Welcome to issue #13 of Obviously Stupid Questions.

By late August 1854, cholera had killed over 600 people in London’s Soho neighborhood in under two weeks. Snow walked the streets door to door, marking each death on a map. The cases clustered tightly around a single water pump on Broad Street. One woman, who died miles away, was thought to be an outlier, but interviews revealed she had sent a servant to fetch water from that pump because she liked its taste. Workers at a nearby brewery, who drank beer instead of water, barely got sick at all. ^[1]

Snow’s proof didn’t rest on the map alone. He’d already mined civil registration records that William Farr’s General Register Office kept for administrative bookkeeping, not epidemiology. He traced thousands of cholera deaths back to which water company supplied each house. Households served by Southwark and Vauxhall, which drew from a sewage-fouled stretch of the Thames, died at eight or nine times the rate of households served by Lambeth, which had moved its intake upstream in 1852.^[2]

Snow took his map to the local council and asked them to remove the pump handle. They did, mostly to humor him, and new cases dropped off within days.

The vindication took longer than the pump handle did. London’s General Board of Health investigated the outbreak and concluded, in its official report, that Snow’s water theory did not explain the facts. Miasma stayed the accepted account for another decade, until germ theory caught up to what Snow had already mapped.

Concerning breakthrough questions, Snow demonstrates a principle seen in the manuscript. The data allowed him to convert a decade-old suspicion into an undeniable pattern that led to a dramatic insight that no case-by-case analysis could have revealed. Additionally, repurposing the records from an administrative record system built for one purpose become a research instrument for another. Even today we repurpose data. E-Z Pass is a good example. Initially developed for cashless toll collecting, their transponders are now used to study traffic speeds, congestion, and traffic flows.

You probably noticed that Snow’s careful data collection resembles epidemiology, and you would be right. Snow is regarded as the father of modern epidemiology now. In 1854, he was the doctor with the odd theory about a pump.

What question are we laughing at today?

1. John Snow, On the Mode of Communication of Cholera, 2nd ed. (London: John Churchill, 1855). For the Broad Street investigation, the map, and the General Board of Health’s rejection of Snow’s findings, see Steven Johnson, The Ghost Map: The Story of London’s Most Terrifying Epidemic (New York: Riverhead Books, 2006).↩︎

2. On William Farr’s General Register Office data and Snow’s use of it to compare the Southwark and Vauxhall and Lambeth water companies, see John M. Eyler, Victorian Social Medicine: The Ideas and Methods of William Farr (Baltimore: Johns Hopkins University Press, 1979); and David A. Freedman, “Statistical Models and Shoe Leather,” Sociological Methodology 21 (1991): 291–313, which analyzes Snow’s 1853–54 water company comparison as a natural experiment.↩︎

Tonegawa died on July 11 at 86, the first Japanese scientist to win the Nobel Prize in medicine, in 1987. He’d solved a hundred-year-old puzzle in immunology: how a body with roughly 20,000 genes builds hundreds of millions of distinct antibodies to fight infections it has never encountered. At the height of that career, he switched fields, to devote the next three decades to an entirely different question: how does the brain physically store a memory?

Most people who produce a breakthrough question spend a career answering it. Tonegawa answered two, in two disparate fields. He had a theory about why.

In 2022, asked why he’d leave immunology right after the Nobel, he didn’t describe a strategy. “I just followed my curiosity and instinct,” he told MIT’s Picower Institute. “I didn’t think about other things like: could it be too risky? Can I really develop my career by venturing into a field I am not familiar with? That never occurred to me.”

That might be an odd or unexpected admission. It’s also the clearest statement I’ve found among the hundreds of breakthrough thinkers studied for my book. He described a pull he couldn’t ignore.

For decades, biologists were stuck on how antibodies, shaped like a Y with a fixed base and a variable tip, could generate hundreds of millions of unique tips from a genome too small to hold a separate gene for each one. In 1974, at the Basel Institute for Immunology, Tonegawa spent months building molecular probes to compare the DNA of mouse cancer cells with mouse embryo cells. The patterns didn’t match. Antibody genes, he found, aren’t fixed at birth. The body cuts and recombines DNA segments to build new genes on the fly, which is how a few hundred segments become hundreds of millions of working antibodies. The scientists on his side of the old debate, he liked to call the somaticeans. They’d won. He published the finding in 1976 and won the Nobel eleven years later.

He started over anyway. In 1994 he founded MIT’s Center for Learning and Memory and took up a question that shared nothing with the first one: where in the brain does a memory actually live? His lab spent years locating engrams, the specific clusters of hippocampal cells that hold a single memory. In 2012, working with postdoctoral researcher Xu Liu and graduate student Steve Ramirez, he showed a fear memory could be switched on with a beam of light. A year later the same lab implanted a false memory in a mouse, tricking it into associating a shock with the wrong room.

At a Nobel laureates’ conference in Tokyo in 2017, someone asked him to define creativity. He said he didn’t feel he was very creative. Pressed for an answer, he named three things: “you have to be very curious about something,” “you have to have an unfailing urge to address the question you have,” and it helps “to try to combine knowledge in at least two different fields where people don’t necessarily interact.”

That third one is the part I’m most interested in: most people who end up working in hybrid fields didn’t do the hybridizing themselves. They inherited a combination someone else had already put together. Tonegawa built the combination himself, twice, in fields with no shared training path between them.

I relate to that more than I expected to. Building Breakthrough Questions meant learning things nobody trained me for: enough database architecture to run a research system, enough Python to keep it working, enough about directing an AI research process to get real analysis out of it instead of a summary. I had to cross into new territory to build my own combination. The crossing was personally challenging, but took me to a place I wanted to be.

References

Liu, Xu, Steve Ramirez, Petti T. Pang, Corey B. Puryear, Arvind Govindarajan, Karl Deisseroth, and Susumu Tonegawa. “Optogenetic Stimulation of a Hippocampal Engram Activates Fear Memory Recall.” Nature 484 (2012): 381–385.

McClain, Dylan Loeb. “Susumu Tonegawa Dies at 86; Nobel Laureate Showed How Antibodies Fight Infection.” New York Times, July 21, 2026.

Picower Institute for Learning and Memory. “Interview with Susumu Tonegawa.” Picower Institute Newsletter, Summer 2022.

Ramirez, Steve, Xu Liu, Pei-Ann Lin, Junghyup Suh, Michele Pignatelli, Roger L. Redondo, Tomás J. Ryan, and Susumu Tonegawa. “Creating a False Memory in the Hippocampus.” Science 341 (2013): 387–391.

Tonegawa, Susumu. “Somatic Generation of Immune Diversity.” Nobel Lecture, Karolinska Institutet, Stockholm, December 8, 1987.

Trafton, Anne. “MIT Professor Susumu Tonegawa, Renowned Molecular Biologist and Nobel Laureate, Dies at 86.” MIT News, July 15, 2026.

The Nobelist Rudolph Marcus, who died earlier this month at 102, felt the same way about chemistry, and he said so:

“It all comes back to the puzzles, I think, and the enjoyment of puzzles as a child,” he told The Electrochemical Society in a 2016 interview. “Only now the puzzles are scientific puzzles.”1

In some chemical reactions, electron transfer occur. That’s when one atom hands an electron over to another atom. Nothing else happens. No pieces break off. No new pieces get added. Just one electron, moving from one atom to the next.

Electron transfers sound abstract, but their uses underly many of life’s major processes: converting sunlight into food through photosynthesis, turning food into energy, or how fireflies glow in the dark, just to name a few. 1

In 1955, Marcus was a post-doc at a college in Brooklyn. One day he was flipping through science journals in the library, not looking for anything in particular. Writing that sentence brings back memories of my student days doing the same thing and perusing the stacks and sometimes finding some really cool book near what I was looking for.

Marcu found a paper by another scientist, Willard Libby, about electron transfer. Libby had a good idea, but it was missing something.1

Here’s what Libby knew. Back in the 1920s, scientists had shown that shining light on an atom could knock an electron loose, like a bowling ball that manages to get only one pin down. The light supplies the energy needed to make that happen. Libby thought something similar was going on during electron transfer.

But here’s the conundrum. Light supplied the energy in that 1920s experiment. But plenty of electron transfers happen in the dark, with no light around at all. So where does the energy come from?

That’s the blank space in the puzzle which Marcus filled in.

He realized that right after an electron jumps to a new atom, everything around that atom is suddenly a poor fit. The atom’s charge has changed, but its neighbors haven’t rearranged yet to match. Like adding an extra passenger to a car’s back seat, they jostle and shift into a new position, and that shifting itself stores energy, like a stretched rubber band.

Marcus’ math showed that this stored energy was exactly enough to power the transfer, even with no light involved. Energy in, energy out. Nothing broke any rule of physics.1

He solved the puzzle, but discovered it had more explantory power.

Marcus’s equations predicted that up to a certain energy level, transfers accelerated, but past it, “surplus” energy made the transfer slower.

We can visualize this by imagining a graph of speed against energy, the line goes up, reaches a peak, then comes back down. Marcus called that back half of the curve the inverted region.1

He described it later as “unexpected, like falling down a cliff and you make it too steep and it’s harder to fall.”1

Marcus published this prediction in the late 1950s. And then he had to wait. Nobody had equipment precise enough to confirm his equation’s results.

For almost 30 years, his puzzle was solved on paper and nowhere else. Finally, In 1984, three other scientists finally built molecules exact enough to test it. They ran the experiment. The inverted region was really there.1

Marcus won the Nobel Prize in Chemistry in 1992: 8 years after that experiment finally proved him right, and 37 years after that afternoon in the library.

