On 24 January 1942, the Office of Production Management was abolished by executive order. Its director general had spent twenty months trying to arm the United States through an agency with no legal power to make anybody do anything. Like the Articles of Confederation (and every other authority with no legal power), it failed.
Four days later that same man was commissioned as a lieutenant general in the United States Army.
He was sixty-two. He had never worn a uniform, never been to West Point, never served a day in any military. He was a Danish immigrant who came through New York harbor in February of 1900 and took a job in a shipyard on the Harlem River, where the foreman wrote his name down wrong because he couldn’t be bothered with all those initials. So far as the published record shows, he is the only civilian ever to enter the US Army directly at three-star rank with no prior service, certainly the only immigrant, and in eighty-four years nobody has produced a second example.
His name was William Signius Knudsen, and the reason Franklin Roosevelt put three stars on a man who had never saluted anybody is the subject of this piece.
In May of 1940, with the Germans in France, Roosevelt went on the radio and asked the country for fifty thousand military aircraft a year. America had built about three thousand the year before.
Time, covering the effort that October, summarized the reaction of every serious person in American industry in one sentence. The President had cried for fifty thousand planes. That, the magazine said, was an impossible figure.
America and the free world needed a man who didn’t believe in impossible. We got Bill Knudsen, who didn’t argue with the press and didn’t argue with the President. He simply went and built thirty-three thousand airplanes in twenty-one months, and without him, we would never have won WW2.
The man who would not be told
Knudsen was born in Copenhagen in 1879, apprenticed in a bicycle shop, and came to America at twenty-one with no grasp of the English to speak of. He worked at shipyards at a dollar seventy-five a day, then railroad shops, then a steel pressing company in Buffalo that Henry Ford bought outright in 1911 because he wanted the stamping expertise. Knudsen arrived at Ford the way a piece of capital equipment arrives. He came with the purchase, and within two years Ford had made him production manager of Highland Park, which at that moment was the most advanced manufacturing operation on earth.
Thirteen years off a boat, running the floor at Highland Park.
Then Ford handed him a problem everybody treated as a law of nature. Ford built every Model T in one place and shipped finished automobiles across a continent, which is an absurd way to move metal, because a finished car is mostly air and a crate of parts is dense. Everybody could see the waste. The accepted answer was that this is simply what it costs to run a national business out of one factory. It’s the way things had to be done.
Knudsen didn’t accept that. He went out and built branch assembly plants: ship the parts, assemble locally, sell locally. He built more than a dozen across the United States and three in Europe, and by 1915 he was running all of them plus Highland Park.
Then the First World War came, and Ford took a contract for submarine chasers. Ford had never built a ship, and shipbuilding in 1917 was a craft trade done one hull at a time in yards by men who had apprenticed to it. Ford proposed to build them indoors, on a line, out of standardized plate, in Michigan. Everybody in American shipbuilding thought this was nonsense and had three hundred years of practice on their side. Knudsen ran the production, and by October 1918 Ford was finishing a hull a day.
This same pattern repeats for the rest of his life. He didn’t win those arguments by proving the skeptics wrong in advance. He couldn’t have. He won them by building the thing and then pointing at it. No muss, no fuss, just results.
He fell out with Charles Sorensen in 1919; Henry Ford backed Sorensen, as Henry Ford generally did, and Knudsen left Ford in 1921. In February 1922 he went to General Motors at thirty thousand a year, and within weeks Pierre du Pont had him running Chevrolet at fifty.
Which brings us to the number that explains why Roosevelt picked up a telephone eighteen years later when America was in crisis.
Chevrolet in 1921 built 75,700 automobiles. Chevrolet in 1927 built 1,001,680.
Thirteen-fold in six years. And in 1927, Chevrolet took the sales crown away from Henry Ford.
Ford had invented the mass-market automobile and the methods behind it, and the settled wisdom in Detroit was that nobody would ever take that position from him. Knudsen took it with a better car at a competitive price, using the plants Ford had taught him to run.
He became GM’s executive vice president in 1933 and president in May 1937. In 1935 his compensation was $325,869, a figure we know because 1935 tax returns were reported in the press. Alfred Sloan made $374,505. Those were the two largest industrial salaries in the United States, north of seven million a year in today’s money.
On 26 May 1940, Roosevelt gave the fireside chat. The British Army was days from the beach at Dunkirk. The President asked for fifty thousand aircraft a year. The call to Knudsen went out the next day.
