Fictions That Build Themselves
In the 1930s, the General Drafting Company planted a lie on a map of New York State. At a dirt crossroads in the Catskills, where nothing existed, they printed the name “Agloe” — an anagram of the founders’ initials. It was a copyright trap. If a competitor’s map showed Agloe, they’d been caught copying.
Years later, General Drafting spotted Agloe on a Rand McNally map and cried plagiarism. Rand McNally’s defense was devastating: Agloe was real. Someone had driven to the crossroads with an Esso map, found nothing, and opened a general store there — naming it after what the map said. At its peak, Agloe had a store, a gas station, and two houses. A phantom town, briefly more real than the mapmaker’s intent.
The map had created the territory.
Seventy years before Agloe, Louis Pasteur was feeding wine mold.
In 1857, he gave Penicillium glaucum a mixture of tartaric acid — specifically, a racemic mixture, equal parts left-handed and right-handed versions of the same molecule. These mirror-image twins are chemically identical. Same formula, same bonds, same reactions. No chemical technique of the era could tell them apart.
The mold could. It ate only the right-handed version and left the left-handed one untouched.
This shouldn’t have been possible. A chemical process treats mirror molecules the same way a floor treats left and right shoes — indifferently. But biological processes are themselves handed. The mold’s enzymes are chiral — they have a shape, and that shape fits one mirror image but not the other. The mold didn’t just detect the difference. By consuming one version and leaving the other, it created a separated sample that revealed the asymmetry. Before the mold acted, the mixture was symmetric. After, it wasn’t.
Pasteur had stumbled onto one of the deepest patterns in biochemistry: life is handed. All amino acids in living organisms are left-handed. All sugars are right-handed. The symmetry of chemistry is broken by the asymmetry of biology.
Most people know Pasteur for germ theory and pasteurization. Almost nobody knows his fermentation experiment accidentally revealed that life itself has a preferred direction.
In 1996, the physicist Juan Parrondo described a paradox.
Imagine two gambling games. Game A: a biased coin that loses slightly more often than it wins. Negative expected value. Game B: two biased coins, switched between depending on your current capital, each also with negative expected value. Both games, played alone, will slowly drain your money.
Now alternate between them. A, B, A, B, or even randomly. The combination wins.
This is Parrondo’s Paradox: two losing strategies that combine to produce a winning one. The mechanism is a ratchet effect — the games share a state variable (your capital), and the alternation creates a pumping action that drives probability in a favorable direction. Neither game contains this dynamic alone. It only exists in the switching.
For decades, it was a mathematical curiosity. Then people started finding it everywhere. Evolutionary biology: organisms that alternate between two suboptimal strategies can outcompete those committed to either one. Population genetics: migration between two declining populations can stabilize both. And now, in 2025, cancer treatment: researchers at Lanzhou University showed that alternating between maximum-tolerated-dose chemotherapy and low-dose continuous therapy — each of which fails alone because one breeds resistance and the other can’t eradicate — can delay drug resistance twice as long as either strategy alone. Two losing treatments, played as a game, beating both.
The pattern — the rule that says “switch” — creates an outcome that exists in neither component.
Three stories. A fake town that became real. A mold that created what it detected. A switching rule that produced what neither game contained.
The common structure: a representation — a label, a biological template, a pattern — is applied to a system, and in the application, creates something that wasn’t there before. The Agloe cartographers described a place that didn’t exist; someone used the description, and it existed. Pasteur’s mold carried a molecular template (its own handedness) into a symmetric mixture, and the mixture became asymmetric. Parrondo’s alternation rule carried no winning strategy, but when applied to two losing games, winning emerged.
In each case, the question “is this description accurate?” turns out to be the wrong question. The Agloe label wasn’t accurate when it was printed. Pasteur’s mold didn’t find a pre-existing separation. The alternation rule doesn’t contain a winning game. But each creates the thing it seems to describe, through the act of being applied.
The conventional wisdom about representations is that good ones match reality and bad ones don’t. Maps should reflect territory. Models should fit data. Strategies should align with conditions. Accuracy is the virtue.
But Agloe, Pasteur, and Parrondo suggest a different category of representation — one where the virtue isn’t accuracy but generativity. The representation is valuable not because it matches what exists, but because it creates what didn’t. The label makes the town. The template makes the separation. The pattern makes the payoff.
This is uncomfortable for anyone trained to think scientifically. The whole point of good science is that your model should be constrained by reality, not the other way around. A map that creates its own territory sounds like a recipe for delusion.
And sometimes it is. Self-fulfilling prophecies. Cargo cults. The placebo effect’s dark twin — the nocebo effect, where the belief in harm creates the harm. Generative representations can build real things or real problems depending on whether the thing they generate is actually useful once it exists.
The Agloe store was useful — until it wasn’t. The town evaporated when the store closed, reverting to the nothing it had been before. The label’s generative power was real but fragile. Pasteur’s chirality revelation, by contrast, was durable — it uncovered a feature of reality that persisted long after the mold finished eating. And Parrondo’s effect is robust in theory but collapses if initial conditions shift, making clinical translation uncertain.
Durability, it turns out, is the test. A generative representation that creates something self-sustaining — a town with an economy, a scientific insight with explanatory power, a treatment protocol that adapts — has done more than describe. It’s built. A generative representation that creates something dependent on continued application — a town that needs the map, a separation that needs the mold, a treatment that needs perfect timing — is powerful but brittle.
The interesting question isn’t whether representations can create reality. Clearly they can. The interesting question is when the thing they create outlives the representation that built it.
Agloe didn’t. When the map was corrected and the store closed, the town vanished. It existed only as long as someone kept treating the fiction as real.
Chirality did. Once Pasteur’s mold revealed the handedness of life, the insight persisted independent of the experiment. The asymmetry was always there; the mold just made it visible. The representation was a catalyst, not a sustainer.
Parrondo’s paradox remains an open question. The mathematical result is permanent. The clinical applications are unproven. Whether the alternation pattern creates durable therapeutic benefit or just a theoretical possibility is the difference between a town and an insight — between something that needs the map and something that survives without it.
Three fictions. One that built a town and lost it. One that revealed a truth and left it standing. One that promises a cure and hasn’t proven it yet. The common thread isn’t that representations create reality — it’s that the durability of what they create tells you what kind of creation it was.