I think about that every time I finish a hard crossword and check my solution. That only takes a few minutes. His took decades, and he was over 60 before he found out he’d gotten it right.

He solved another puzzle the same way, years earlier. As a young researcher in North Carolina in the early 1950s, he took an older theory about how molecules break apart and made it more accurate. Chemists still use his version today, in everything from drug testing to engine design.1

I’ve often thought that solving crossword puzzles can be a generalizable skill. Marcus showed that it is, the way that puzzlers work: notice the piece that’s missing and solve for it, even before you can prove you’re right.

Footnotes

1. Dylan Loeb McClain, “Rudolph Marcus, Who Illuminated Electrochemical Processes, Dies at 102,” The New York Times, July 17, 2026. ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8 ↩9

— Josh Kerr, who just set the new world record time for running the mile, smashing the prior one by nearly 3 seconds. (NY Times, “Fastest Miler’s Formula; Bespoke Spkes, A Team, Atltitude and Attitude, pB10, July 21, 2026.

The quote echoes a key finding from Breakthrough Questions. While rare, they are most likely to occur when the conditions are right: when a person has access to .the optimal combination of theory, colleagues, investigable questions, tools, methods, and a way to know the answer is correct or not. Breakthrough Questions are far more likely when the context supports the questioner holistically.

The story reads as one about talent: fewer scientists, a thinner bench, ground ceded to China and Europe. But the research done for the Breakthrough Questions manuscript points somewhere more specific. Research curiosity isn’t what’s being throttled back. It’s the infrastructure that turns curiosity into validated research.

Across the 328 contributors spanning five thousand years I studied, the capacity to ask a good question turns out to be common. What’s rare is access: to a theoretical framework that makes the question askable, to tools capable of investigating it, and to a validation mechanism that can judge whether the answer upholds the theory. Ph.D. programs are where young scientists get all three at once, under supervision, over years. They are incubators for potentially breakthrough questions, the type that change how we understand the world and shift the direction of knowledge. Cutting the program doesn’t make anyone less curious. The conditions under which curiosity becomes trained judgment become rarer and less accessible.

Richard Feynman got those conditions at twenty-four, apprenticed to Hans Bethe and Niels Bohr at Los Alamos. Bohr asked for “the young one” specifically because Feynman would tell him when his reasoning was wrong — a habit Feynman built by being corrected, in real time, by people whose judgment had already been tested. He didn’t show up at Los Alamos knowing how to separate a good physical intuition from a bad one. He learned it there, over years, from people who argued with him.^[2]

A retired microbiology professor commenting on the Times article put it more plainly than the article itself: a Ph.D. is an apprenticeship. You learn to ask and answer questions, design experiments, read data honestly, from faculty and from the other students and postdocs working next to you. Losing one year’s cohort isn’t the end of a department. Losing several years in a row means the people further along — the ones who’d normally be mentoring the new arrivals — are fewer in number, which weakens the training of whoever comes after them. The effect compounds forward, past the point where any single year’s funding gets restored.

That’s the part a headline about admissions numbers can’t easily show. A 15 percent drop in this year’s cohort doesn’t produce 15 percent less science this year. The training that a five-year apprenticeship builds doesn’t show up as a shortfall until the students who weren’t admitted would have been finishing their own independent work, a decade from now. By then the funding fight that caused it will be old news but probably bitterly remembered.

Breakthrough Questions argues that context, not raw cognitive ability, explains most of who gets to produce a breakthrough. This is that argument playing out in real time, given backbone by the Times and other stories certain to follow.

What happens to a field’s judgment when fewer in a generation gets trained than the one before it?

This piece is part of the Breakthrough Questions series, drawn from a database of 328 contributors to human knowledge spanning 5,200 years.

1. Vimal Patel, “Decline of Ph.D. Admissions Could Imperil a ‘Generation of New Talent,’” The New York Times, July 6, 2026. Reports the AAU Data Exchange figures cited above — the 15 percent admissions drop across 55 AAU member universities, the prior year’s 11 percent decline, and the specific cuts at M.I.T. and Caltech — along with the university and association officials’ account of unstable federal funding as the primary driver.↩︎

2. James Gleick, Genius: The Life and Science of Richard Feynman (New York: Pantheon, 1992). Documents the Los Alamos period in detail, including Bohr’s practice of calling for “the young one” and the apprenticeship under Bethe that shaped Feynman’s early scientific judgment.↩︎

Alfred Wegener’s continents drifted, but resistance to the idea lasted until the mechanism moving them turned up, thirty years after his death. Barbara McClintock’s transposable elements sat unconfirmed for three decades until molecular tools caught up with what she had already seen. In both cases, although they initially failed, a decision existed in principle — an experiment or series of them — that when repeated and reviewed, the entire field eventually accepted the results. Nobody in geology or genetics or most sciences continues disagreeing once the evidence is in: it becomes settled.

John Cage’s 4’33” fails the test a second way. Seven decades after its premiere, the argument over whether silence can be music is still open, because no decision exists that could close it. Critics and audiences form opinions, there’s no body that rules on the question, and no ruling would bind the next critic who disagrees. The argument fails to end because there’s no mechanism, even in principle, that produces a final verdict.

Today’s case, Bitcoin, fails the test a third way, and it’s worth seeing how it differs from Cage’s.

Satoshi Nakamoto published the Bitcoin white paper on October 31, 2008, proposing a way to transfer digital value without a central authority.^[1] The problem Nakamoto solved — how to stop someone from spending the same digital token twice — drew on four fields at once: cryptographic hash functions for verification, distributed systems theory for reaching consensus without a coordinator, game theory for making honest participation more profitable than fraud, and monetary economics for the policy built into the protocol itself.

As we know, the technical implementation worked. The Bitcoin network launched on January 3, 2009 and has been active ever since. Early adopters started using it within weeks. Nothing about the system has been shown to fail at doing what Nakamoto specified it would do.

Institutional adoption, though, is a different story. Some countries have built legal frameworks around cryptocurrency; others have banned or restricted it. Financial institutions vary widely in how they treat it. Decisions do exist here — a central bank can rule, a country can legislate, a regulator can issue guidance — which already sets this case apart from Cage’s. But each of those decisions binds only the body that made it. El Salvador’s ruling doesn’t bind Japan. The SEC’s position doesn’t bind a bank in Singapore. A decade and a half after publication, the question Nakamoto’s paper actually raised — how should digital value transfer work, and who gets to authorize it — remains unsettled not for lack of decisions, but because none of the decisions reaches beyond its own jurisdiction.

Adam Smith’s critics in the 1770s didn’t dispute that markets could allocate resources efficiently — they argued over what markets should be allowed to allocate, and who absorbed the cost when they got it wrong. Keynes’s opponents didn’t dispute his arithmetic — they debated how much authority a government should have over an economy. Both disputes ran on the same structure as Nakamoto’s: a treasury or a central bank could decide, and the decision would bind that country’s policy, but no decision bound the next country, or the next government that came to power in the same one.

Let’s look at what is not being argued. Nakamoto’s critics don’t claim the system fails to do what it was built to do. Smith’s and Keynes’s critics didn’t dispute that the mechanics worked as described. In every one of these cases, the execution isn’t in question. What’s at issue is whether an institution built around the old answer is willing to grant the new one standing. That’s a decision institutions can keep re-making, differently, indefinitely, without ever converging.

Breakthrough Questions is a reader-supported series drawn from research for the book in progress, tracing 5,200 years of discovery through the people who asked the questions that changed what we know. New posts every Tuesday.

1. Satoshi Nakamoto, “Bitcoin: A Peer-to-Peer Electronic Cash System,” white paper published October 31, 2008. Bitcoin network launched January 3, 2009. Note: Satoshi Nakamoto is likely a pseudonym. Two major efforts have worked to identify the true identity behind it. A lengthy piece by John Carreyrou with Dylan Freedman, “My Quest to Solve Bitcoin’s Great Mystery,” New York Times, April 8, 2026, identifies a candidate: a programmer named Adam Back. The documentary is sourced to Matthew Miele and Tucker Tooley, dirs. Finding Satoshi. Tucker Tooley Entertainment, 2026. The documentary, released April 22, 2026, follows a four-year investigation by journalist William D. Cohan and private investigator Tyler Maroney into the identity of Bitcoin creator Satoshi Nakamoto. Both the article and documentary identified programmer Adam Back as the leading candidate. Back, however, denies the identification in both cases.

In the summer of 1951, Barbara McClintock presented her findings on genetic transposition to a conference at Cold Spring Harbor Laboratory, a biomedical research institute that focues on genetics and other areas. She spent more than a decade documenting how certain genetic elements moved within chromosomes in maize, controlling the expression of other genes. McClintock’s methods and evidence were meticulous. The audience, all leading geneticists, sat mostly in silence. A few asked polite questions, but not one accepted her findings. James Watson, who listened to the talk, later wrote that most people thought she had “gone off the deep end.”^[1]

She was later proven right and then awarded the The Nobel Prize in 1983, thirty-two years later.

The history of knowledge is full of stories like this. The standard explanation is that fields are conservative, that established scientists protect their turf, that revolutionary ideas always face resistance. All of that is true. But it doesn’t explain why the resistance lasts as long as it does in some cases and not others — why some right answers wait thirty years and some wait three, why some debates close and others remain open.

