02 · The Reframe

THE SHORTEST PATH — A History of the Art and Science of Problem-Solving

Königsberg, 1735

The East Prussian city of Königsberg sat astride the river Pregel, split by the water into four pieces of land — two riverbanks and two islands — connected to one another by seven bridges. Sunday strollers there had, for some time, amused themselves with a standing challenge: could a person walk a route through the city that crossed every one of the seven bridges exactly once, and no bridge twice? Nobody had managed it. Nobody had proved it was impossible either. It was, by every reasonable measure, a trivial puzzle — the kind of thing argued over on an afternoon walk, not brought before an Academy of Sciences.

It reached Leonhard Euler anyway, relayed through a chain of acquaintances: Heinrich Kühn, a mathematics professor in Danzig, mentioned it to Carl Leonhard Gottlieb Ehler, the city's mayor and an amateur enthusiast, who wrote to Euler in St. Petersburg asking for a solution "worthy of your great genius." Euler's first reaction, preserved in a letter he wrote on March 13, 1736 to Giovanni Marinoni, an Italian astronomer at the Viennese court, is almost dismissive: "This question is so banal, but seemed to me worthy of attention in that neither geometry, nor algebra, nor even the art of counting was sufficient to solve it." He goes on to wonder, half to himself, whether the puzzle belongs to a discipline the philosopher Gottfried Leibniz had once called for but never built — a geometria situs, a "geometry of position," concerned with arrangement rather than measurement (Sachs, Stiebitz & Wilson, "An Historical Note: Euler's Königsberg Letters," Journal of Graph Theory, 1988).

Three weeks later, replying directly to Ehler, Euler is blunter still — and reveals, in the same breath, exactly how little he thinks of the kind of solution he is about to hand over: "Thus you see, most noble Sir, how this type of solution bears little relationship to mathematics, and I do not understand why you expect a mathematician to produce it, rather than anyone else, for the solution is based on reason alone, and its discovery does not depend on any mathematical principle" (Sachs, Stiebitz & Wilson, 1988). Read that carefully. The man who is, at this exact moment, inventing what would become graph theory is telling a mayor, in writing, that what he has just done barely counts as mathematics. He is wrong — spectacularly, historically wrong — and the way he is wrong is the entire subject of this chapter.

Here is what Euler actually did, presented to the St. Petersburg Academy on August 26, 1735 and published the following year as Solutio problematis ad geometriam situs pertinentis ("The solution of a problem relating to the geometry of position") (MAA Convergence; Seven Bridges of Königsberg, Wikipedia). He noticed that the exact streets a walker takes within a landmass, the exact spot on a bridge where a foot lands, the width of the river, the layout of the buildings — all of it is irrelevant to the question being asked. The only fact that matters is which landmasses each bridge connects. So he threw the city away. In its place he drew four points, one for each landmass, and seven lines connecting them, one for each bridge — the first drawing in history of what we would now instantly recognize as a graph. Every possible walking route through Königsberg's actual, cobblestoned, physically detailed streets became, in this new picture, nothing more than a sequence of dots visited by following lines.

Once the city has been stripped down to four dots and seven lines, the puzzle nearly solves itself. Consider any landmass that is not the start or end of the walk. Every time the walker arrives at it over one bridge, they must leave over a different one — arrivals and departures pair up, bridge by bridge. That means any landmass visited only in passing must have an even number of bridges touching it: each visit uses up exactly two, one in and one out. Count the bridges touching each of Königsberg's four landmasses in the real map, and every single one comes out odd — three, three, three, and five. A walk that starts and ends at different points can tolerate exactly two landmasses with an odd count (the start and the end); Königsberg has four. No route exists, full stop, for any walker, using any order of streets whatsoever — and Euler's argument proves it in roughly the space of two sentences, with no need to try a single one of the geographically enormous number of possible walking routes through the actual city.

