Part One of a two-part reflection. Here I tell the story of Fermat’s Last Theorem, the schoolchild-simple equation that took humanity three and a half centuries to prove, and I end by opening a stranger question about what it even means to solve something. Part Two, The Witness Paradox, follows that question wherever it leads.

An equation a child can read

Write down aⁿ + bⁿ = cⁿ. A ten-year-old can read it. It has three letters, one exponent, a plus sign, and an equals sign. There is nothing hidden in the notation, no exotic symbol, no specialized vocabulary. And yet this small string of characters generated a mystery that outlasted empires, defeated the greatest minds of every generation for over three hundred years, and was only laid to rest in the final decade of the twentieth century.

That gap between how easy the question is to ask and how brutally hard it was to answer is the whole reason the equation is famous. It is one of the purest examples in all of human thought of a problem that looks like it should take an afternoon and instead takes an age.

First, the part that works

Before the impossibility, there is a version that behaves beautifully.

When the exponent is 2, the equation becomes a² + b² = c², the Pythagorean theorem, the relationship between the sides of a right triangle. Here, solutions are everywhere. The famous one is 3, 4, 5: nine plus sixteen equals twenty-five. Then 5, 12, 13. Then 8, 15, 17. There are infinitely many, and mathematicians have known how to manufacture them on demand for over two thousand years. Pick any two positive whole numbers, call them m and k with m larger than k, and compute:

  • one side as m² − k²
  • another as 2mk
  • the longest side as m² + k²

Feed in m = 2 and k = 1 and out pops 3, 4, 5. Feed in m = 3 and k = 2 and out pops 5, 12, 13. The machine never runs dry. For n = 2, the equation is not just solvable. It is generous.

So the natural question, the one any curious person would ask next, is: what about cubes? What about a³ + b³ = c³? What about fourth powers, fifth powers, higher? If squares give us endless triangles of whole numbers, surely cubes give us something.

They give us nothing. And that nothing is the entire story.

The margin that started it all

Around 1637, a French lawyer and amateur mathematician named Pierre de Fermat was reading a copy of an ancient Greek text on arithmetic. Beside the section dealing with Pythagorean triples, he jotted a note in the margin. He claimed that for any exponent greater than 2, the equation has no solutions in positive whole numbers at all, and then added the most tantalizing sentence in the history of mathematics: he had discovered a truly marvelous proof of this, which the margin was too narrow to contain.

Then he died without ever writing it down.

We do not know whether Fermat actually had a proof. Most mathematicians today strongly doubt it. The tools that eventually cracked the problem did not exist in his century and would not exist for another three hundred years, so if he had a genuine general proof, it would have to have been something completely different from anything anyone has since found, and nobody has ever reconstructed it. The likeliest explanation is that Fermat had a proof for a specific small case, convinced himself the same idea would generalize, and was simply mistaken. He was a brilliant mathematician who was occasionally wrong, like everyone.

But the note survived, published after his death by his son, and it turned into a challenge that would haunt the discipline.

Three centuries of siege

Here is the strange thing about the claim: it is easy to test and impossible to finish.

Anyone can check that there are no small cubes that work. You can check thousands of cases, millions, and today billions with a computer, and you never find a counterexample. But checking cases, no matter how many, proves nothing about the infinite ocean of cases you did not check. Mathematics does not accept “we looked really hard and didn’t find one.” It demands a proof that covers all numbers at once, forever. That is the wall every generation slammed into.

The greatest names in mathematics chipped away at pieces of it. The case of cubes was settled in the 1700s. Fourth powers had actually been handled by Fermat himself with a method he really did leave behind. One by one, individual exponents fell: a proof for n = 5, a proof for n = 7, whole families of exponents dispatched by clever arguments in the 1800s. There was even a mathematical crisis when a celebrated attempted proof turned out to rest on a hidden false assumption about how numbers factor, a mistake so instructive that repairing it created entire new branches of mathematics.

But every one of these victories was partial. Proving the theorem for the exponent 5, or 7, or a thousand specific exponents, still leaves infinitely many exponents untouched. After three hundred years of effort by history’s finest, the general statement remained exactly what Fermat had left it: an assertion nobody could actually prove.

It became, in a sense, the most famous unsolved problem in mathematics. Not because it was important to any application, but because it was so simple to state and so humiliatingly resistant to being finished.

The unlikely path to the answer

The resolution, when it finally came, arrived from a completely unexpected direction. It did not come from someone cleverly manipulating the equation itself. It came from a deep and initially unrelated conjecture about two very different kinds of mathematical objects, curves defined by cubic equations on one side, and richly symmetric functions on the other, which a suggestion in the mid-twentieth century proposed were secretly, profoundly connected.

In the 1980s, someone noticed the bombshell: if Fermat’s equation had a solution, that solution could be used to build one of these curves with impossible properties, a curve so strange it could not possibly obey the deep connection the conjecture predicted. In other words, if this grand unrelated conjecture were true, then Fermat’s Last Theorem had to be true as well. The centuries-old riddle had been quietly bolted onto one of the central questions of modern mathematics.