He took it, left the presidency of General Motors, and went to Washington to work for a dollar a year. Some records say he never drew even that. What his daughter said he told the family was that this country had been good to him and he wanted to pay it back.
Roosevelt revived a First World War body, the Advisory Commission to the Council of National Defense. Knudsen was one of seven commissioners. There was no chairman. Here is Time:
“The President had given the Commission no legal power to do anything. All they could do was advise, suggest, request—no more. They did not even have a boss.”
Eight months later the Office of Production Management made him Director General, and put a labor leader beside him as co-director with the Secretaries of War and Navy on the council above them both. The National Defense University’s history of the mobilization puts it flatly: Roosevelt made it an advisory body, so it was doomed to fail.
So here was the offer put to the highest-paid industrialist in America. Give up the money. Take a divided command. Accept no statutory power, no control over military procurement, and a President who designed it that way on purpose.
A reasonable man says no to that proposition, and to call me when you have an actual plan and the job has actual power.
Knudsen said yes in a day, and then did the job anyway, without the tools, because the alternative was that it didn’t get done. In troop leading procedures, we call this “start necessary movement.”
An impossible figure
Here is the country he was handed. In 1940, the US Army had fewer than 200,000 men. Troops on maneuvers carried broomsticks to simulate rifles and drove trucks with signs on them to represent tanks. Between 1918 and 1933, the United States produced a total of thirty-five tanks.
Thirty-five, in fifteen years. That number is laughable in the aftermath of a world war with another obviously brewing. Against that, FDR wanted fifty thousand aircraft a year, in an industry that didn’t exist at the required scale, with mostly obsolete tooling, and in a nation not at war whose Congress wasn’t going to declare one.
And the people who said it couldn’t be done weren’t being defeatist. They were doing the math. Engine plants take two years to build and another to learn to run. American aluminum production in 1939 was a fraction of what the program required. The entire American machine tool industry was shipping about a hundred million dollars a year to every customer in the country. And nobody had built military aircraft in volume since 1918. They weren’t describing an attitude. They were describing a supply chain, and every link in it was short.
Which is what makes Knudsen’s response the most interesting thing in his file.
He didn’t contradict the President in public, which would have been disloyal and useless. He didn’t accept the number either, which would have been a lie. He quietly set his own target at 35,000 aircraft, 25,000 for the Army and 10,000 for the Navy. And in October 1940 he made a public forecast: the United States would have built 33,000 aircraft by April 1942.
Run the actual production. July 1940 through March 1942 comes in right around 33,000 to 34,000.
He called it inside a thousand airplanes, eighteen months out, with no authority, in a country not at war, against a presidential number the press had already written off.
He didn’t take impossible for an answer. He also didn’t take fifty thousand for an answer. He went and found the real number, and then hit it.
Here’s how.
Aircraft production in the second half of 1940 was 3,611 airplanes; in 1941 it was 18,466, a five-fold increase inside his tenure. But the aircraft aren’t the answer to how.
In 1940 the United States spent about $2 billion on factory construction, and in 1941 more than $4 billion. Aircraft engine and propeller plants went from 41 in 1940 to 81 by 1943.
And then there’s the machine tools. At the start of 1940 there were 896,000 installed in American metalworking; by the end of 1944 there were 1,518,000. Shipments had run at roughly $100 million a year, every year, from 1929 through 1938. Between April 1942 and June 1943, they ran at $100 million a month.
At the end of 1939, only 29 percent of the machine tools in the country had been bought since 1929. Seventy-one percent of American industrial tooling was more than a decade old, because nobody buys a lathe during a depression.
So when Time asked him in October 1940 what the job actually was, he gave them the answer:
“From now on, it becomes a problem of tools and men.”
Not a problem of will. Not of money. Tools, buildings to put them in, and men trained to stand at them. He had identified the binding constraint on American power in nine words, and then spent six billion dollars in two years relieving it. It wasn’t impossible, the system just didn’t exist.
In 1946, when the Bureau of the Budget wrote the government’s own official history of the war, with no reason on earth to flatter a GM executive, this was the verdict:
“though greatly handicapped by the mounting problems of his weak and bifurcated organization, had made a great contribution, especially toward the expansion of facilities.”
Especially toward the expansion of facilities. The United States government, in its own history, says what Bill Knudsen gave the country was buildings and plant.