Notice what happened to the size of the problem. Before the reframe, "does a walk exist?" is a question about an intractably large space of possible paths through real streets, checkable only by trying route after route and hoping to either find one or run out of patience. After the reframe, it is a question about four small integers and whether they are even or odd — checkable, quite literally, on one hand. Nothing about the city changed. What changed was which of the city's features Euler agreed to see as part of the problem, and which he agreed to throw away as noise. That act of selective forgetting is not sloppiness. It is a proof, embedded in the argument itself, that the discarded detail could not possibly have affected the answer — which is exactly why the resulting two-line argument is not a cheat but a genuine solution, arguably a more complete one than exhaustively checking every route ever could be, since checking finitely many routes never rules out one you didn't think to try, while Euler's parity argument rules out all of them at once, forever, by construction.

What a Representation Actually Is

Chapter 00 introduced the idea, borrowed from Allen Newell and Herbert Simon's 1972 formalization of problem spaces, that any problem can be described by a set of states, a set of legal moves between states, a starting state, and a goal. What Euler's bridges make vivid is that this description is never handed to you by the world. Someone — often you, often unconsciously — has to decide what counts as a "state" and what counts as a "move," and that decision is where nearly all of a problem's real difficulty actually lives.

A representation, in this technical sense, is a choice about three things: which differences between situations you will bother to track as different states; which transformations between states you will permit yourself to call legal moves; and what you will count as the goal — the target configuration that ends the search. Before Euler, the implicit representation of the bridge puzzle distinguished states by exact physical location: which specific street, which specific bridge underfoot, facing which direction. Under that representation the problem space is enormous, because two routes that differ only in some irrelevant meandering through a courtyard count as different states requiring separate consideration. Euler's representation instead distinguished states only by which bridges remain uncrossed and which landmass the walker currently occupies — a description that treats every meandering path within a single landmass as the same state, because for the question being asked, they are the same. He did not simplify the problem by ignoring difficulty. He simplified it by proving, correctly, that a huge amount of what looked like difficulty was never actually relevant to the answer.

This is the discipline that separates a good reframe from mere hand-waving, and it is worth stating plainly because it will recur, in different clothes, in every chapter that follows. Throwing away information you shouldn't throw away does not shrink a problem — it breaks it, producing a confident wrong answer or no answer at all. Keeping information you don't need does not make a solution more careful — it just keeps the search space needlessly large. The right representation is the one that keeps exactly the distinctions the goal actually depends on, and collapses every distinction that doesn't change the answer into a single state. Finding that boundary — what to keep, what to discard — is not a mechanical step you can outsource to a procedure the way Ahmes' scribes outsourced arithmetic to false position. It is closer to a judgment call, made correctly by Euler here and, as chapters to come will show, made incorrectly by very smart people a great many times before someone else found the right cut.

The Schoolroom That Keeps Changing Shape

Carl Friedrich Gauss is supposed to have performed the same kind of move, at a vastly smaller scale, as a schoolboy in Braunschweig sometime in the 1780s. A teacher — bored, or wanting quiet, depending on which telling you read — set the class the tedious busywork of adding up every whole number from 1 to 100. Young Gauss, the story goes, produced the correct answer, 5,050, almost instantly, while his classmates were still laboriously carrying sums. His method, as usually told: pair the numbers from opposite ends of the list — 1 with 100, 2 with 99, 3 with 98 — and notice that every pair sums to the same value, 101. There are fifty such pairs. Fifty times 101 is 5,050. What began as one hundred sequential additions collapses into a single multiplication.

That is a genuine and useful reframe, worth walking through carefully because it is structurally identical to Euler's, at a much smaller scale. The naive representation of "sum the numbers 1 to 100" treats the problem as a sequence: a running total, updated by one small addition at a time, a hundred steps from start to finish — a long, thin, unavoidably serial problem space. Gauss's representation instead treats the same hundred numbers as fifty unordered pairs, exploiting the fact that addition does not care what order its inputs arrive in. Reorganizing the numbers around their pair-sums, rather than around their sequence, throws away information (which number came "first") that the goal — the total — never needed in the first place, and replaces a hundred-step walk with a single leap.