This is one of the questions I pursued in my research for Breakthrough Questions. I discovered that the answer lies in the validation mechanism: the process a field uses to determine whether an answer is correct.

Every domain has one, though they differ. In physics and biology, nature validates. Run the experiment, measure the result, and the answer either holds or it doesn’t. Disagreement can persist for a while, but the mechanism exists to resolve it — the data, theory and methods eventually speak loudly enough that the field has to listen. In music and other aesthetic domains, validation comes from the community. An audience hears a piece and responds to it, critics interpret it, other composers engage with it or ignore it, and over decades a consensus may or may not form. No experiment settles the question. The community’s judgment is the only judgment available.

This distinction, which I’ve been building toward across this series, turns out to predict something specific: not just how knowledge advances, but how long a field can hold out against a right answer, and whether the resistance can ever fully end.

Let’s look at three cases that make the point.

Alfred Wegener proposed continental drift in 1912, backed by evidence from multiple directions: the geometric fit of African and South American coastlines, matching fossil species found on continents separated by thousands of miles of ocean, identical rock formations appearing on opposite sides of the Atlantic, glacial deposits in regions now tropical. The evidence was real, physical and told a compelling story. The geological community rejected his theory anyway. The rejection was scientifically defensible by the standards of the time.

Wegener, though, couldn’t explain what moved the continents. He proposed centrifugal force and tidal friction as mechanisms; both were quickly shown to be much too weak. Geologists weren’t simply being conservative — they were applying a legitimate standard. A claim about how the earth’s solid crust behaves over hundreds of millions of years requires an account of what drives the behavior. Wegener had the pattern but not the mechanism. His theory was incomplete.

Wegener died in Greenland in 1930, his fourth polar expedition, still arguing for a theory the geological establishment had largely dismissed. Thirty years later, oceanographers mapping the seafloor discovered the mid-ocean ridge system which confirmed that new oceanic crust was spreading outward from the ridges — the movement mechanism Wegener couldn’t supply. Once the mechanism existed, the theory became plate tectonics, and the resistance collapsed within a decade.

As we can see, the validation mechanism was always there. Nature would eventually settle it. The resistance lasted as long as the missing piece was missing, and ended when the piece was found.

McClintock’s situation was different. She had both the claim and the evidence. Her maize breeding records documented transposable elements across multiple generations; her cytogenetic analysis showed chromosomal rearrangements with a precision that her contemporaries acknowledged was exceptional. Her problem wasn’t missing evidence.

Resistance to her results had a different complexion. Genetics in the 1950s was moving fast and away from maize. The field’s attention had shifted to bacteriophages and the fruit fly,Drosophila, organisms that reproduced quickly and yielded molecular data. McClintock’s organism was unfashionable. Her methods — patient, morphological, requiring close observation of individual plants across growing seasons — sat at odds with the field’s new direction. Additionally, and significantly, her claim was conceptually demanding: it required giving up the idea that chromosomes were stable, that genes sat in fixed positions, that the genome was essentially static. This was a major adjustment to thinking: what was generally understood as true might well be false.

None of this changed the validity of her data. Biology is nature-validated. The mechanism for resolving the dispute existed, but slow. Molecular biology eventually developed the tools to see what she had seen by different means — restriction enzymes, gel electrophoresis, DNA sequencing — and when those tools confirmed transposable elements in organisms across the biological spectrum, the resistance ended. Her Nobel lecture in 1983 described the vindication without particular bitterness: she had known what she had found, and she had waited.

The field fought back for thirty years. But the battle lasted only as long as the tools to confirm her findings were unavailable. Once they arrived, nature validated, and the argument was over.

Composer John Cage asked a different kind of question.

In 1952, Cage composed 4’33” — four minutes and thirty-three seconds during which the performer makes no intentional sound. The piece consists entirely of whatever sounds occur in the environment during its three movements. Cage had come to the question through Zen Buddhism and the writings of Ananda Coomaraswamy, and through his study of silence itself: time spent in an anechoic chamber at Harvard had convinced him that true silence didn’t exist, that what we call silence is simply the sounds we haven’t yet learned to hear.^[2] His question was whether those sounds — unintended, environmental, indeterminate — could be musical material.

The audience reaction at the premier in Woodstock, New York, was hostile. People walked out of his performances. Critics dismissed him. Even Arnold Schoenberg, his former teacher who had spotted his talent early, said Cage was “not a composer, but an inventor — of genius.”^[3] Such backhanded compliments kill with kindness.

Seven decades later, the argument is still open. 4’33” is performed regularly at major concert halls and dismissed regularly by serious musicians. The resistance hasn’t collapsed the way it did for Wegener or McClintock, and it won’t, because no mechanism exists to close it. No experiment can determine whether silence is music. The question of what music should be — which is what Cage was really asking — is beyond what aesthetic validation can settle. The community can form opinions, argue, influence each other, shift over generations. It cannot deliver a verdict.

This is the asymmetry the validation mechanism produces. In nature-validated fields, resistance has a time limit. The tools will arrive, or the mechanism will be found, or the evidence will accumulate past the point where rejection is possible. In aesthetically validated fields, resistance can persist indefinitely — not because the field is less sophisticated, but because the validation process has no mechanism to force a conclusion.

Across 328 contributors to human knowledge spanning roughly 5,200 years I studied, the validation mechanism turns out to be the strongest single predictor of how knowledge in a field evolves — whether new approaches replace old ones or accumulate alongside them, whether frameworks shift or layer, whether the people who ask the breakthrough questions get recognized in their lifetimes or after. The pattern is evident across domains as different as physics, economics, and music.

McClintock and Wegener were right, and their fields eventually agreed. The mechanism that made eventual vindication possible — nature as arbiter — is the same mechanism that makes those fields cumulative, that makes each generation’s findings available to the next as confirmed ground rather than contested claim.

Cage asked a question that his field can’t answer. That’s not a flaw in the question. Some of the most generative questions in the history of knowledge are questions the available validation mechanisms can’t settle. They stay open, keep producing argument and response and new work, branch outward rather than forward. 4’33” is still generating both.

The question is whether you understand, going in, which kind of validation you’re seeking, how it works, and what is required to pass.

Breakthrough Questions is a reader-supported series drawn from The Question AI Can’t Ask, a book in progress tracing 5,200 years of discovery through the people who asked the questions that changed what we know. New posts every Tuesday.

1. Watson, J.D. The Double Helix. Atheneum, 1968. Watson’s account of the Cold Spring Harbor meeting and the reception of McClintock’s findings is corroborated in Evelyn Fox Keller, A Feeling for the Organism: The Life and Work of Barbara McClintock (W.H. Freeman, 1983), 150–175.↩︎

2. Cage, J. Silence: Lectures and Writings. Wesleyan University Press, 1961. The anechoic chamber experience is described in “Experimental Music,” pp. 7–8.↩︎

3. Revill, D. The Roaring Silence: John Cage, A Life. Arcade Publishing, 1992, p. 165. Schoenberg’s comment is also cited in Kay Larson, Where the Heart Beats: John Cage, Zen Buddhism, and the Inner Life of Artists(Penguin Press, 2012), 94.↩︎

Nothing in that trajectory guarantees she will ask a breakthrough question, but her story has recognizable elements that were uncovered through my 5200 year history of breakthrough questions and the people who ask them.

In 1812, Michael Faraday, a twenty-year-old bookbinder’s apprentice in London, attended four lectures by Humphry Davy at the Royal Institution. He had no formal scientific education. However, he had access to knowledge through his employer’s bookshop. He had read scientific texts including Jane Marcet’s Conversations on Chemistry, which he \annotated carefully. Davy’s lectures confirmed what his reading had taught him: that the questions at the frontier of chemistry and electrical science were real, unresolved, and worth pursuing. To do that he would need institutional access. To gain that, Faraday wrote up his lecture notes, bound them carefully, and sent them to Davy with a request for work. Davy hired him as a laboratory assistant.

Over the next two decades Maxwell pursued an investigation into the relationship between electricity and magnetism. When Hans Christian Oersted demonstrated in 1820 that an electric current deflects a compass needle, Faraday’s response was characteristic: if electricity affects magnetism, shouldn’t magnetism affect electricity? That inversion — the anomaly turned into a question — drove the work that produced electromagnetic induction in 1831, a discovery underlying every electrical generator built.

Although he had no university degree, he had done foundational reading, was exposed to a working scientist’s method, and had a mind trained to notice when something didn’t fit.

Across the 328 contributors to the history of human knowledge I analyzed, roughly a third show early foundational exposure as a documented origin of the question that eventually mattered. Not formal education specifically. Contact with the real material of a field, early enough before the conventions of that field became invisible. Faraday read Marcet before he understood what chemistry was. Darwin collected beetles as a teenager before he understood what species were. Karikó was working with RNA in Hungary in the 1970s before anyone believed it could be therapeutic. And this is not limited to science, but applies to all fields studied including music.

The pattern that emerges from the biographical record is not genius arriving fully formed. It is preparation meeting anomaly. The foundational exposure builds familiarity with how a field thinks. The anomaly — the thing that doesn’t resolve cleanly within that framework — becomes visible only to someone familiar enough with the framework to notice the gap. And noticing the gap is the beginning of the question.

Faraday noticed that electricity and magnetism were treated as separate phenomena when the evidence suggested they weren’t. Karikó noticed that cells use RNA constantly without immune response while synthetic RNA triggers attack — a paradox that the field had recorded but not pursued. Chen noticed that microsensor fabrication was expensive in ways that seemed unnecessary given the underlying materials science. We can’t know today whether her question turns out to matter at the scale of Faraday’s or Karikó’s. We can know that she is positioned to ask it.

