preserving. That absence turned out to be the advantage. In 1992, Finland offered Estonia its old analog telephone exchange. Estonia refused. It chose to skip analog entirely and build digital from the ground up.³³

In February 2000, the Estonian parliament declared internet access a human right, the first country to do so. The following year, X-Road went live. It is a decentralized data exchange layer connecting every government database. No central hub stores all the data. Each ministry manages access to its own systems, but every system can talk to every other system. It handles about 2.2 billion transactions a year, by e-Estonia's count.³⁴

The results accumulated. Tax filing: three to five minutes, pre-filled from employer and bank data, ninety-five percent of citizens file online. In the United States, the average filer of the main federal return spends about thirteen hours. Digital prescriptions, 100 percent electronic since 2010. Business registration in five to fifteen minutes, no notary, no office visit. Online voting since 2005, the first legally binding internet elections in the world. By 2023, more than half of all parliamentary votes were cast online. As of December 2024, Estonia became the first country where 100 percent of government services are available digitally.³⁵

In April 2007, the system was tested. A coordinated cyberattack hit government websites, banks, and media for nearly three weeks, widely regarded as the first major act of cyber warfare in the world. The digital infrastructure held. Estonia rebuilt what was damaged and NATO established its Cooperative Cyber Defence Centre in Tallinn in response. The system that started with nothing survived the thing designed to destroy it.³⁶

Three escape routes closed at once.

"Only rich countries can do this." Estonia's GDP per capita is less than half of the United States'.

"Only small countries can do this." India's Aadhaar system has enrolled more than 1.4 billion people with biometric identification. The scale objection does not hold.³⁷

"Government is just slow." Government is slow here. It is not slow everywhere.

The cave

The process was there for a reason. In 1960, storing records underground made sense. Nuclear war was not hypothetical. The Soviet Union tested a fifty-megaton bomb the following year. Underground, inside stone, 220 feet of earth overhead. That was the right answer.

In 2024, it was not the right answer. It had not been the right answer for decades. Nobody removed it because nobody who could remove it bore the cost of keeping it. The retiree who waited months for a pension check had no authority to change the system. The administrator who could change it had no cost to bear. The process sat between them, 220 feet underground, undisturbed. It was easy to ignore, and nobody who could fix it had a reason to.

Then someone questioned it. Two engineers. Six months. The timeline collapsed. More than $100 million in failed modernization attempts over thirty years, and the solution was two people who understood databases and cared about the outcome. The problem was structural, not technical. The people who experienced the delay and the people who could end the delay were separated by a gap no feedback could cross.

Every project in this piece failed the same way. The engineering was not impossible. The process that surrounded it was never questioned. Each layer was added for a reason. The reason passed. The layer remained. The cost accumulated downstream, year after year, paid by people who had no voice in the system that produced it. Commuters on Eglinton Avenue. Riders who waited ninety-eight years for three subway stations. Taxpayers funding $126 billion for a train that has not laid track. A woman in a car in a tunnel that used the wrong glue.

Incheon opened on time because the system connected the person who built the airport to the outcome of whether it opened. Estonian tax returns take three minutes because someone decided that waiting was not inevitable. It was a design choice. Kansai has never lost a suitcase because the standard does not stop at the blueprint. It reaches the last person in the chain, and that person bears the responsibility of whether the bag arrives.

The process is there for a reason. But nobody checks whether the reason is still there.

The mine is still in Boyers, Pennsylvania, 220 feet underground. It was sold as safe from the Soviets, and it outlasted them. The process that lived inside it was not designed to survive a question. It did not.