But — and this is where the chapter's stated house rule about disputed anecdotes has to be honored rather than smoothed over — there is real reason to doubt that this specific, vivid version of the story is an accurate record of an actual 1780s classroom. The science writer Brian Hayes went looking for the story's origin for a 2006 American Scientist column, "Gauss's Day of Reckoning," and traced its earliest known appearance to an 1856 memorial biography of Gauss written after his death by Wolfgang Sartorius von Waltershausen — more than seventy years after the event it describes, based on Sartorius's personal recollections of conversations with an elderly Gauss, not on any contemporary record from the schoolroom itself. Hayes then tracked the story through dozens of later retellings and found the details drifting from version to version: the exact numbers being summed change, the teacher's name and motive change, and the method Gauss is credited with using is not always the pairing trick — some versions have him using a different, though related, arithmetic shortcut entirely (Hayes, "Gauss's Day of Reckoning," American Scientist, 2006). The honest conclusion is not "the story is false" — something like it plausibly happened, and Gauss undeniably possessed exactly this kind of number sense from childhood — but that the crisp, cinematic version most people repeat has been shaped by a hundred and seventy years of retelling into something cleaner than the historical record can actually support. As chapter 00 already flagged in passing, the same caution that applies to Alexander at Gordium applies here: a story can be pedagogically true and biographically unverifiable at the same time, and a book about clear thinking should say so rather than pretend otherwise.

What survives the uncertainty is the mathematics, and the mathematics does not need the schoolboy to be real to make its point. Whether or not a nine-year-old actually stunned a Braunschweig classroom in the 1780s, the pairing trick is a clean, small-scale demonstration of exactly the move this chapter is about: a problem that looks sequential and slow under one representation (running sum, one hundred additions) becomes a single fast operation under another (fifty constant-sum pairs, one multiplication), and nothing about the underlying arithmetic changed to make that possible — only the arrangement in which it was viewed.

Descartes' Bridge

Chapter 01 already introduced René Descartes as a man chasing a universal method for all discovery. One specific technical result from that chase deserves a closer look here, because it may be the single most productive representation change in the history of mathematics: the coordinate system he laid out in the 1637 appendix La Géométrie.

Before Descartes, geometry and algebra were, for most practical purposes, two separate toolkits addressing two separate kinds of question. Geometry, in the Euclidean tradition, worked with figures — circles, lines, triangles — manipulated by physical-feeling constructions with compass and straightedge, and proved things about them through chains of logical argument grounded in diagrams. Algebra worked with symbols and equations, manipulated by symbolic rules, largely detached from any picture. A circle, geometrically, was "the set of all points a fixed distance from a center" — a definition you could draw but not easily calculate with. Descartes' innovation was to fix a pair of perpendicular reference lines — what we now call axes — and describe every point in the plane by its distance along each (La Géométrie, Britannica). Once every point is a pair of numbers, every curve through those points becomes an equation relating the numbers, and vice versa: that same circle is now simply every pair (x, y) satisfying x² + y² = r². Finding where two curves intersect — a question that used to require an inventive geometric construction, different for every pair of shapes — becomes a single, mechanical act of solving two equations simultaneously, the same technique regardless of whether the curves in question are circles, parabolas, or shapes nobody had ever successfully drawn together before.