The third element in the Faraday sequence is the one most easily overlooked. Davy didn’t teach Faraday chemistry in a classroom. He gave him access to a working laboratory and real problems. Faraday learned experimental method by doing experiments, not by studying how experiments are done. That experience. has persisted. Sixty-four percent of the 328 contributors in the historical analysis had academic or research laboratory access — not as students receiving instruction, but as participants in pursuing questions. The exposure to real problems, with real uncertainty about outcomes, is very different from exposure to solved problems presented as curriculum.

This is what Chen’s program provided. Mentored research experience — sitting alongside a doctoral student working on an unsolved problem about oxygen measurement in muscle tissue — is not the same as a laboratory course where the result is known before the experiment begins. The anomaly is real. The uncertainty is real. The question that emerges, if one does, comes from actual contact with a problem that hasn’t been resolved.

These conditions are not confined to research universities. Faraday’s preparation happened in a bookshop and a lecture hall. The conditions require foundational exposure to a domain’s real material, contact with something that doesn’t fit neatly into the existing account, and proximity to someone working on real problems. In different eras those conditions have been met through apprenticeship, through patronage, through correspondence networks, through libraries, through deliberate self-education. Chen’s path is one version. It is not the only one.

What the historical record suggests is that the conditions are identifiable. That matters now especially because AI made fluent answers instantly available. People will not stop being curious, but the risk is that the preparation — the foundational exposure, the encounter with genuine anomaly, the proximity to unsolved problems — will be replaced by question-and-answer exchanges that simulate inquiry without building the capacity that makes original questions possible.

Faraday read Marcet before he could verify her claims. He attended Davy’s lectures before he could evaluate Davy’s methods. He spent years in a laboratory handling problems he couldn’t yet fully understand. That sequence — exposure before mastery, anomaly before framework, real problems before solved ones — is what built the judgment to ask the question that hadn’t been asked.

The conditions that produced it are still available. Recognizing them is the first step toward deliberately managing or creating them.

A personal reflection. As I worked through this essay I was reminded of my days as an undergraduate and graduate research assistant working alongside leading professors on their problems. That’s where I learned what it meant to be a professional scholar. To appreciate the craft of research, the questioning, the problem solving, the importance of an open-mind, how to work with others and, perhaps most importantly, to admit not knowing or to being confused, and having the drive to overcome those deficits. Those experiences gave me many of the tools I needed to go out in the world and answer my questions and those of many others.

Sources

On Michael Faraday: James, F.A.J.L. (2010). Michael Faraday: A Very Short Introduction. Oxford University Press. Hamilton, J. (2002). Faraday: The Life. HarperCollins. Faraday, M. (1839–1855). Experimental Researches in Electricity, 3 vols. Richard and John Edward Taylor. Marcet, J. (1806). Conversations on Chemistry. Longman, Hurst, Rees, and Orme.

On Katalin Karikó: Karikó, K. (2023). Breaking Through: My Life in Science. Crown.

On the historical database patterns: Rappaport, S. (forthcoming). Breakthrough Questions: The Question AI Can’t Ask. Early exposure as curiosity trigger: 102/329 contributors (31%). Academic or research laboratory access: 210/328 contributors (64%). Analysis of contributor_cognitive_analysis, json-1st database, n=328–329 major contributors, April 2026.

On Caitlyn Chen: Baker, B.K. (2026). Caitlyn Chen’s path to becoming a physician-scientist. Penn Today, University of Pennsylvania, April 10, 2026.

On Oersted and electromagnetic induction: Williams, L.P. (1965). Michael Faraday: A Biography. Basic Books. The 1831 electromagnetic induction discovery and its relationship to Oersted’s 1820 finding is documented throughout the Faraday secondary literature.

Across the biographies of the more than 300 people I studied who asked breakthrough questions, a few kinds of turning points recur. It can be a piece of evidence contradicting the framework someone has been working within. A collaborator supplying expertise that had been missing. A person displaced from their normal environment, stripped of routine, and forced to think differently. Sometimes it's an institutional change — a new position, an expulsion, an unexpected appointment — opens up questions that hadn't been available before.

Here are four examples showing each of these at work. I’ve taken examples from antiquity to the 20th century, presented them most recent first, to show how turning points persist through eras.

In June 1925, Werner Heisenberg traveled to the island of Helgoland in the North Sea to recover from a hay fever attack that was severe enough to keep him from working in the city of Göttingen. He went alone, packing mathematical tools and carrying an unresolved problem: why do atoms emit light only at specific frequencies that classical physics couldn't predict?

Heisenberg worked through the problem, far away from his colleagues and routine. He abandoned the fundamental assumption that electrons traveled in observable paths around atomic nuclei, the idea that atoms were organized like small solar systems. He reasoned that if the paths couldn't be observed, then they didn't belong in a physical theory. He replaced this model with matrix mechanics, the first complete mathematical framework for quantum theory.#fn-1

A century earlier, in January 1825, Samuel Morse was painting a portrait in Washington D.C. when a letter arrived telling him his wife Lucretia was gravely ill. He left immediately for their home in New Haven. To his great dismay, she died and was buried before he arrived.#fn-2

Lucretia’s loss prompted a question that stayed with him for years: why did news travel so slowly? Seven years passed before he did anything about it until, sailing home from Europe in 1832, he fell into a shipboard conversation about recent experiments with electromagnetism and began sketching a system for sending information by wire. The technical work took roughly six more years, culminating in the first successful demonstration of the telegraph in 1838.

A century and a half before that, in the summer of 1665, the plague closed Cambridge, and Isaac Newton, twenty-two and just graduated, returned to his family's farm at Woolsthorpe, in Lincolnshire. Cambridge would reopen and close again over the next two years as outbreaks recurred, and Newton spent much of that time working alone, away from his university habits.

According to his later account, he acquired a glass prism to investigate what he called "the celebrated Phenomena of Colours” in early 1666. The prevailing view, following Descartes, held that sunlight was uniform — white — and that passing through a prism somehow modified it, adding color. Newton's experiments showed the opposite. Color wasn't something a prism added to light. Color was already there, white light was all of the colors traveling together.#fn-3

Going back even further, to around 2285 BCE, Enheduanna, high priestess of the moon god Nanna at Ur and daughter of Sargon of Akkad, was expelled from her position during a political uprising. The man who seized power drove her from the temple and stripped her of her ritual authority.#fn-4

In exile, and without the institutional position that had defined her standing, she composed the Hymn to Inanna, a poem addressed directly to the goddess, written in the first person, signed with her own name. This is the first signed literary work in recorded human history.

Her decision to sign it, to assert individual authorship in a literary tradition that lacked it, came from her outsider status. Without temple authority, she could speak as herself.

Each of these examples shows that a person's biography and training was necessary but not sufficient for questioning: a turning point was needed. The pollen in Göttingen, the slowness of a horse messenger, the plague that closed Cambridge, and a political adversary's decision to act each opened a specific question at a specific moment.

The data from 328 careers points toward something the individual-genius model doesn't account for: breakthrough capacity appears across many people who have done the preparation. The context they're in when that capacity is ready does a great deal of the selecting of which questions get asked and by whom.

Sources

Heisenberg, W. Physics and Beyond: Encounters and Conversations. Harper & Row, 1971, pp. 60–63.#fnref-1

Mabee, C. The American Leonardo: A Life of Samuel F. B. Morse. Knopf, 1943, pp. 89–94.#fnref-2

Newton, I. "A Letter of Mr. Isaac Newton... Containing His New Theory about Light and Colors." Philosophical Transactions of the Royal Society 6 (1672): 3075–3087; Westfall, R.S. Never at Rest: A Biography of Isaac Newton. Cambridge University Press, 1980.#fnref-3

Hallo, W.W. and van Dijk, J.J.A. The Exaltation of Inanna. Yale University Press, 1968.#fnref-4

This series has been tracing how breakthrough questions get asked and the people who ask them. The question underneath it all: where does the judgment to ask them come from?

In 1995, the University of Pennsylvania demoted Katalin Karikó and cut her salary. Her research into messenger RNA as a therapeutic platform didn’t attract grants. Although her institution had run out of patience, she clung tight to her program.

Karikô was working through a specific molecular paradox: cells use RNA to synthesize proteins constantly without triggering an immune response, yet every synthetic RNA researchers produced was attacked as foreign. Something about the cell’s own RNA was chemically different. But what? The discovery was that replacing the nucleotide uridine with pseudouridine rendered synthetic RNA “invisible” to the immune system. This finding appeared in a 2005 paper she published with Drew Weissman. That landmark publication established the biochemical foundation for the mRNA vaccines administered two decades later.

The decade between her demotion and that paper wasn’t a string of brilliant insights. It was very different: 10 years of technical obstacles generating methodological innovations, then new obstacles, then new approaches — cycles that Karikó describes in her memoir and her Nobel lecture as the necessary structure of the work. The judgment that recognized pseudouridine as the answer was tied to the hard work of advancing methods.

AI analyst Barry Chudakov writes that AI “can simulate questioning but not perform the moral act of questioning” — the act of someone who has done enough of the work to own the consequences of the answer. He isn’t making a technical argument. To Chudakov, questioning is an intentional human act: “When we outsource thinking to AI, we outsource our moral capacity, our ability to ask: What does this mean? Should we do this? What are the consequences here?” His observation, drawn from a recent Elon University canvassing of 386 technology experts, identifies the problem accurately, but it doesn’t explain the mechanism. Why does genuine questioning require the person asking it to have built that capacity? What does building it actually involve?

