---

The closing chapter of the book, reproduced in full — on how a field can hold the answer for a century and never think to ask the question.

Every reader of this work eventually arrives at the same question, and it is a fair one: why is this argument coming from a retired data engineer rather than from a constitutional scholar? The chapter below answers it, and not with biography. It sets out twenty cases — in law, medicine, geology, astronomy, linguistics, and financial regulation — where the evidence was published, available, and unread for decades, and it identifies who finally asked. The pattern is not that outsiders are usually right. It is that a question lying outside a field’s settled sense of what matters does not get refused. It never comes up.

The chapter also states plainly what this work owes to a set of tools, and shows what that dependence looks like in cases where it can be checked against something physical: a word read inside a scroll nobody had opened, ink found in scan data everyone had already examined, a planet located in an archive astronomers had combed for years. In each, a machine supplied reach and a person supplied judgment — and the field’s own experts confirmed the result before it counted for anything.

A field can fail at its own edge in two ways. Sometimes the field is working with what it has, and someone arrives with something better — a sharper instrument, a larger set of records, a method nobody had tried. Call that failure the better tool. Sometimes the field already holds everything the answer requires. The numbers are printed in its own annual report. What is missing is a question. Call that failure the unasked question.

The two are not equally serious. The better tool indicts nobody; a geologist in 1900 cannot be blamed for lacking satellite imagery. The unasked question means the field held the answer, looked at it, and did not see it — and the reason was never a shortage of evidence.

The argument in this book is an unasked-question case. The regulations are published in the Federal Register. Patent expiration dates are public. The window stickers are on the cars. It required somebody to ask what happens to a design when the patent runs out and the federal standard does not.

This chapter does three things. It shows how constitutional doctrine goes blind, and who has noticed. It shows why the person who notices almost never wins the case, which matters because the author of this book cannot bring one. Then it turns to fifteen years of results where machines did the searching. Nobody aimed them.

Two rules run through all of it. Every case gets labeled — better tool or unasked question — and where a case is some of each, which part is which. And where the popular version of a story is wrong, the correction is given instead of the half that helps. Rounding a case toward the answer you want is the exact error these cases charge other fields with.

Thomas Kuhn gave the problem its vocabulary in The Structure of Scientific Revolutions. A field operating normally does not test its own foundations. It works inside a settled framework that tells its members which questions count. That framework makes ordinary work efficient, and it makes certain facts invisible while they sit in plain view. A question outside the field’s settled sense of itself does not get refused. It never comes up.

Movement One: The Unasked Question

Start with Harry Markopolos, because this case has the same shape as the argument here. A financial analyst, he was asked to copy Bernard Madoff’s investment returns and could not, because they were not possible. He told the Securities and Exchange Commission in 2000, again in 2001, and in 2005 sent a document titled “The World’s Largest Hedge Fund is a Fraud.” The agency did not act. The scheme collapsed in December 2008 and took roughly sixty-five billion dollars with it. The Commission’s own Inspector General later found it had received credible complaints and failed to act. Markopolos did not find a document; he read numbers the agency already had, with no standing and no way to make anyone look. The unasked question, and note the kind: a regulator failing to see what sat in its own files, delivered by hand, three times.

Aiko Herzig-Yoshinaga had no law degree and no institutional standing; as a teenager she had been imprisoned at Manzanar. In the National Archives she found what forty years of litigation had not — the government’s original copy of General DeWitt’s Final Report, with passages a later version had removed. They showed the claim of military necessity made to the Supreme Court in 1944 had been made while contrary evidence sat in the file. Fred Korematsu’s conviction was vacated in 1983, and in Trump v. Hawaii (2018) the Court said Korematsu had been gravely wrong the day it was decided. The unasked question, in its most damning form. The evidence sat in the government’s own files for forty years.

Bowers v. Hardwick (1986) upheld a Georgia sodomy law and stated that laws against homosexual conduct had ancient roots. Seventeen years later a group of historians filed a brief in Lawrence v. Texas (2003) taking that statement apart from sources in print the whole time. The category of person Bowers described did not exist in the period Bowers pointed to. The Court overruled Bowers. The unasked question, and the cleanest one in law. A court asserted a fact, and for seventeen years nobody checked it against scholarship that already said otherwise.