This is a representation change in exactly Euler's sense, just aimed at an entire discipline rather than one puzzle. The "states" of a geometric problem — points, curves, intersections — get redescribed as algebraic objects, and the "operators" — construct this line, bisect that angle — get redescribed as algebraic manipulations: substitute, factor, solve. Problems that had resisted geometric construction for centuries — including some, like general angle trisection, that Greek geometers' tools genuinely could not solve, a fact only provable using this same algebraic machinery two centuries later — suddenly had a common, systematic language in which to be attacked. Two disciplines that had developed almost independently for two thousand years were, after one reframe, doing the same arithmetic underneath different vocabularies. Chapters 04 through 06 rely on exactly this fusion whenever they reach for calculus or linear algebra to solve problems that started out looking purely geometric or purely combinatorial; none of that machinery has anywhere to stand without Descartes' coordinates underneath it.

The Man Who Turned His Back on the Bar

Not every reframe in this book's history comes from a mathematician with a pen. On the evening of October 20, 1968, in Mexico City's Estadio Olímpico Universitario, a twenty-one-year-old American named Dick Fosbury won Olympic gold in the high jump by doing something no serious competitor had ever done at that level: he went over the bar backward, head and shoulders first, arching his spine so his back faced the ground below.

Fosbury had not started out trying to reinvent his sport. As a high schooler in Medford, Oregon, following the loss of his younger brother Greg to a drunk-driving accident, he took up track partly to give himself something to hold onto, and struggled badly at the high jump using the standard technique of the era, the "straddle," in which a jumper clears the bar face-down, rotating one leg over first. Using the straddle, Fosbury routinely failed to clear even five feet — a height that barely qualified him for competition (Smithsonian, "The Fosbury Flop"). In the spring of 1963, mid-meet and mid-failure, he changed his body position instinctively rather than deliberately: instead of rotating face-down over the bar, he arched backward over it, landing on his neck and shoulders. It looked, to at least one local newspaper photo caption the following year, like "a fish flopping in a boat" — and the name stuck (Smithsonian, "The Fosbury Flop").

The reframe here is worth stating precisely, because it is easy to describe it too loosely as simply "trying something new." The straddle technique accepts a goal implicitly stated as get your whole body, moving forward and rotating, cleanly past a raised bar — and treats every improvement as a search within that fixed frame: more speed on the approach, more leg flexibility, a faster rotation. Fosbury's flop instead exploits a genuine fact of physics that the straddle leaves on the table: a jumper's center of mass does not have to clear the bar at the same height as the rest of the body. By arching deeply backward, with the head, arms, and legs all hanging below the arched spine at the peak of the jump, a flopping jumper's center of mass can pass underneath the bar even while every part of the body technically passes over it — meaning a jumper can, in principle, clear a given bar height with less total energy than a technique whose center of mass has to rise above the bar directly (Stanford, "The Physics of the Fosbury Flop"). This is not a change in effort or talent. It is a change in what "clearing the bar" was allowed to mean — a redefinition of the legal states of the problem that the straddle's own frame had never questioned.

Crucially, this reframe was only survivable because something in the environment had changed first, independent of any high jumper's cleverness. Traditional high-jump landing pits, through most of the twentieth century, were shallow beds of sand or wood shavings — fine for a straddle jumper landing feet or side-first, lethal for someone landing on their head and neck at speed. It was only the spread of deep foam landing mats through American university and high-school track programs during the 1960s — precursors to the "Port-A-Pit" design later commercialized by the inventor Donald Wallace Gordon — that made a backward, head-first landing survivable at all; Fosbury's own high school was among the earlier adopters of such matting, after his coaches, alarmed at how he was landing, arranged for the school to receive foam retired from a local university's upgraded pits (Smithsonian, "The Fosbury Flop"). Change the representation of the jump without changing the landing surface, and the same idea is simply a fast way to break your neck. This is a detail worth sitting with: sometimes the operators available to you — the legal moves your environment permits — are not fixed facts of nature but recent, contingent, and changeable, and part of finding a shorter path is asking whether an operator you have been treating as fixed is actually just a habit someone else could unfix.