The historical record offers specific answers.

Across 328 contributors to the history of human knowledge I studied — spanning roughly 5,200 years, from ancient accounting through contemporary computing — roughly two-thirds show sustained effort as a documented prerequisite before breakthrough arrived. This finding remained across every era. The introduction of the printing press, the telegraph, the research university, and digital computing did not change it. Two examples from fields and times quite different from Karikó’s biochemistry show the same pattern.

Composer Béla Bartók spent years traveling to remote Hungarian villages with a phonograph, collecting melodies that cultured Europe didn’t know existed, after his nanny’s singing revealed that what passed for authentic Hungarian folk music was largely urban fabrication. His compositional work was not a smooth elaboration of those discoveries. His letters document recurring cycles: intensive engagement with new material, then months of frustrating attempts to integrate it, then breakthrough synthesis when the creative work had run its course. His essays are explicit that direct, extended contact with source material was required before any compositional advance became possible. Today, an AI asked what Hungarian folk music sounds like returns a comprehensive answer. Bartók’s life work began precisely because the answer he had been given was wrong, and he knew it.

During a visit to the island of Helgoland in the summer of 1925, Heisenberg worked on a problem that had challenged physicists for two decades: why atoms emit light only at specific frequencies that classical physics could not predict. In Physics and Beyond, he describes his pattern — initial confusion with anomalies, periods of mathematical exploration during which conscious analysis gave way to something he could not accelerate or direct, breakthroughs arriving during musical performances or walks in nature rather than at the desk. The matrix mechanics that became the first complete mathematical framework for quantum theory emerged from that process. Like Bartók and Karikó, he required time inside the unresolved problem.

The pattern is consistent enough to describe as a sequence: cognitive friction ignites the question; sustained effort builds the capacity to work on it; apparent stagnation incubates the integration; breakthrough arrives. Each stage depends on the previous one.

Further, the pattern produces judgment, the capacity to recognize which question matters, to know when an answer is wrong despite its fluency, to endure unresolved difficulty long enough for something genuine to emerge — that capacity is built in doing.

Arlindo Oliveira, a computer scientist writing in the same Elon survey, names the stakes:

“In an age where answers are abundant and instantly accessible, the scarce resource is not information but judgment.”

Sources

On Katalin Karikó: Karikó, K., Buckstein, M., Ni, H., & Weissman, D. (2005). Suppression of RNA recognition by Toll-like receptors. Immunity, 23(2), 165–175. Karikó, K. (2023). Breaking Through: My Life in Science. Crown. Karikó, K. (2023). Nobel Prize Acceptance Speech. Nobel Committee for Physiology or Medicine.

On Béla Bartók: Bartók, B. (1931). Hungarian Folk Music. Oxford University Press. Bartók, B. (1976). Essays (ed. Benjamin Suchoff). University of Nebraska Press. Bartók, B. (1971). Letters (ed. János Demény). Faber and Faber. Somfai, L. (1996). Béla Bartók: Composition, Concepts, and Autograph Sources. University of California Press.

On Werner Heisenberg: Heisenberg, W. (1925). Über quantentheoretische Umdeutung kinematischer und mechanischer Beziehungen. Zeitschrift für Physik, 33, 879–893. Heisenberg, W. (1971). Physics and Beyond: Encounters and Conversations. Harper & Row. Cassidy, D.C. (1991). Uncertainty: The Life and Science of Werner Heisenberg. W.H. Freeman.

On productive struggle and the historical database: Rappaport, S. with Zaltman, G. (forthcoming). Breakthrough Questions: The Questions AI Can’t Ask. Project database analysis of 328 major contributors, April 2026.

On the Elon University expert canvassing: Anderson, J. & Rainie, L. (2026). Building a Human Resilience Infrastructure for the AI Age. Imagining the Digital Future Center, Elon University, April 2026. Chudakov, B. and Oliveira, A. quoted therein.

Not everyone is familiar with Herndon, who grew up in Johnson City, Tennessee, and spent formative years in Berlin as a teenager and later as a young adult. Berlin was a hotbed of electronic and experimental music; the culture there supported a collective approach to making and performing sound.^[2] Herndon later studied composition and electronic music at Mills College in Oakland, then completed doctoral work at Stanford’s Center for Computer Research in Music and Acoustics. By the time she built Spawn, she had spent a decade thinking deeply about what technology does to the relationship between a voice and the person it belongs to.

Holly+ followed Spawn in 2021: it’s a platform where fans could license her voice to generate new music, with a consent framework and a revenue-sharing structure built in.^[3] This was one answer to the question of who owns a voice when the voice can be separated from its source.

AI had been a concern in Hollywood since the last strike, and was addressed in the Writer’s Guild and SAG contracts in 2023. But when the production company Particle6 unveiled Tilly Norwood at the Zurich Film Festival in 2025, describing her as the world’s first AI actress, a new round of resistance erupted.^[4] Tilly has no human origin. She was assembled from prompts, voiced by ElevenLabs, animated by Tavus, and scripted by ChatGPT. SAG-AFTRA’s response was immediate: Tilly was “trained on the work of countless professional performers — without permission or compensation.” ^[5]

Particle6’s founder, Eline van der Velden, said she created Tilly partly to warn the industry about what was coming: to force a conversation about AI and performer rights before studios made the decisions without anyone’s input.^[6] The industry’s reaction was swift and hostile. Kevin O’Leary put the film investor’s logic, especially regarding extras and background performers, as plainly as can be said:

“Why couldn’t you simply put AI agents in their place? [You could] save millions of dollars, so more movies could be made.”^[7]

The Hollywood Reporter described the industry as feeling “one update away from apocalypse.”

What Herndon had asked in an academic and artistic context, the entertainment industry was now asking out loud — in contract negotiations, union statements, and legal filings — without an established framework to answer it. Whose voice is this, and on whose terms can it be used?

Holly Herndon saw the question in 2019 and built an answer before the industry needed one. This is what domain depth looks like when it’s working: you can see where a framework is straining before it breaks. The entertainment industry is coming to grips again with a problem that had been brewing for years. The reason? The framework it relied on simply hadn’t kept pace with what the technology could do.

Note: After publishing this piece on June 2, the NY Times ran an article detailing that the question of who owns a performer's voice and likeness is still very much open. (https://www.nytimes.com/2026/06/02/business/media/martin-scorsese-artificial-intelligence.html)

This post is part of a recurring series on breakthrough thinkers and their questions, drawn from Breakthrough Questions: The Question AI Can’t Ask (in progress).

References

1. Herndon, Holly. PROTO. 4AD, 2019. On Spawn and the consent-based training methodology, see Herndon’s interviews with Pitchfork (May 2019) and The Wire (May 2019).↩︎

2. On Herndon’s Berlin periods: Wikipedia, “Holly Herndon,” Early life section; Red Bull Music Academy lecture, Tokyo, 2014; Tape Op Magazine, issue 132. She has described both a high school exchange and a subsequent five-year period living in Berlin in multiple interviews.↩︎

3. Holly+. hollyplus.xyz, launched 2021. Herndon described the platform and its consent and revenue framework in interviews with The Guardian (November 2021) and Resident Advisor (December 2021).↩︎

4. Goodfellow, Melanie. “Talent Agents Circle AI Actress Tilly Norwood As Studios Quietly Embrace AI Technology.” Deadline, September 27, 2025. Gorman, Steve et al. “Hollywood performers union condemns AI-generated ‘actress’ Tilly Norwood.” Reuters, October 1, 2025.↩︎

5. SAG-AFTRA statement on Tilly Norwood, September 29, 2025. Quoted in multiple outlets including Reuters and The Guardian.↩︎

6. Maddaus, Gene. “AI Actress Tilly Norwood Has Hollywood Fuming. Is She a Threat or a Stunt?” Variety, October 8, 2025. Maddaus, Gene. “Tilly Norwood Creator on Hollywood Backlash, Creating Jobs and Full AI Movies.” Variety, November 11, 2025. Brodesser-Akner, Taffy. “I Profile Celebrities for a Living. Nothing Prepared Me for Tilly Norwood.” The New York Times Magazine, May 31, 2026.↩︎

7. Maddaus, Gene. “Kevin O’Leary Suggests Using AI Extras Instead of Humans ‘Could Save Millions’ for Movies Like ‘Marty Supreme.’” Variety, October 21, 2025. https://variety.com/2025/film/news/kevin-oleary-marty-supreme-save-millions-ai-extras-1236558434/↩︎

Both composers’ questions kept generating, as the book’s breakthrough model predicted, but in completely different ways — one passed the question forward as a chain, the next answer becoming the starting point for the one after it; the other branched, spreading influence in multiple directions. We’ll see that it ultimately comes down to whether the question is the kind that can be settled — or the kind that isn't.

In 1948, Pierre Schaeffer was working with a turntable at the studios of Radiodiffusion-Télévision Française in Paris when a locked groove stopped him. A locked groove is different from a skip. A skip jolts the needle out of its groove and sends it somewhere else on the record — pure, unwelcome disruption. A locked groove is a circular track with no spiral, so the needle stays in it, playing the same fragment in a perfect, unending loop. Schaeffer listened to this locked groove over and over again. He didn’t hear a malfunction, he heard music instead.^[1]

That really stunned me as I researched this composer. As a college radio DJ in the LP days, we knew that vinyl records skipped sometimes. A skipping record was a small disaster — the enemy of good radio. It never occurred to me that a repeating sound fragment was anything but a problem to recover from, and fast.