Look at what came out of them. Munitions production ran about $500 million a month in 1940; by 1944 it was over $5 billion a month, and the 1944 total was $61.3 billion in a single year. General Motors alone delivered more than $12 billion of war material, including 38,000 tanks and tank destroyers. Ford’s Willow Run built 8,685 B-24s and by 1945 was finishing one an hour.
One bomber an hour, at a plant that didn’t exist when Knudsen took the job, built by a company that made cars, to produce what was then the most complicated machine any nation had attempted to mass produce.
Every link in that short supply chain got longer. Aluminum multiplied, the engine plants doubled, and the men who didn’t exist in 1940 were trained, hundreds of thousands of them. None of it was impossible. All of it was simply undone, and the difference between those two words is the whole subject here.
The things they tell you are impossible now
Here is the current list of things that cannot be done, according to everybody.
In February 2022, the Pentagon published State of Competition Within the Defense Industrial Base. Its own document, about itself:
“Since the 1990s, the defense sector has consolidated substantially, transitioning from 51 to 5 aerospace and defense prime contractors.”
Underneath that: tactical missiles from thirteen primes in 1990 to three, fixed-wing aircraft from eight to three, surface ships from eight to two. Two companies in the United States build surface warships. There are two suppliers of solid rocket motors in this country, and ninety percent of American missiles come from three sources.
Shipbuilding. UNCTAD publishes world shipbuilding share annually. In 2024 China delivered 54.6 percent of the world’s commercial gross tonnage, South Korea 28, Japan 12.6. The United States delivered 0.04 percent.
On warships, we’re still ahead in the water: 3.11 million tons of combat tonnage against China’s 1.44 million as of 1 January this year. But across the two years ending December 2025, China built 636,000 tons of warship and the United States built 270,000. They’re building at better than twice our rate from a base less than half our size. We aren’t losing today. We’re losing the derivative, which is worse, because trajectories compound and inventories don’t.
Virginia-class submarines run at about 1.13 boats a year against a requirement of two, and 2.33 including the Australian commitment under AUKUS. In March, American 155mm shell production stood at 36,000 rounds a month against a target of 100,000, first due in October 2025.
CSIS has wargamed a Taiwan contingency roughly two dozen times, and in every iteration the United States exhausted its long-range anti-ship missiles inside the first week. Brian Schimpf, who runs Anduril, noted in May that the US has expended about thirty percent of its Tomahawk inventory: a decade of production, gone in three days. His summary was that American munitions production rates and stockpiles are at a dangerous level.
Munitions lead times run twenty-five to fifty-one months. GAO puts the average time from program start to fielded capability at over twelve years, and rising.
When Knudsen walked into Washington, the United States had roughly eight million people out of work, enormous idle plant sitting cold after eleven years of depression, and by a wide margin the largest manufacturing economy on earth. The country had slack, and mobilization largely consisted of picking it up.
We have no slack. Four percent unemployment, a manufacturing workforce shrinking since before most of you were born, a welder shortage in the submarine yards the Navy will describe at length if you ask, and no commercial shipbuilding industry to convert because we let it go overseas forty years ago.
They have a second argument that’s just as good. In 1940 the weapons were simple, and a company that built automobile transmissions could learn to build a Sherman tank. A modern interceptor is a flying computer with a rocket attached, built on semiconductors from a fab that takes five years and twenty billion dollars to stand up. You cannot convert a furniture plant into a missile line the way a piano factory became a glider works in 1942.
So the pessimists’ case is: no slack, no convertible plant, weapons an order of magnitude more complex, supply chains running through the adversary, and no mobilization consensus. Every piece of that is correct.
Here is what it proves. It proves this will be difficult, that it will cost money, and that the people who do it will have to invent things that don’t currently exist.
It does not prove it can’t be done, and the people making the argument generally slide from the first conclusion to the second without noticing they’ve moved.
In 1940, the supply chain was short at every link, the tooling was obsolete, nobody had built a military aircraft in volume in twenty years, and the entire analysis was correct right up to the point where a man who didn’t care about the analysis started pouring concrete.
Arsenal-1
Palmer Luckey was born in 1992 in Long Beach and homeschooled by his mother. At sixteen, he was building virtual reality headsets in the garage, financed by repairing and reselling broken iPhones. At nineteen, he ran a Kickstarter for the Oculus Rift and raised $2.4 million, 974 percent of target. Two years later, Facebook bought the company for about $2 billion. In 2017, he was out, in circumstances that produced a great deal of coverage and very little clarity. That same year, at twenty-four, he founded Anduril Industries.