Fosbury struggled even after his 1963 breakthrough — his own college coach at Oregon State initially pushed him back toward the straddle, worried the flop could not be coached to consistency, and Fosbury nearly lost his scholarship over inconsistent results before a breakthrough 1968 season culminating in a Pac-8 title, an NCAA championship, and finally Mexico City, where he cleared 2.24 meters for Olympic gold and an Olympic record, using a technique he was still, at the time, the only competitor in the field employing (World Athletics, "50 years since the day Dick Fosbury revolutionised the high jump"). He never jumped higher than that night. He didn't need to, for the argument this book is making: within four years, at the 1972 Munich Olympics, 28 of the 40 competitors in the men's high jump final were using some version of his technique; by 1980, the straddle had essentially vanished from elite competition (World Athletics; Smithsonian, "The Fosbury Flop"). A field of specialists spent decades searching harder within one representation of "how to jump." One outsider changed the representation, once, and the entire field's search space was permanently reshaped around his answer.

Why Reframes Feel Like Cheating

There is a common thread running underneath Euler's letters, the Gauss legend, Descartes' coordinates, and Fosbury's back-first leap, and it is worth naming directly because it will recur for the rest of this book: a genuine reframe very often does not feel, to the people watching it or even to the person performing it, like "real" work in the domain being reframed. Euler said so explicitly, in writing, about his own solution — it "bears little relationship to mathematics." Fosbury's own world-record-holding rival, Valery Brumel, was quoted at the time calling the flop "an aberration," not a technical advance. A reframe often looks less like mastering a domain's existing difficulty and more like refusing to engage with it on its own terms — which is, in a sense, exactly what it is. The difficulty a domain's practitioners have spent years learning to fight is precisely the difficulty that a good reframe makes irrelevant, and there is no reason the practitioners who mastered fighting it should be the first to recognize that the fight was optional.

This is why the theoretical apparatus matters and is not mere decoration. A representation is a specific, nameable thing — a choice of states, a choice of operators, a choice of goal — and asking "what representation am I using right now, and did I choose it or did I inherit it?" is a concrete, repeatable question, not a vague call to "think outside the box." Euler's states were landmasses and bridge-crossings, not streets. Gauss's states were pairs, not a sequence. Descartes' states were coordinate pairs, not geometric figures. Fosbury's states were centers of mass, not bodies treated as rigid wholes facing a fixed direction. In every case, the difficulty of the original problem turns out to have been at least partly an artifact of an unexamined representation — and in every case, once the new representation is stated, the remaining work is comparatively short. That gap — vast search before the reframe, short walk after it — is the shortest path this book's title promises, and chapter 03 turns to the deliberate, teachable techniques several traditions have since built for finding that reframe on purpose, rather than waiting for a Euler, a Gauss, or a Fosbury to stumble into it.

The Move

Move 02 — Change the representation until the difficulty dissolves.

• Ask what you are allowed to throw away. Euler discarded every street in Königsberg and kept only which bridge touched which landmass. Before searching harder inside a hard representation, ask which of its details the goal actually depends on — and prove, don't just assume, that the rest is safe to discard.
• Ask what you are needlessly holding apart. Gauss's pairing trick works because addition doesn't care about order; the "sequence" was a habit of representation, not a requirement of the arithmetic. Look for structure your current framing is hiding by insisting on an arrangement — order, direction, sequence — the goal never actually required.
• Check whether the environment's rules are fixed or just unquestioned. Fosbury's reframe of the jump was only viable once foam mats replaced sand pits — an external operator, not a mental one, had to change first. When a reframe seems physically or practically blocked, ask whether the blocking constraint is a law of nature or a fixable habit of your environment.
• Expect it to feel illegitimate. Euler called his own breakthrough barely mathematics. If your reframe doesn't provoke at least a flicker of "that's not really solving the problem" from someone who knows the domain well, it may not be cutting deep enough yet.

Next — Chapter 03 · The Heuristic Age: what happens once solvers stop waiting for a reframe to arrive by accident and start building deliberate, teachable methods for finding one on purpose.