AI Cracks 80-Year Math Puzzle Using Everyday Tech

A general-purpose AI just outsmarted an 80-year-old math puzzle, and the approach it used could change how we all tackle tough problems.

OpenAI's AI model disproved a long-standing conjecture about the "planar unit distance problem," a deceptively simple geometry question first posed by legendary mathematician Paul Erdős in 1946. The question asks: if you place points on a flat surface, how many pairs can be exactly one unit of distance apart?

For eight decades, mathematicians believed the traditional "square grid" arrangement was the best answer. The AI proved them wrong by discovering an infinite family of examples that work even better.

Here's what makes this breakthrough truly exciting: the AI wasn't custom-built for mathematics. It was a general-purpose reasoning model, the kind that could theoretically help anyone solve problems in their own field.

Think of it like discovering that a regular hammer can fix complex car problems instead of needing an expensive specialized tool. Suddenly, the solution becomes accessible to everyone, not just experts with rare equipment.

The AI succeeded by taking a different approach than most people expect. Instead of trying to prove the existing theory was correct, it searched for a counterexample to show the theory was wrong.

A team of external mathematicians has verified the proof, confirming the AI's discovery is legitimate. This marks a genuine advance in combinatorial geometry, a field that studies how objects can be arranged in space.

Why This Inspires

This breakthrough shows us something valuable beyond mathematics. When facing a stubborn problem, sometimes the smartest move isn't pushing forward harder, it's looking for what breaks the current assumption.

AI expert Dr. Lance Eliot points out that most people missed this lesson. Instead of always having AI "plow forward into a morass," we should consider using it to find those crucial counterexamples that reveal when we're on the wrong track entirely.

The fact that everyday AI accomplished this suggests specialized tools aren't always necessary for breakthrough thinking. Regular technology, applied thoughtfully, can crack problems that have stumped specialists for generations.

This discovery opens doors for researchers across fields who might have assumed they needed expensive, customized AI systems. Sometimes the tools we already have are more powerful than we realize.

The math community initially felt a twinge of disappointment, hoping AI had generated a complete new proof rather than disproving an old conjecture. But finding what doesn't work is often just as valuable as finding what does, because it clears the path forward.

Mathematics just got a little more solved, and problem-solving everywhere just got a little more accessible.