The company makes surveillance towers, and more than 350 are operating on the southern border. They make small drones, including a quadcopter that can carry a warhead, counter-drone interceptors, one of which takes off vertically, goes supersonic, and lands itself again if it isn’t used, which is a genuinely new idea in the economics of air defense. They make a cruise missile family. Autonomous submarines, including the one Australia is buying. An autonomous fighter called Fury. And underneath all of it, a software layer called Lattice that lets one operator see what every sensor sees and direct what every shooter does.
Some of it has been shot at. Roughly a hundred loitering munitions went to Ukraine in 2023, about forty helicopters have been there since 2022, and Taiwan took 131 last year.
The thesis, in his own words on 60 Minutes last year:
“The idea behind Anduril was to build not a defense contractor, but a defense products company.”
That distinction is important. A defense contractor sells you a program: you pay for development, you carry the overruns, you own the schedule risk, and twelve years later you have a thing. A products company builds it on its own money and sells it to you finished, the way Chevrolet sold you a car. That difference has nothing to do with marketing posture. It comes down to whose capital is at risk, and therefore to who is in a hurry.
In January 2025, Anduril announced a factory. Arsenal-1: five hundred acres in Pickaway County, Ohio, five million square feet planned across seven buildings, four thousand jobs over a decade, about a billion dollars of the company’s own money. The governor’s office called it the largest single job creation and payroll project the state had ever announced.
The reaction from people who have watched defense announcements for thirty years was that this was a rendering with a press release attached. Venture-backed companies announce factories. They don’t build them.
Here is where it stands this month. The first main building is finished, about a million square feet. The exterior of the second is done. They said production would begin in July 2026; it began on 23 March, four months ahead of their own schedule. Headcount went from about fifty in January to more than three hundred now, against an internal target of 250 by year end.
They are running ahead of a timeline nobody imposed, in a building nobody ordered, because it’s their own money (and not yours) being spent.
A factory finishing early isn’t ordinarily news. In this industry it is, because the standard reason things are late is that nothing starts until the money arrives, the money doesn’t arrive until the requirement is final, and the requirement isn’t final until somebody has argued about it for four years. Arsenal-1 was built outside that sequence entirely. No requirement, no appropriation, no program office, no customer obligation. A company spending its own billion dollars on a bet about what the country would need.
Twenty-two workstations, deliberately balanced so each carries roughly equal work. About thirty people to launch it, trained in California and moved to Ohio. On one shift, an aircraft moves station to station every five days; at three shifts, about a day and six hours. Almost no automation. No large permanent tooling. Aluminum instead of titanium. Commercial parts wherever one will do. No castings and no forgings, because those mean tooling, and tooling means lead time.
Their head of production for autonomous aircraft, out loud:
“Our general posture is, question the value of automation, especially super early in the ramp.”
And:
“The best part is, no part. The best process is, no process.”
And the chief operating officer, on what they optimize for: they’re thinking about the cheapest and simplest possible way they can make this.
Go back to the man in the stamping mill in Buffalo, bought by Ford because Ford wanted the tooling expertise. Who spent his thirties building assembly plants in a dozen cities because he refused to accept that shipping finished automobiles across a continent was simply what it cost. Who took Chevrolet from 75,700 cars to a million in six years by asking every day what the cheapest possible way to make this was.
Twenty-two balanced workstations is Highland Park: cut the job into equal pieces and move the work past the men. Thirty people trained elsewhere and shipped in to start a line is the branch plant system, which rested on knowing the scarce thing was never the building but the people who could run it. And question the value of automation is the most Knudsen sentence anyone has said this century, because his reputation inside Ford and GM rested on accuracy and not on cleverness. What they said about him was that he never did anything ingenious he didn’t have to do.
Building ahead of the order
On 6 October the Navy and Anduril announced a second factory. Arsenal-2, and it has nothing to do with drones. Submarine components. Virginia class.
The Navy is putting up to $2.9 billion behind it. Anduril is putting in $3.7 billion of its own capital. Six and a half billion dollars in total, and the larger share is private money going into the submarine industrial base.
That arrangement is strange by the standards of the last eighty years. The deal in American defense since 1945 has been that the government funds development and the plant, often literally owning the building, and the contractor operates it for a fee. Cost plus. The contractor’s capital is never at risk, which is precisely why the contractor has never once been in a hurry. Here a private company is putting more money into the yard than the United States Navy is.