Josephine Goldmark was research director of the National Consumers League and had no law degree. Of the hundred and thirteen pages filed for Oregon in Muller v. Oregon (1908) under Louis Brandeis’s name, she and her researchers supplied about ninety-eight — factory reports, medical testimony, foreign labor statutes. Roughly two pages were legal argument. The Court upheld the law and took formal notice of the material; the Brandeis brief is the ancestor of every filing that brings a court facts rather than precedent. Two corrections. Some of the medical evidence was later refuted. And Muller rested on women’s physical dependence — limits imposed on women and not on men — a ground contested ever since; this book takes no position on the dispute. Praise the method. Do not praise the holding. The unasked question. The reports were public; nobody had put them in front of a court.

Kenneth and Mamie Clark were psychologists, not lawyers. The doll studies did not decide Brown v. Board of Education (1954). They appear in footnote eleven, supporting a conclusion the Court had already reached on constitutional grounds. What the Clarks supplied was a finding the Court was willing to state — that segregation produced a sense of inferiority in children. The unasked question, with the causal claim kept small. They did not win Brown. They gave the Court a sentence it could write.

The pattern outside law is the same. The Vienna General Hospital ran two maternity clinics, one staffed by doctors, the other by midwives. From 1840 through 1846 the death rate from childbed fever in the doctors’ clinic was 98.4 per thousand births; in the midwives’ clinic it was 36.2. The figures were printed in the hospital’s own annual returns, and because women were admitted on alternate days with no medical selection, the obvious objection was not available. Ignaz Semmelweis saw the cause in 1847: the doctors came to deliveries from the dissecting room. He ordered chlorine handwashing, and deaths fell. One correction. Semmelweis was not an outsider. He was a credentialed physician in the very clinic whose figures he read. If a theory of blindness requires an outsider to break it, that theory deletes the best case medicine has. The unasked question.

Pellagra killed thousands of Americans a year in the early twentieth century, most of them poor and Southern, and the consensus was that it was infectious. Joseph Goldberger made an observation that required no instrument: in the asylums where inmates fell ill by the hundreds, the staff never did. An infection does not respect a payroll. So it was diet. The resistance is the useful part. Southern physicians, newspapers and officials rejected a finding that blamed regional poverty. The objection was not to the evidence but to what the answer implied about them. The unasked question.

In 1946 the radiologist John Caffey described a pattern in infants — multiple fractures healing at different times, with bleeding around the brain — that does not happen by accident. The films sat in hospital record rooms for sixteen years, until C. Henry Kempe named the battered-child syndrome in 1962. The first state reporting laws passed in 1963; by 1967 all fifty states had one. The unasked question, with a sixteen-year gap and a five-year legislative answer.

Yuri Knorozov is the purest case. A sixteenth-century Spanish friar, Diego de Landa, wrote down what he took to be the Maya alphabet. It and photographs of the surviving Maya books were in print by the end of the nineteenth century. Scholars used it and failed, because everyone read the list as an alphabet. In 1952 Knorozov, in Leningrad, read it as a syllabary: when de Landa asked for the letter B, his informant had given him the syllable be. Now the correction, and it makes the case stronger. It is not true that nobody asked. Many asked. What Knorozov brought was not evidence and not an instrument but a different idea about what kind of writing it was. That is the shape of the argument in this book. The Patent Clause has been read continuously since 1787. Reading a text wrongly is not the same as never reading it. That is why forgetting is the right word for what happened to the Clause’s second half. The unasked question, and the strongest example here.

Robin Warren and Barry Marshall must be labeled honestly, and the label is mixed. In 1938 Doenges examined 242 human stomachs after death, found spiral organisms in forty-three percent, and connected them to no disease. In 1954 Palmer studied tissue from a thousand people and concluded they were contamination from the mouth. Medicine already believed the stomach was sterile; Palmer’s study told the field what it already believed, on a sample too large to argue with, and the question closed for thirty years. Warren took the organisms seriously again in 1979; Marshall joined him in 1981. What broke the resistance was new capability: the bacterium was grown in 1982, and Marshall drank a dose in 1984 and made himself ill. So the slides were an unasked question for decades, and the answer came from the better tool. The sharper point: the organisms had been counted in nearly half a large sample — and then explained away.

