An OpenAI model has disproved a central conjecture in discrete geometry Explains the problem

 

What problem did the AI solve?

Mathematicians have been thinking about a geometry puzzle for almost 80 years.

The question was:

If you place n points on a flat surface, how many pairs of points can be exactly 1 unit apart?

This is called the Unit Distance Problem, created by Paul Erdős in 1946.

Easy example

Imagine points on graph paper.

A square grid creates many pairs that are exactly distance 1 apart:

• • • •

• • • •

• • • •

For decades, mathematicians believed this square-grid style was basically the best possible arrangement.

That belief became a famous conjecture (an unproven mathematical belief).

What did the OpenAI model do?

The OpenAI reasoning model found a completely different construction that creates even more unit-distance pairs than mathematicians thought possible.

So the AI didn’t just “solve” the conjecture.

It actually proved the conjecture was wrong.

That’s why the headline says:

“disproved a central conjecture”

Why is this important?

Because:

• Thousands of mathematicians had thought about this problem.
• The conjecture survived for decades.
• The AI found a counterexample humans missed.

According to mathematicians quoted in the article, this is considered a genuine research breakthrough.

What math trick did the AI use?

This is the wild part.

The problem looks like simple geometry…

…but the AI used advanced ideas from:

• algebraic number theory
• class field towers
• Golod–Shafarevich theory

These are extremely deep math areas usually unrelated to simple point geometry.

So the AI connected two distant branches of mathematics in a creative way.

That’s why researchers are impressed.

What does “$n^{1+\delta}$” mean?

The old belief was roughly:

$u(n) \leq n^{1+o(1)}$

Meaning the number of unit-distance pairs grows only a little faster than linear.

The AI showed configurations where:

$u(n) \geq n^{1+\delta}$

for some fixed positive number $\delta$.

That means the growth is genuinely polynomially larger than expected.

Why people are calling this historic

Researchers say this may be:

• the first time an AI independently solved a major open math problem,
• using original reasoning,
• not just copying known proofs.

Famous mathematicians like Tim Gowers and Noga Alon praised the result publicly.

One quote from Tim Gowers basically said:

If a human submitted this proof to a top math journal, he’d recommend accepting it immediately.

Why this matters beyond mathematics

People are excited because this suggests future AI systems may help with:

• physics research
• medicine
• chemistry
• engineering
• discovering new scientific ideas

Not just answering questions.

The key thing is:

the AI wasn’t only calculating — it appears to have shown creative mathematical reasoning.