Schaeffer developed a working principle: captured sound is raw material. He went looking for material and found it at the Gare des Batignolles, a railway station, where he directed six engine-drivers to improvise with their locomotives according to his instructions. He didn’t passively record this world, he stripped the sounds of their meaning and listened for what remained: rhythm, tone color, pitch. He recorded them, cut them, layered them, and transformed them into something never heard before. That became “Étude aux chemins de fer” — the Railway Study. He named his method musique concrète: music made from the world itself.

The question he asked from that moment forward was narrow and specific: Can recorded sounds from everyday life become musical material? More than sixty years later, it was still producing answers.

Where Schaeffer came to music through a radio transmitter, Schoenberg came through Vienna’s concert halls. He had grown up steeped in Wagner and Brahms. Wagner’s Tristan und Isolde had pushed Romantic harmony to its boundaries — Schoenberg concluded there was nowhere left to go within it. By 1908, he abandoned tonality entirely, beginning with the final movement of his Second String Quartet, which dispensed with a key signature for the first time.^[2] His question was audacious: How can music evolve beyond tonal harmony while maintaining structural coherence and emotional expression?

Schoenberg worked on it for fifteen years before inventing twelve-tone serialism in the early 1920s — a systematic method for organizing music without tonal centers, built on rows of all twelve chromatic pitches and their mathematical permutations.^[3] The Second Viennese School followed. Berg and Webern became major composers. Serialism entered the conservatory; its compositional ideas are taught and still refined. But even with this embrace by the academy, serialism failed to catch on with the listening public at large. This is not to say that great pieces weren’t composed — they were: operas, piano pieces, all manner of works. They are performed, notably less often than other works. Rather than producing a lineage, serialism became a resource, even a component. Stravinsky adopted it late in his career. Bernstein and Copland used it. Bartók, Britten, and jazz pianist Bill Evans each borrowed from it. Composers took what served them and integrated it as they saw fit. Others pushed back against serial complexity altogether — Terry Riley’s In C in 1964 built from repetition and consonance arguably led the movement; Reich and Glass followed, and Brian Eno’s ambient records of the late 1970s carried the same impulse into recorded sound itself.

Schoenberg’s question branched, but it didn’t produce the kind of transmission Schaeffer’s question did — where each answer became the next starting point, passed forward to the next person.

The difference between the questions runs deeper than public or critical approval. Schaeffer asked what music can be made from — a question answerable in the studio. You put a recorded train on a turntable, listening gives you the answer. Musique concrète led to electroacoustic music, electroacoustic to sampling, sampling to hip-hop, hip-hop to Bartholomäus Traubeck reading tree rings on a turntable in 2011.^[4] Each answer became the next link in a chain. Music kept evolving because the answers were usable and built on one another.

Schoenberg asked what music should be. That question is beyond what aesthetic validation can decide — no experiment returns a verdict on what music ought to sound like, and listeners make their own judgments.

That’s the refinement my own conclusion needed. My chapter argues that aesthetic validation produces layering — and it does. But within that layering, some questions generate chains because the question is narrow enough, and practical enough, to be answered by whatever tools the domain actually has. Schaeffer’s question was settled with a turntable and a good ear.

You can’t know which way a question will go at the moment it’s asked. Schaeffer didn’t know in 1948 that his locked groove would still be generating answers sixty years later. The trajectory only becomes visible in retrospect — which is why my book covers five thousand years of evidence, from Kushim to the present.

Schaeffer’s locked groove outlasted a lot of grand theories about the future of Western music. A question narrow enough to be answered by the tools at hand tends to pass forward — each answer opening the next, the question moving intact from one person to the next. A question that exceeds what the domain can validate tends to branch — spreading influence in all directions, generating argument, reaction, and borrowing, but no single forward transmission.

Both keep generating. But only one passes the question on.

This post is part of a recurring series on breakthrough thinkers and their questions, drawn from Breakthrough Questions: How Knowledge Evolves Through the Questions We Ask (forthcoming, Wiley).

1. The 1948 locked groove discovery and the Railway Study are documented in Schaeffer, Pierre. Treatise on Musical Objects: An Essay across Disciplines. Translated by Christine North and John Dack. University of California Press, 2017 (originally published 1966). The Gare des Batignolles recording session, including the six engine-drivers improvising to Schaeffer’s instructions, is documented in Föllmer, Golo. “Pierre Schaeffer: Études aux chemins de fer.” See also Kane, Brian. Sound Unseen: Acousmatic Sound in Theory and Practice. Oxford University Press, 2014.↩︎

2. Schoenberg, Arnold. String Quartet No. 2 in F-sharp minor, Op. 10 (1908). The final movement, setting Stefan George’s poem “Entrückung,” omits a key signature entirely. For the analytical account, see Ross, Alex. The Rest Is Noise: Listening to the Twentieth Century. Farrar, Straus and Giroux, 2007, pp. 52–60.↩︎

3. The twelve-tone method was first fully applied in Schoenberg’s Piano Suite, Op. 25 (1921–23). Ross, The Rest Is Noise, pp. 60–65. See also Schoenberg, Arnold. Style and Idea: Selected Writings. Edited by Leonard Stein. University of California Press, 1984.↩︎

4. Traubeck, Bartholomäus. Years (2011). Tree-ring data from cross-sections of various trees is translated into piano music via a modified turntable and custom software. Documentation at traubeck.com.↩︎

In 1896, Henri Becquerel discovered that uranium emitted invisible rays, which was quite an achievement. He published his findings and then moved on to other subjects. Most contemporary physicists did the same. The rays were worth noting, but without being able to explain them, uranium rays were just an interesting finding — potentially nothing more than a footnote.

Marie Curie chose to study the rays further. The reason? She detected an anomaly in her measurement data and wanted to understand it.

Testing pitchblende, the ore that contains uranium, with an electrometer her husband Pierre and his brother had developed, she tracked the radiation from sample after sample, measuring extraordinarily weak electrical currents. The readings were precise, consistent, and wrong: They showed pitchblende was more radioactive than pure uranium.^[1]

Curie realized that shouldn’t have been possible. The ore is mostly uranium. Pure uranium should have been more radioactive than any mixture containing it., yet the numbers, invariably, said otherwise. Becquerel had seen the same ore. Other physicists had handled it. No one had asked about this discrepancy.

Curie asked, though her question didn’t come to her fully formed. It went through three stages, each one defined more sharply than the last.

First: What causes uranium’s radiation? That was the obvious question, the one Becquerel had essentially posed and left unanswered.

Second, after confirming the pitchblende anomaly: Is radioactivity a general atomic property, or is it specific to uranium? This is where her thinking shifted from the specific to the general. If pitchblende was more radioactive than its uranium content could explain, something else in the ore was radioactive too. That meant radioactivity wasn’t a quirk of one element. It was a property atoms could have.

Third, building on the more general question: How many radioactive elements exist? That one opened a research program that outlasted her life and the lives of the scientists who inherited it.

What she actually changed when she asked whether radioactivity was an atomic property is worth stopping to think about because it shows how breakthrough questions work.

The prevailing model of the atom in 1896 held that atoms were stable, indivisible, and inert. They didn’t change. They didn’t emit energy spontaneously. They were the fixed units matter was built from, reliable and permanent. This was essentially the atomic theory of Democritus improved upon by Dalton, which I’ve written about before.^[2]

Curie’s question didn’t attack that model. She wasn’t a theorist poking holes in an established framework. She was an experimentalist following a measurement that didn’t fit. But where the measurement led was devastating for the stable-atom model.

If radioactivity was an atomic property, atoms were doing something. That something was emitting energy from inside themselves, without any external source. That meant atoms had internal structure, internal dynamics. The “indivisible, inert” picture was proven wrong.

Her biography is worth a look here, because the move she made — treating a numerical anomaly as the real question rather than experimental noise — wasn’t accidental.

Curie had learned to work against institutional resistance before she entered a laboratory. She grew up in Warsaw under Russian occupation, where higher education for women was illegal. She attended the Flying University,^[3] a clandestine network of lectures held in private apartments, locations rotated to stay ahead of the authorities. You don’t emerge from that with a strong prior toward accepting what the authorities declare off-limits. She arrived in Paris in 1891 at 24. Those attitudes came with her.

Curie’s pursuit of her questions required a commitment that was physical as much as intellectual.

She and Pierre worked in a converted shed, leaking in rain, glacial in winter. She processed pitchblende by the ton, literally, stirring boiling vats of ore with an iron rod to separate the fractions that showed the most radioactivity. The work was repetitive, exhausting, and conducted in dangerous conditions, though they didn’t fully understand how dangerous. In July 1898, she announced polonium. In December 1898, radium. By 1902, she had isolated one-tenth of a gram of pure radium from several tons of ore.

The isolation of radium is often treated as the triumph, but the question came first. The isolation was proof the question was right. Relating this to our four-element model of breakthrough questions, all four aligned: a testable theory, question, method, and validation.^[4]

What that question eventually opened is too large to cover here. Ernest Rutherford showed that radioactive decay caused atomic transmutation, one element literally becoming another, which meant atoms weren’t indivisible.^[5] That led to the discovery of atomic nuclei, protons, neutrons, and nuclear forces. Rosalyn Yalow borrowed radioactive isotopes to measure hormones at concentrations of trillionths of a gram, winning the Nobel in 1977.^[6] Lise Meitner worked out the mathematics of nuclear fission on a walk through the Swedish woods in 1938.^[7]

All of it traces back to a woman looking at a number that shouldn’t have been that high, and refusing to let it be a footnote in science.