Two more from the physical sciences. J Harlen Bretz read the scoured landscape of eastern Washington in 1923 as the work of a single catastrophic flood; geology treated catastrophe as disreputable, and in 1927 senior geologists, several of whom had never visited the ground, took his argument apart. Joseph Pardee supplied the missing water in 1942 — a glacial lake in Montana that had drained catastrophically. The unasked question, with a missing fact; roughly fifty years. Fritz Zwicky in 1933 measured how fast galaxies moved inside the Coma cluster against how much matter he could see, found the cluster should have flown apart, and named what held it together dark matter. The calculation used published measurements and standard physics. It was largely neglected until the 1970s. Mixed: the 1933 result was an unasked question, and the resolution took better tools.

For decades, parents were told to put babies to sleep on their fronts. The reasoning was never tested. A 2005 systematic review in the International Journal of Epidemiology pooled the studies as they appeared, year by year, and asked when the answer would have become visible. By 1970 the pooled odds ratio for front sleeping was 2.93 — nearly three times the risk. The advice continued in childcare books until 1988. The reviewers’ estimate: a review done in 1970 might have prevented more than ten thousand infant deaths in the United Kingdom alone. The unasked question, and possibly the cleanest one in this book. The studies were on library shelves, and the method that found the answer was to add up what was already there. What was missing was the decision to look.

One qualification. No book gathers these constitutional cases into a single account of how doctrine goes blind and who notices, though Sam Erman, Peter Irons, and Lani Guinier and Gerald Torres have each written nearby. No subscription law-review database was searched; the claim is that nothing surfaced, not that nothing exists.

Movement Two: Why the Person Who Notices Rarely Wins the Case

Every case above has the same structure. Somebody outside the field supplied the material. Somebody inside the field carried it to a decision. The two jobs never sat in the same person. Search for the pure form — an outsider whose own research produced a Supreme Court holding in a case that person brought — and you find nothing. What follows is an account of why.

Clarence Earl Gideon wrote his own petition from prison. He is the story everyone reaches for, and he does not fit. The Court appointed Abe Fortas, and the winning argument was Fortas’s; Gideon v. Wainwright (1963) supplied the case the Court wanted, not the theory. Susette Kelo, whose house was condemned for a development, discovered nothing and lost in Kelo v. City of New London (2005). What matters is what happened after: more than forty states tightened their eminent-domain laws, and the development was never built. A loss can produce a reform. Keep that in mind.

Brewster Kahle came closest. An engineer and archivist, not a lawyer, he made a structural observation of exactly the kind this book makes. When Congress removed the requirement that copyright owners renew their copyrights, it turned an opt-in system into an opt-out one. Works stopped entering the public domain by default. He lost. The Ninth Circuit, in Kahle v. Gonzales (2007), treated it as the argument the Supreme Court had rejected in Eldred v. Ashcroft (2003). The court admitted Eldred had left open whether “limited Times” puts a real ceiling on a term, then left that ceiling to Congress. And the Solicitor General then cited Kahle against the next challenger; his defeat became a tool in the government’s hand. The distinction this book rests on is that Kahle was about extending a term. The argument here alleges no extension of anything. Whether that distinction holds is for a court to say.

The reason the two jobs never sit in one person is arithmetic and procedure. In one recent set of 6,273 petitions to the Supreme Court, 6.3 percent were granted or sent back for reconsideration. Of the 3,785 filed by people too poor to pay the fees, 98.78 percent were denied, against 85.97 percent of the 2,488 paid petitions. A person representing himself may represent only himself. When the Court wants a case it appoints a lawyer, which separates the person who found the thing from the person who argues it, by design.

Pierre Azoulay and two colleagues measured the gate from the other side. They found 452 star scientists in the life sciences who died unexpectedly and tracked their fields afterward. Publication by scientists who had never worked with the deceased rose 8.6 percent. That is the mechanism behind every case above, measured. Not that outsiders are right. That gatekeeping is real, and that it relaxes when the gatekeeper dies.

