In June 1993, a Princeton mathematician gave three lectures in Cambridge under a title that sounded deliberately dull. At the very end of the last one, almost in passing, he announced that a puzzle from 1637 was finally solved. He had told almost nobody he was even working on it.
A library book and a ten-year-old’s promise
Andrew Wiles was about 10 years old when he stopped at his local library on the way home from school and picked up The Last Problem, a book by Eric Temple Bell.
It told the story of Pierre de Fermat’s famous margin note: no whole numbers can make a cube equal the sum of two cubes, and the same goes for every higher power. Fermat claimed a proof but never wrote it down.
The boy was struck by the contrast: a statement a child could understand, that nobody had ever proved. He decided he would be the first, then realised his knowledge was far too limited and shelved the dream.
The 1986 news that reopened the door
Wiles became a specialist in elliptic curves, objects that seemed to have nothing to do with Fermat. Then, in the summer of 1986, Ken Ribet proved a surprising link.
In plain terms: a solution to Fermat’s equation would create a curve so strange that it would break another famous prediction, the Taniyama-Shimura conjecture, which says every such curve hides a deep, regular structure. Prove that conjecture for the right kind of curves, and Fermat’s claim would be true automatically.
The catch was that almost everyone considered the conjecture out of reach. Wiles’s former supervisor John Coates said it seemed “impossible to actually prove.” Ribet later said Wiles was “probably one of the few people on earth who had the audacity to dream that you can actually go and prove [it].”
Six years, one confidant
Wiles was 33, and the problem had landed squarely in his own field. He dedicated all of his research time to it for more than six years, in near-total secrecy, releasing older work in small segments as separate papers to cover his tracks. The only person he confided in was his wife.
His June 1993 lectures at the Isaac Newton Institute were called “Modular Forms, Elliptic Curves and Galois Representations.” As Ribet told The New York Times, there was no hint in the title that Fermat would be discussed. Wiles proved his general result, then, “seemingly as an afterthought,” noted that it meant Fermat’s Last Theorem was true. The announcement drew wide press coverage.
The gap a referee found
Then came peer review. Nick Katz, one of the referees, sent Wiles a series of clarifying questions that led him to recognise a gap. One critical step, meant to put a limit on the size of a certain mathematical group, did not hold.
Without that piece, there was no proof. Wiles spent almost a year trying to repair it, first alone, then with his former student Richard Taylor. By the end of 1993, rumours were spreading that the proof had failed.
Twenty minutes of disbelief
On the morning of 19 September 1994, Wiles was on the verge of giving up. He was taking one last look at a method, only to understand why it would never work. Instead, he realised that the very method that kept failing held exactly what his own abandoned approach from three years earlier had been missing.
“It was so indescribably beautiful; it was so simple and so elegant,” he later said. “I couldn’t understand how I’d missed it and I just stared at it in disbelief for twenty minutes.” He called it “the most important moment of my working life.”
The corrected proof filled two papers, one co-written with Taylor, which took up the entire May 1995 issue of the Annals of Mathematics: 129 pages, more than seven years of research. John Conway called it “the proof of the [20th] century.”
A margin far too small
The proof relies on 20th-century tools that did not exist in Fermat’s lifetime, a strong hint that Fermat’s own argument, if it ever existed, was flawed. He never mentioned the general case again for the rest of his life.
Wiles was 41 when he fixed the proof, over the age limit of 40 for the Fields Medal, so the International Mathematical Union gave him a special silver plaque instead in 1998.
The story is still not over: in 2024, Kevin Buzzard and his collaborators received a five-year grant to translate the proof into a form a computer can check. A promise made by a ten-year-old in a public library is now being verified by machines.