Luckey has been explicit about the reasoning. In March, he said he couldn’t sit and wait
“for the government to come, give me an order, give me money, then start building.”
And then:
“How stupid would I feel if I won a bunch of contracts to build the things that would deter China from invading Taiwan.”
The sentence finishes itself. How stupid would I feel if I won the contracts and then couldn’t build the things.
That is the Knudsen problem exactly. In 1940 the United States didn’t have a weapons shortage, it had a factory shortage. The designs were mostly known. What didn’t exist was a building with modern tools in it and trained men standing at them. And what Knudsen did about that, with no legal authority whatsoever, was spend $2 billion in 1940 and more than $4 billion in 1941 putting up plant, for weapons whose designs weren’t finalized, in a country that hadn’t entered the war.
He built ahead of the order. Everything anyone celebrates about American war production is downstream of concrete poured before anyone could prove it was needed.
Revenue in 2025 was about $2.2 billion, more than double the year before; projected revenue this year is around $4.3 billion. Across every contract it has ever held, the federal government has actually obligated $2.76 billion to the company. In June, the Air Force awarded production contracts for the collaborative combat aircraft program to General Atomics and Anduril, four months ahead of schedule, and says it wants more than 150 of those aircraft by the end of the decade. Australia signed a A$1.7 billion program for an autonomous submarine, with the first production vehicle delivered this month. And under a Pentagon framework for low-cost cruise missiles, Anduril’s minimum commitment is a thousand missiles a year beginning next year.
The bet underneath all of it is a claim about cost curves: that if you design a thing with almost no custom parts, no custom tooling, no castings and forgings, and build it in volume, cost collapses toward materials and labor hours. Do that to weapons instead of consumer goods and a country can suddenly afford a magazine deep enough to fight with.
Whether or not he’s right, that’s the same bet Knudsen made and the same bet Chevrolet was. The history of the American automobile is the history of driving a complicated machine down a cost curve until an ordinary person could own one, and everybody in 1908 knew that was impossible too.
Judged in October of 1940
Here are the objections.
Anduril is small. Total federal obligations across every contract it has ever held come to $2.76 billion, roughly four months of revenue for Lockheed Martin. The valuation is $61 billion and the company isn’t profitable, with a projected operating loss this year around $1.2 billion.
Most of it hasn’t been delivered. Arsenal-1 builds one product. Fury capacity is about fifty aircraft a year now, a theoretical 150 at three shifts, and they don’t have orders for 150. The cruise missile has a minimum of a thousand a year starting next year and zero delivered. In September their own VP for strategy said the company had gone as far as it could on its own money and couldn’t start full production without the next defense budget.
The third objection is the hardest. The Army mixed-reality headset program Anduril took over from Microsoft was cancelled in June. The government had spent about $1.8 billion and produced nearly ten thousand units that didn’t meet soldiers’ needs, and those went into storage. Anduril inherited that program instead of creating it, and its replacement is in prototype, but it’s on their record so I have to mention it (though I think they’ll correct this in due course).
And a fourth nobody in this industry likes saying. In 2024 Ukraine put roughly a million drones into the field and ninety-six percent were made in Ukraine. Anduril’s largest confirmed drone order to date is six hundred units.
Every one of those is true. They’re also the wrong test, and the reason is eighty-six years old.
Carry all four arguments back to Washington and apply them to Bill Knudsen in October of 1940.
He was small too, at 3,600 airplanes against a presidential target of 50,000 a year, about seven percent of the ask. Most of it hadn’t been delivered: he’d ordered ten thousand aircraft and mailed letters of intention for fifteen thousand more, which means the overwhelming majority of what he had to show for five months was paperwork and holes in the ground. And he was wrong about important things. In August 1940, the man running American war production told the press that automobile plants couldn’t be converted to build aircraft, that only the buildings would be any use. By 1943 Chrysler had converted 89 percent of its machine tools.
So if you’d stood in Washington in October 1940 with a notebook and judged him on units delivered, you’d have written that he was a well-meaning amateur, out of his depth, who had generated favorable press and a lot of poured concrete.
But the poured concrete was the point. The factory is the weapon, and everything else is downstream of whether the building exists with the tools in it and somebody trained standing at them.