Andrew Wiles, a mathematician who had been captivated by the problem as a ten-year-old reading about it in a library book, saw his opening. He worked in near-total secrecy for seven years, an almost unheard-of thing in a collaborative field, telling almost no one what he was really doing. In 1993 he announced a proof to a stunned audience. Then, during the review process, a subtle gap appeared in the argument, and for over a year it looked as though the dream had collapsed at the finish line. Working with a former student, Richard Taylor, Wiles finally repaired the gap with a clever combination of approaches, and in 1994 the proof was complete and correct.

Three hundred and fifty-seven years after a lawyer scribbled in a margin, the theorem was a theorem for real.

So did anyone “solve” it?

This is where language needs care, because two very different things get called “solving,” and confusing them is the most common mistake people make about this problem.

Nobody ever solved the equation in the sense of finding numbers that satisfy it for a power of 3 or higher, because no such numbers exist. There is nothing to find. That was never the goal and never possible.

But the theorem, the statement that no such numbers exist, was absolutely solved, in the sense of being proven, by Wiles in 1994. For three and a half centuries the honest thing to say was “nobody has been able to prove this yet.” After 1994, that sentence became false. The problem is closed. It is one of the great settled results of mathematics.

So the poetic phrase “the equation nobody could solve” is a story about the past. Today the correct statement is that it took humanity an astonishingly long time, and mathematics of breathtaking depth, to prove something a child could understand the question to.

Why it still fascinates

The lasting curiosity of aⁿ + bⁿ = cⁿ is not that it is hard. Plenty of things are hard. It is the specific, almost cruel mismatch between the surface and the interior.

On the surface it is arithmetic, the kind of thing that lives in a middle-school classroom, one small step past the Pythagorean theorem everyone already knows. Underneath, the only known path to its truth runs through some of the most abstract and demanding machinery ever built by mathematicians, ideas that took generations to invent for entirely different reasons.

There is also the human drama that no other equation can match: the maddening marginal note and its lost “marvelous proof,” the parade of geniuses who threw themselves at it and won only fragments, the crisis of a false proof that accidentally birthed new mathematics, and finally a boy who fell in love with the problem, disappeared for seven years, stumbled at the last moment, and got back up.

And there is a quieter lesson underneath all of it. The proof that no solutions exist did not come from staring harder at the equation. It came from discovering that this humble little question was secretly tied to vast, distant regions of mathematics that no one had connected to it before. The answer was hiding not inside the problem but in the relationships between problems. That is often how the deepest questions get resolved, not by force, but by revelation of a link nobody expected.

The equation is simple. The impossibility it describes is absolute. And the road to proving that impossibility is one of the most beautiful journeys the human mind has ever taken. That is why, long after it stopped being a mystery, it remains one of the great stories.

A closing paradox: even solved problems need a witness

It is tempting, in an age of powerful machines, to imagine that hard problems will now simply fall, that we could point a tireless artificial intelligence at every open question and wait for the answers to pour out. And to some degree tools do accelerate discovery. But the story of Fermat quietly warns against believing the tool is the whole of it.

Consider what the breakthrough actually required. Not raw computation, because no amount of checking cases would ever have finished the job, and everyone knew it. What was required was a human deciding the problem mattered enough to spend a life on, recognizing an unexpected bridge between distant fields, and having the taste to know which of a thousand possible directions was worth walking. The willing and the creativity were the scarce ingredients. The machinery, however impressive, does not supply the wanting.

And here a genuine paradox appears. Suppose you tried to remove the human entirely: you set a system running forever to solve every problem in the world, sealed off, with no human access at all. Ask the simple question, solved for whom? A proof no one ever reads is not performing the function a proof exists to perform. Mathematics was never merely the production of true statements. It was the act of understanding, of convincing, of folding a result back into what people can now see and build upon. A correct answer sitting in a sealed box that no one opens has the same practical standing as no answer at all.

The trap tightens when you notice how you would even learn that the machine had succeeded. To know it solved anything, you need access to its results, and the instant you have that access, the premise of the sealed box is broken. The two conditions, knowing that problems are being solved and no human ever having access, cannot both hold. The scenario quietly eats itself.

This reveals something easy to miss: solving is not only an event out in the world. It is also an event that has to happen inside a mind. The answer has to land somewhere, be understood by someone, change what someone can now do. The human is not merely the one who starts the search. They are also the one who completes it by understanding what was found. Remove the human from both ends and you do not get a universe brimming with solved problems. You get a very expensive silence.

Fermat’s theorem took three and a half centuries precisely because it was waiting for a mind ready to understand it. That, in the end, is what a solution is: not a string of symbols that happens to be true, but a piece of the world that has finally been understood by someone capable of caring that it was.

Continue reading: Part Two, The Witness Paradox.

Author’s note: this text is still awaiting a proper revision from me. I’m publishing it in this form because reading it on screen makes the revising easier. Spot something off? Tell me.