The docket. No reported federal decision has reached the merits of a claim that federal regulation, without extending any patent term, preserves practical exclusivity in an invention whose patent has expired. The statutory neighbors — patent-term extension and adjustment, orphan-drug exclusivity, biologics — were checked one at a time; in none was the constitutional question reached or raised.

Why has nobody argued this? Generic drug makers fight regulatory exclusivity constantly. Three honest answers. Congress gave generic manufacturers a statutory path to market, and a company with a cheaper remedy does not reach for the Constitution. Durable goods have no such path — there is no abbreviated application for a patent-expired engine — so the blockage is total rather than temporary, which is exactly why it raises a question the drug cases never had to reach. And a would-be remanufacturer must establish that the injury is real before anything has been built. That is a reason the question has gone unasked. It is not a reason the answer is no.

This is why the book is written as a brief for a plaintiff rather than as a complaint. The author found the argument. The author cannot carry it. The honest response is to write for the person who can, not to pretend to be him. Adoption is the verification. That is mechanics, not modesty.

Movement Three: When the Machine Asked

The last group is different, because in it nobody aimed the machine. In AD 79 Vesuvius buried Herculaneum and with it a villa library of roughly 1,800 papyrus scrolls, cooked to charcoal. They are unreadable because the ink is carbon, the papyrus is carbon, and unrolling one destroys it. In 2019 Brent Seales put two of them through a particle accelerator and produced high-resolution X-ray scans. In 2023 the scans were released publicly as an open competition with a seven-hundred-thousand-dollar grand prize.

Two people acted on those scans, and the difference between them is the point. Casey Handmer is a physicist with no papyrology and no Greek. He built nothing. He looked at the scans for hours. He noticed a faint texture, like cracked mud, in shapes that looked like Greek letters, and published it on his blog in August 2023. The carbon ink is invisible against carbon papyrus; what Handmer saw was the disturbance the ink had left in the surface it dried on — the mark of a thing rather than the thing. It was the first substantial evidence of ink inside an unopened scroll. The unasked question. He looked at a picture everyone had, for longer than anyone else had, without knowing what he was supposed to see — which is precisely why he saw it.

Luke Farritor was twenty-one, a computer science undergraduate at Nebraska–Lincoln. He read Handmer’s post and did the thing Handmer could not: took the crackle as training data, built a model to find more of it, and iterated until the model returned a word: πορφύρας — purple. By December, Farritor, Youssef Nader and Julian Schilliger had recovered about two thousand characters and shared the grand prize. This one is the better tool, and the labels matter here because the sequence is the point. Seales built the instrument. Handmer, with no standing, noticed something in data everyone possessed. Farritor, with no standing either, built a machine that could find more of it. The field’s own experts said whether it was Greek. Four steps: an institution supplied the capability, an outsider the observation, a second outsider the machine, and insiders the judgment about whether any of it was true.

That is the arrangement behind this book, and it should be said plainly rather than left for a reader to infer. The author used machine assistance to assemble the argument. He says so here, because the alternative is letting a reader find it out. What the Herculaneum sequence shows is that the assistance was one step of four. Somebody has to notice. Somebody else has to check. The middle step is the only one a machine performs.

Christopher Shallue was a software engineer at Google, not an astronomer. NASA’s Kepler archive was public and had been worked over for years by professionals who had extracted thousands of planets from it. They had concentrated on strong signals, because a weak dip in brightness is usually noise. Shallue’s idea was that a neural network might sort the weak ones — that the constraint had been human attention rather than the data. He trained a network on signals scientists had already labelled and, with Andrew Vanderburg, the astronomer who supplied the domain judgment, pointed it at 670 systems known to hold several planets. It found Kepler-90i, an eighth planet the earlier methods had missed. A hybrid, and the halves are worth separating. The network was a new tool. But the data was public and had been examined. What had not happened was anyone asking whether the signals the professionals had reasonably skipped were worth a second look. The question was the contribution. The network was how he could afford to ask it.

