Robot Brain Breaks an 87-Year-Old Math Puzzle Nobody Could Solve!
Quick Summary
A mathematician at Harvard University used an AI model to disprove a famous math problem that had stumped experts for 87 years. The AI found the answer in a solution so short it could fit in a single text message. Scientists around the world are amazed.
What Happened?
Imagine a puzzle box that nobody has been able to open for nearly 90 years. Thousands of brilliant minds have tried. They all failed. Then one day, a computer program cracks it open in seconds.
That is almost exactly what happened with something called the Jacobian conjecture.
A conjecture is an idea in mathematics that people believe is probably true — but nobody has been able to prove or disprove it yet. Think of it like a rumor that everyone thinks is true but nobody has found proof for.
The Jacobian conjecture was first written down by a German mathematician named Ott-Heinrich Keller in 1939. It was so famous and so difficult that a top mathematician added it to his list of the biggest unsolved math problems of the 21st century.
For 87 years, some of the smartest people on Earth tried to crack it. Some claimed they had — but each time, other experts found tiny mistakes in their work.
Then, on July 19, 2026, Harvard mathematician Levent Alpöge sat down with an AI model called Fable 5, made by a company named Anthropic. He asked it to help him think through the problem. Together, they found something incredible: a counterexample.
A counterexample is like catching a rule breaking its own rule. Imagine if someone said, “All swans are white.” If you find just one black swan, the rule is busted. Alpöge and Fable 5 found a “black swan” for the Jacobian conjecture — proving it was actually false.
The solution was just 216 characters long. That is shorter than this paragraph! Alpöge posted about it on social media. He wrote, simply, that the conjecture was false.
Other mathematicians around the world checked the work carefully. They confirmed: it was correct.
Why Does It Matter?
This is big news for two reasons.
First, solving — or in this case, disproving — a famous old math problem is extremely rare. It’s like finding a map that rewrites our understanding of a whole country.
Second, it shows just how powerful AI is becoming at helping scientists think through hard problems. AI is not replacing mathematicians. Instead, it is working alongside them like a super-fast thinking partner.
Many mathematicians say this is the hardest problem an AI has ever helped crack. And experts say the pace of these AI-powered math breakthroughs is speeding up fast.
Big Words
- Conjecture — An idea in math or science that people believe might be true, but that nobody has proven or disproven yet.
- Counterexample — A single example that shows a rule or idea is NOT always true, instantly proving the rule is wrong.
- AI model — A computer program trained on huge amounts of information that can help people think through difficult problems very quickly.
- Disprove — To show that something believed to be true is actually wrong.
- Mathematician — A scientist who studies numbers, patterns, and logical rules.
Fun Fact
The Jacobian conjecture was so famous that a top mathematician included it in his year-2000 list of the greatest unsolved problems for the entire 21st century — and it got solved (well, disproved!) in less than 30 years!
Think About It
If you could use an AI to help you solve any mystery or puzzle in the world, what would you pick — and why?
Sources
- Smithsonian Magazine — ‘A.I. Disproves a Decades-Old Mathematical Idea, the Biggest Conjecture That the Tech Has Played a Role in Yet’, July 27, 2026
- The Next Web — ‘AI just disproved the 87-year-old Jacobian conjecture’, July 21, 2026
- The Week — ‘The Jacobian conjecture disproved by AI, explained’, July 29, 2026
- The Conversation — ‘hello there the jacobian conjecture is false thanx: why a tiny social media post has mathematicians rethinking AI’