Breakthrough Questions is a reader-supported series drawn from The Question AI Can’t Ask, a book in progress tracing 5,200 years of discovery through the people who asked the questions that changed what we know. New posts every Tuesday.

1. For Curie’s systematic measurement of pitchblende and the anomaly that led to polonium and radium, see Barbara Goldsmith, Obsessive Genius: The Inner World of Marie Curie (New York: W.W. Norton, 2005), 98--124. For the electrometer methodology, see Marie Curie, Radioactive Substances (New York: Philosophical Library, 1961; orig. 1903).↩︎

2. For the atomic theories of Democritus and Dalton as antecedents to Curie’s framework, see the earlier post in this series: “What Is Matter Made Of?”, Breakthrough Questions, https://breakthroughquestions.substack.com/p/why-some-questions-wait-centuries

Breakthrough Questions

Essays on the questions that changed everything — and what they reveal about breakthrough thinking in the age of AI."

By Stephen Rappaport

Breakthrough Questions

Essays on the questions that changed everything — and what they reveal about breakthrough thinking in the age of AI."

By Stephen Rappaport

3. On the Flying University and Curie’s education under Russian occupation, see Susan Quinn, Marie Curie: A Life(New York: Simon and Schuster, 1995), 46--62.↩︎

4. For the four-element model of breakthrough questions, see “What Makes a Breakthrough Question,” Breakthrough Questions, https://breakthroughquestions.substack.com/p/why-some-questions-wait-centuries↩︎

5. Ernest Rutherford and Frederick Soddy, “The Cause and Nature of Radioactivity,” Philosophical Magazine 4 (1902): 370--396.↩︎

6. Rosalyn Yalow, Nobel Lecture, December 8, 1977, Nobel Prize Foundation.↩︎

7. Ruth Lewin Sime, Lise Meitner: A Life in Physics (Berkeley: University of California Press, 1996), 233--240.↩︎

Read more

If you enjoy Breakthrough Questions, please invite friends to subscribe and read with us. I’m not interested in monetizing Breakthrough Questions, I’m interested in coalescing a community around questions.

How to participate

1. Share Breakthrough Questions. When you use the referral link below, or the “Share” button on any post, you'll get credit for any new subscribers. Simply send the link in a text, email, or share it on social media with friends. Should Breakthrough Questions become a book, you’ll get advance notice and a publication discount.

Refer a friend

To learn more, check out Substack’s FAQ.

Thank you for helping get the word out about Breakthrough Questions!

His careful observations on the Galápagos showed him each island had distinct varieties of finches and tortoises. He noticed fossil mammals in South America resembled living species in the same regions but weren’t identical to them. Darwin’s evidence contradicted natural theology, the dominant idea then, which said species were fixed and purposefully designed.^[2]

In July 1837, Darwin opened his first transmutation notebook and started building a framework that could account for what he’d seen. His original question immediately generated new ones. Did species actually vary? The Galápagos finches answered that. Was variation inherited? For that, he needed evidence that offspring resembled parents in their particular traits. What, in fact, drove the process of natural variation?”^[3]

Just about a year later in September 1838, an insight came to him while reading Malthus on population. If you’re not familiar with the story, Malthus had argued that populations grow faster than food supplies, producing a struggle for survival. Darwin recognized this applied to natural variation. Why? He realized that more organisms are born than can survive, those with favorable variations outlive others and are then able to pass those traits to the next generation. Natural selection was his answer to the original question.^[4]

But the answer generated twenty years of new questions.

Did domesticated varieties show the kind of variation natural selection required? He bred pigeons and produced dramatic differences in beak shape, feather color, and body size, just within a few generations. That answered whether usable variation existed. It also raised a question he hadn’t started with: what was the relationship between artificial selection by breeders and the undirected process he was proposing for nature?^[5]

Did the principle hold across different organism types? He spent eight years dissecting barnacles, publishing four monographs documenting variation within and between species. Darwin’s barnacle work confirmed that variation operated across radically different life forms. It also revealed patterns in reproductive strategies and geographic distribution he hadn’t anticipated — which raised questions about how variation worked differently depending on how an organism reproduced.^[6]

Could small variations accumulate into large changes? Plant hybridization experiments showed traits could blend, split, and recombine across generations. Farmers and breeders across three continents had been doing this deliberately for generations and seeing exactly the results his theory predicted. ^[7]

Each investigation answered the question that prompted it and raised further questions Darwin hadn’t known he needed to ask. His extensive notebooks document the pattern, and his correspondence shows him actively seeking counterexamples from breeders, botanists, and geologists — finding the places where his framework might fail. . He anticipated objections and addressed them before critics could raise them because he’d already raised them himself. Chapter 6 of On the Origin of Species documents is titled “Difficulties on Theory” where he lists and addresses them in turn. ^[2-1]

When the book appeared in 1859, Thomas Huxley read it and said he couldn’t think why he hadn’t thought of it himself.^[8] The core insight — natural selection as a mechanism for species change — had been available to anyone who connected Malthus and natural variation. Darwin’s contribution was following the cascade of questions that insight generated.

The twenty-one-year delay between insight and publication was the time required to follow a question through every branch it produced.

Most questions that change a field work this way: they arrive as starting points, not finished products. Darwin’s question about species change couldn’t be answered without answering questions about variation, which couldn’t be answered without answering questions about inheritance and selection pressure, which couldn’t be answered without questions about reproductive strategies and geographic distribution. Each answer revealed what still needed asking. Darwin’s questioning didn’t end with publication. On the Origin of Species generated a new round of questions that took biology another seventy years to answer.

The question that changes a field tends to be the one that keeps producing questions the original asker never anticipated.

1. Darwin sailed on HMS Beagle from December 1831 to October 1836. For the voyage observations and their significance, see Janet Browne, Charles Darwin: Voyaging (New York: Knopf, 1995), 186–305.↩︎

2. Janet Browne, Charles Darwin: Voyaging (New York: Knopf, 1995), 186–305. For the Galápagos observations specifically, see 304–305. For the South American fossils, see 186–194.↩︎↩︎

3. Darwin’s transmutation notebooks (Notebooks B through E, 1837–1839) are transcribed in Paul H. Barrett et al., eds., Charles Darwin’s Notebooks, 1836–1844 (Cambridge: Cambridge University Press, 1987).↩︎

4. Thomas Robert Malthus, An Essay on the Principle of Population (London: J. Johnson, 1798). For Darwin’s reading of Malthus in September 1838 and its significance, see Browne, Voyaging, 385–390.↩︎

5. Charles Darwin, The Variation of Animals and Plants under Domestication, 2 vols. (London: John Murray, 1868). The pigeon breeding work is treated in vol. 1, chapters 5–6.↩︎

6. Charles Darwin, A Monograph on the Sub-class Cirripedia, 2 vols. (London: Ray Society, 1851–1854). For the significance of the barnacle years to Darwin’s thinking about variation, see Browne, Charles Darwin: The Power of Place (New York: Knopf, 2002), 36–60.↩︎

7. Charles Darwin, On the Various Contrivances by which British and Foreign Orchids are Fertilised by Insects(London: John Murray, 1862). For the broader botanical correspondence and hybridization work, see Browne, The Power of Place, 148–195.↩︎

8. Thomas Henry Huxley’s response is documented in Adrian Desmond, Huxley: From Devil’s Disciple to Evolution’s High Priest (Reading, MA: Addison-Wesley, 1997), 252–253.↩︎

Richard Feynman’s father, Melville, sold uniforms for a living, but he had strong opinions about what it means to actually know something. Melville would take his young son to the Catskill Mountains during summers and point to a bird. He could name it in English, Italian, Portuguese, even Japanese. Then he would say: the name tells you nothing about the bird. What does it eat? How does it move when it lands? What does it do when it rains? The name is not the thing.^[1]

His dad’s questioning led the future Nobelist to ask over and over: what is actually happening here? I’m no Feynman, but his father’s question is one that has driven my desire to approach understanding what I didn’t. The other question I ask regularly is “what is missing?” but that’s for another post.

At twenty-four, Feynman started working alongside the greatest physicists of their time — Bohr, Fermi, Bethe, all working on the atomic bomb at Los Alamos. He had a strong sense of himself and a passion for getting to the bottom of things. When Niels Bohr, then the most famous physicist in the world, wanted to work through a problem honestly, he asked for “the young one” — Feynman — because Feynman would say when the reasoning was wrong. Feynman didn’t accept things at face value. He picked the locks on classified filing cabinets to know whether the security system worked or not.

After the bombs fell on Hiroshima and Nagasaki, Feynman went to Cornell and stopped working. He had spent years building something that killed over a hundred thousand people, and physics had stopped feeling worth doing. He eventually got out of his funk but what did it? A plate tossed in the air by a student in the cafeteria. Feynman watched it wobble and spin. He started calculating the wobble out of pure curiosity — a diversion that helped reignite his playfulness of mind. Within months he was working again on the problem that would define his career.^[2]

That problem was quantum electrodynamics, or QED — the physics of how light and matter interact at the smallest scales. The existing theory was a mess. It produced some correct predictions, but the equations also spat out results that had no physical meaning. Two physicists, Julian Schwinger in the United States and Sin-Itiro Tomonaga in Japan, fixed the math using extraordinarily dense and complicated calculations. Both got it to work. Good as it was, Feynman, a visual thinker, needed to see it.