In the results below, the person holding the instrument was an insider, working on data his own field had been generating for decades. That takes the operator’s credentials out of the argument entirely. These are demonstrations that a field can stare at its own evidence for a century and not see what is in it.

Doctors have photographed the back of the eye for about a hundred years. Ryan Poplin and his colleagues trained a system on retinal photographs from 284,335 patients. From the photograph alone it estimated age, predicted whether the patient smoked and whether the patient would have a major cardiac event — and identified the patient’s sex almost perfectly. No ophthalmologist can do that. The unasked question, without qualification. The instrument was new. The information was not.

Andrew Beck’s team at Stanford turned a system called C-Path on breast-cancer slides. Breast cancer had been graded for decades by reading the cancer cells themselves, under a scheme largely unchanged since the 1920s. C-Path found that features of the stroma — the ordinary connective tissue around the tumour — predicted survival more strongly than the tumour did. Two corrections. It is partly a better-tool case. And the ninety years sometimes attached to the story belongs to the grading scheme, not to how long the stroma went unread. Nothing hid the stroma. Specialists walked past it for the better part of a century — not because they could not see it, but because they were confident they already knew which part of the slide mattered. The unasked question. That sentence is the thesis of this book, and a pathology slide is where it can be proved.

Two more, briefly. Mayo Clinic researchers trained a system to detect a weak-pumping heart from an ordinary electrocardiogram, a tracing recorded since the early 1900s, then tested it in a randomized trial across 120 primary care teams covering 22,641 adults; new diagnoses within ninety days rose from 1.6 percent to 2.1 percent — a modest effect, stated as a modest effect, and a randomized trial, which almost nothing else in this movement has. Nicolas Coudray’s team trained a system on lung-cancer slides stained with the dyes pathologists have used since the nineteenth century, and from the stained image alone it predicted, moderately well, which genes were mutated. A tumor’s genetics leave a visible mark in the tissue, and a hundred and fifty years of pathologists did not know it. Both the unasked question.

AlphaDev and AlphaTensor are a different trick, often confused with this one. AlphaDev searched for faster ways to sort short lists of numbers. It also claimed that for the three-element case no shorter program existed. Cassio Neri showed otherwise, with counterexamples and a faster plain implementation. AlphaTensor found a way to multiply certain small tables of numbers with fewer multiplications than the best method known since 1969 — in a restricted arithmetic, not ordinary arithmetic. Manuel Kauers and Jakob Moosbauer answered within about a week and went past the machine, cutting the five-by-five case from ninety-six multiplications to ninety-five, on a small computer, starting from the machine’s output.

The machine results are the better tool. But look at what these two cases produced. AlphaDev asserted an impossibility and was wrong, and being wrong in public provoked somebody to check. AlphaTensor showed that ground the discipline had stopped digging in was not exhausted, and within days mathematicians who had held the capability the whole time went further than the machine. So the machine results are the better tool, and the human responses to them are the unasked question. What the machines dislodged was not a technical barrier. In one case it was a false claim. In the other it was a settled belief that a question was closed.

The Turn

In every case in this movement but the last two, the machine was not told where to look. And in every case, a human being decided what counted as proved. Mayo ran a randomized trial across forty-five sites. The pure-mathematics result the preface describes arrived with a companion paper from nine mathematicians who had checked the proof by hand.

That structure is the structure of this book. Tools helped assemble the argument. Every case in it was read. Every figure was checked against a primary source or marked as unchecked. And a court, not a machine and not the author, will decide whether the argument stands.

Close

Fields build up blind spots at their own edges. People whose training points at those edges are more likely than average to see what is sitting there. That is the whole claim. It is narrower than the story usually told about outsiders, and it is the only version the evidence supports.

Constitutional law is a field with that kind of edge. It has the documents. It has had them the whole time. What it has not had, on this question, is the question. This chapter has explained why the author is not among those with standing to file it, and why that is the ordinary condition rather than a defect in the argument.

Goldmark did not argue Muller. The Clarks did not argue Brown. Herzig-Yoshinaga did not argue Korematsu. Markopolos did not prosecute Madoff. Each of them found something. Somebody else carried it. The finding is here. The blindfold is described. Somebody with standing has to take it off.