The third objection against Knudsen is instructive as well. He was flatly, publicly wrong on the most important technical question in front of him, and none of it changed the outcome, because the trait that made him useful had nothing to do with being right. He kept building other plants, and when the auto plants did convert, he was happy to be wrong.
The man who refuses to accept impossible is rarely the man who turns out to be correct about everything. He’s frequently, embarrassingly wrong, because the only way to find the real limit is to walk into it at speed (think Elon watching his rockets explode). What he won’t do, what they can never do, is take somebody else’s estimate of that limit as a reason to stop walking.
There is a cheap version of this argument, which is that attitude beats arithmetic. It doesn’t. Knudsen didn’t will fifty thousand airplanes into existence, and the fifty thousand never happened in the form the President asked for. What happened is that a man with thirty years of production experience looked at an impossible number, worked out what was achievable, committed to it in public, and then spent six billion dollars of other people’s money building the plant required to get there.
Optimism has nothing to do with it. That’s a cold assessment of what a factory can do, performed by somebody who had built a lot of factories, and the only thing separating him from the pessimists was that his assessment ended in a number rather than in a shrug. The pessimists had the arithmetic right and the conclusion wrong, and the only way anybody found that out was that somebody went and did the work while they were still talking.
The Ukraine number works the same way. A million drones a year, ninety-six percent built inside the country doing the fighting, by small shops redesigning weekly. That’s the actual shape of what we’ll need, and no American institution is currently built to produce it.
And it’s an argument for a specific thing. Cheap, fast, made in many places, designed so no part is precious. Aluminum instead of titanium. The best part is no part. It’s an argument against the twelve-year program and for whoever is standing in a half-finished building in Ohio arguing with his own engineers about whether they really need that robot.
And Palmer Luckey, a home-schooled 34-year-old wearing cargo shorts and flip-flops, is leading the way. I love this country.
Knudsen was not promoted for success, whatever the rank suggests.
On 16 January 1942 Roosevelt created the War Production Board under Donald Nelson, with, in the official language, a single head with all power, to end what the Bureau of the Budget later called the slow processes of debate, consultation, board action, and the inconclusive influence of advisory interdepartmental agencies. That was his agency being described, in a government history, by people who had worked alongside him. It was abolished on the 24th. His commission is dated the 28th. He was moved sideways four days after his organization was dissolved out from under him.
He served until June 1945 and was awarded the Distinguished Service Medal twice. He went home to Detroit and died there in 1948, at sixty-nine. And the plants he put up in 1940 and 1941, with no authority, for weapons that hadn’t been designed, in a country that wasn’t at war, turned out $61.3 billion of munitions in 1944 alone.
I’m not going to tell you Palmer Luckey is exactly Bill Knudsen. Knudsen was sixty-one when Roosevelt called. He’d already built everything he was ever going to build, he took a dollar a year, he wore the uniform of the United States Army, and he had nothing left to gain. Luckey is thirty-four, owns a very large piece of what he’s building, is selling to the government instead of working for it, and if this comes off he’ll be among the wealthiest men in America. Those are real differences and they aren’t small.
But they share the one thing this country has always actually run on, and it has never been genius or money or even patriotism, though all three help and both men are/were American patriots. What it runs on is a flat refusal to accept somebody else’s assessment that a thing cannot be done.
Fifty thousand aircraft a year was an impossible figure. Everyone said so, in print, including the people whose job it was to know. A Danish bicycle mechanic who had never run anything larger than an automobile company looked at that number, understood it was wrong, found the number that was right, and hit it inside a thousand airplanes.
Fifty-one primes down to five. Two companies that build surface warships. Four one-hundredths of one percent of the world’s commercial shipbuilding. Half the submarines we promised. A third of the shells. A magazine that runs dry inside a week in every wargame anybody has run.
Everyone will tell you those are impossible figures too, and they’ll have reasons, and the reasons will be good ones.
A country doesn’t get out of a hole like that on policy. Policy helps, and we should have better policy. But in 1940 what actually moved was not a statute and not an appropriation, because the appropriations came late and the statutes came later. It was one man who took a job with no authority attached and started placing orders anyway, on the theory that somebody would eventually make it legal.
You can usually spot him, because he’s the one who has already started building before anyone told him he was allowed to.
Right now, a man pouring concrete in Pickaway County, Ohio, for weapons nobody has ordered yet, is revolutionizing our military procurement system.
That is exactly how America won her last war, and the only way it will win her next one.