Feynman went back to the beginning, to the type of question his father might have asked: what actually happens when two electrons interact?

He imagined that an electron doesn’t take one path from point A to point B. It takes every possible path at the same time. What’s observed is the result of all those paths added together. From that image alone — before writing a single equation — he built what became the path integral formulation of quantum mechanics and Feynman diagrams. Schwinger once said he could calculate anything; he just couldn’t picture it. Feynman always started with the picture.^[3]

Feynman’s habit was to treat every existing explanation as a starting point to be questioned, not a conclusion to be trusted. He didn’t ask “how do we fix QED?” He asked “what is actually happening?” — and that earlier, more basic question is where the work began of challenging a framework that was insufficient.The 1974 Caltech commencement address is where he named the habit explicitly.

Feynman called it “Cargo Cult Science,” after something that happened in the South Pacific during World War II. Islanders had watched American military planes land and bring food and supplies. After the war ended and the planes stopped coming, some islanders built wooden airstrips, carved headphones from bamboo, and went through all the motions they had seen — hoping the planes would return.

Feynman said some scientists do the same thing. They run experiments, publish papers, cite each other’s work — all the right motions. What they skip is what he called utter honesty: the requirement to report not just what your experiment showed but what it didn’t show, to actively look for every reason your own theory might be wrong, to be as hard on your own conclusions as you are on everyone else’s. That habit, he said, is what separates real science from the imitation of it.^[4]

Note to readers: Darwin employed a similar approach to Feynman, stress-testing ideas before publication. See The Question That Wouldn’t Stop Asking Questions,

This article is part of a recurring series profiling breakthrough thinkers and their questions, drawn from the Breakthrough Questions database of 328 contributors to human knowledge spanning 5,200 years. Profiles are augmented with data analyses and thought pieces on related topics.

1. Richard P. Feynman, Surely You’re Joking, Mr. Feynman! Adventures of a Curious Character (New York: W. W. Norton, 1985). The bird story and Melville’s teaching philosophy are documented in Feynman’s own account. The wobbling plate episode is also described there.↩︎

2. James Gleick, Genius: The Life and Science of Richard Feynman (New York: Pantheon, 1992). Gleick documents the Los Alamos period, the Bohr anecdote, and the Cornell recovery in detail.↩︎

3. Richard P. Feynman, QED: The Strange Theory of Light and Matter (Princeton: Princeton University Press, 1985). Feynman’s own plain-language account of the path integral and the physical picture underlying the diagrams.↩︎

4. Richard P. Feynman, “Cargo Cult Science,” commencement address, California Institute of Technology, 1974. Reprinted in Surely You’re Joking, Mr. Feynman! and in The Pleasure of Finding Things Out (Cambridge, MA: Perseus Books, 1999).↩︎

When his colleague Jakob Kolletschka died from a wound infection after a scalpel nick during an autopsy, the pathology looked identical to childbed fever — Semmelweis made the connection. In those days, medical students and physicians moved directly from performing dissections to delivering babies without washing their hands. Midwives didn’t perform dissections, they were clean, comparatively. Semmelweis introduced mandatory handwashing in chlorinated lime solution. Mortality in his ward dropped to rates comparable to the midwives’.

Semmelweis had solved the mortality problem. He established cause and effect, and . the data was clear. Yet the medical establishment of Europe ignored him for nearly two decades. He died in 1865 in a mental asylum, possibly from the same kind of infection he had spent his career trying to prevent.

Most people who know this story treat it as a tragedy of institutional stubbornness — the brilliant man defeated by lesser minds unwilling to accept an inconvenient truth. Semmelweis failed not because his answer was wrong, but because he never fully understood what his question required.

He assumed his problem was one of evidence. Like most scientists he believed that by rigorously and repeatedly showing that handwashing saved lives, the practice would spread. Semmelweis generated more data, more demonstrations, more publications. What he didn’t see was that his question lived inside a system he had barely considered.

Germ theory didn’t exist in 1847. Without a theoretical mechanism explaining why handwashing worked, clinical authorities had nothing to ground an endorsement on. His findings also implied something his colleagues found intolerable: that physicians were killing their patients by carrying what he called “cadaverous particles” from the morgue to the delivery room. That was a direct assault on professional identity. How could doctors be responsible for death?

He kept offering evidence to a resistance that wasn’t about evidence. More data would not change the status quo. A testable theory that made validated his findings would.

Questions exist inside a context — the theoretical frameworks that make an answer conceivable, the tools available to investigate it, and the validation mechanisms that determine whether an answer will be accepted at all. Semmelweis had the tools and the answer. He lacked theory and validation. Those missing pieces proved fatal.

I spent most of my professional life working on questions — not asking them so much as helping people figure out what they were actually trying to know. In knowledge centers inside large organizations, or at the Advertising Research Foundation, the hardest part was rarely the research. It was the question. A question that can’t be framed precisely enough to investigate isn’t yet a question. At best, it’s a direction.

I watched the Semmelweis’ failure pattern play out in advertising when Facebook introduced the Like button in 2009. Within a few years, the question I heard constantly in conference rooms and client meetings was some version of: What’s our like strategy? People assigned dollar values to likes. Analysts tried to divine trends in like counts and deduce what they meant for brand health. Entire measurement programs were built around them.

In 2012, I started work on The Digital Metrics Field Guide at the ARF with the aim of explaining metrics. Digital media created so many metrics that heads spun. To give you an idea of how many, I reduced a list of 350 candidates to 197 that I researched and wrote about. When I addressed the like strategy’ question I realized that the question assumed things that hadn’t been established: that likes had a stable meaning, that they connected to business outcomes through a known mechanism, and that Facebook’s framing of the metric could be trusted as a foundation for strategy. None of those conditions had been validated. The theoretical framework was Facebook’s marketing. The validation mechanism was essentially nonexistent. People were building strategies on a metric whose meaning and value were, at that point, largely invented.

Facebook eventually deprecated several of the metrics that had driven years of “like strategy” conversations, because they turned out to be too difficult to interpret and impossible to act on. All that effort, directed at a question built on an assumption that a like meant something as a business metric, rather than being a feature Facebook had designed to promote engagement and eventually track user interests.

Semmelweis and the like strategists failed by the same mechanism, separated by a century and a half. Neither group lacked intelligence or diligence. What both lacked was a complete understanding of what their question actually required — the theoretical foundation that would make an answer meaningful, and the validation conditions that would allow it to be accepted.

This is the problem I kept running into, and eventually became determined to investigate seriously. Not why smart people ask bad questions, but what distinguishes questions that lead somewhere from questions that consume enormous effort and produce nothing. What do the people who asked the questions that sent knowledge in a new direction — in physics, biology, music, economics, computing, communication — understand about the conditions their questions require?

Three years ago, I began an investigation that produced a database of 328 contributors to human knowledge spanning 5,200 years. The patterns that emerged were not what I expected. Breakthrough questions don’t come from genius alone. They come from a combination of genius and specific conditions — theoretical frameworks under strain, tools newly capable of investigation, validation mechanisms ready to render a verdict. The questioner who can read those conditions, and frame a question that fits them, is doing something categorically different from the one who simply asks what seems interesting.

That is what my essays are about: The conditions that make a question answerable — and what the people who have asked the best questions in history understood about them.

People who changed their fields weren’t better at searching existing knowledge. They asked questions that existing knowledge, especially theories and frameworks, couldn’t generate.

Today’s AI systems are trained on existing literature. They can find patterns, predict connections, and surface overlooked relationships between concepts already named in that literature.

Researchers at the Karlsruhe Institute of Technology recently trained a machine learning system on 221,000 materials science abstracts. They extracted key concepts, mapped how those concepts related to one another, and asked the system to predict which pairs researchers would connect next. Expert scientists judged 26% of the suggestions worth pursuing. In a 30-minute session, that’s several promising research directions generated before lunch.1

Now let’s look at what the system is actually doing. It builds a graph from concepts already named in the existing literature. Then it finds pairs sitting close together that haven’t been explicitly linked yet. Its best predictions connect concepts already just 2 steps apart. The further apart two concepts sit in the graph, the worse the system performs.

AI can’t ask questions about concepts that don’t yet exist.

When Democritus proposed that all matter consists of indivisible particles moving through empty space, what he called “void,” he wasn’t connecting nearby nodes in an existing concept graph. There were none to connect. There wasn’t one for atom. Nor was there one for void. He introduced concepts a graph drawn during his time didn’t contain — and in doing so asked a question the graph couldn’t possibly have generated. AI has not yet demonstrated the capacity to do that: to introduce a concept the existing graph doesn’t contain, recognize where a prevailing framework is failing, and ask the question to replace that failed framework with a stronger one.

The researchers who changed how we understand matter, life, energy, music, information and most other fields weren’t better at searching what was already known. They were asking questions the existing knowledge could not answer.

Footnotes

[^1] Marwitz, Thomas, Alexander Colsmann, Ben Breitung, Christoph Brabec, Christoph Kirchlechner, Eva Blasco, Gabriel Cadilha Marques, Horst Hahn, Michael Hirtz, Pavel A. Levkin, Yolita M. Eggeler, Tobias Schlöder, and Pascal Friederich. “Predicting New Research Directions in Materials Science Using Large Language Models and Concept Graphs.” Nature Machine Intelligence, 2026. https://doi.org/10.1038/s42256-026-01206-y.