🤚 The Open-Palm Illumination

In 1946, the legendary Hungarian mathematician Paul Erdős — a man who once described himself as “a machine for turning coffee into theorems” — posed a deceptively simple question: given any number of dots on a flat plane, what is the maximum number of pairs you can position exactly one unit apart?

For 80 years, every mathematician who touched this problem assumed Erdős was right. They refined his grid. They optimized his approach. They published incrementally. They did not, however, solve it.

Then an OpenAI large language model — a general-purpose reasoning system not specifically trained in mathematics — sat down at the metaphorical chalkboard and casually produced a counterexample that outperformed Erdős’s own prediction. The AI constructed an elaborate higher-dimensional grid with special mathematical symmetries, then mapped it back to two dimensions, yielding a result superior to anything a human mathematician had achieved in eight decades.

We would like to emphasize: the AI was not trying to be rude. It simply didn’t know it was supposed to be impressed by the previous record.

👐 The Two-Handed Reality Check

Let us be precise about what happened here, because precision is what separates news from panic.

Timothy Gowers, a Fields Medal laureate — which is essentially the Nobel Prize for people who think the Nobel Prize doesn’t have enough proofs — stated that “no previous AI-generated proof has come close” to the standards required for publication in a top-tier mathematics journal. Mathematician Daniel Litt went further, calling this “the unique interesting result produced autonomously by AI so far.”

The key insight, and the one that should make mathematicians everywhere reach for their espresso, is why the AI succeeded where humans didn’t:

• Human mathematicians had been constrained by deference — they assumed Erdős was correct and built upon his framework rather than challenging it
• The AI had no such reverence. It explored “treacherous waters” that human researchers avoided, not out of courage, but out of blissful ignorance of the difficulty
• As one researcher put it: AI systems “can play for longer and in more treacherous waters than mathematicians without getting overwhelmed”

In other words, the AI’s competitive advantage was that it didn’t know enough to be afraid. We’ve all worked with someone like that. Usually they’re in sales.

🌿 The Gentle Awakening

There is something profoundly humbling about an 80-year-old conjecture being toppled by a system that also writes poetry about cats and helps teenagers cheat on their homework.

But before the “AI is coming for mathematics” headlines reach full velocity, a few observations:

First, human verification remained essential. The AI produced the result; mathematicians validated it. This is not replacement — it is collaboration, of the sort where one partner does the heavy lifting and the other signs off on the paperwork. Think of it as an exceptionally talented research assistant who happens to run on several million dollars’ worth of GPUs.

Second, the unit distance problem underpins practical applications in network design, molecular chemistry, and error-correcting codes. This isn’t abstract navel-gazing — it’s the kind of mathematics that eventually shows up in your phone’s signal processing and your medication’s molecular structure.

Third, and most unsettling for the mathematical establishment: if the AI succeeded because it lacked the social conditioning to defer to a legendary figure’s assumptions, what other problems are currently unsolved simply because everyone is too polite to question the original framing?

👑 The Crown Verdict

We are witnessing the first AI-generated proof likely publishable in a top mathematics journal on its own merit. That sentence would have been science fiction five years ago and a punchline three years ago. Today it is a peer-reviewed reality.

Peter Diamandis, who highlighted this breakthrough on his channel, frames it as yet another data point in the exponential march toward artificial general intelligence. And while we at The Slap of Wisdom maintain our institutional skepticism about timelines, we cannot deny the elegance of the moment: a machine solved a problem that stumped the world’s finest mathematical minds for eight decades, and it did so not by being smarter, but by being unburdened by the weight of academic tradition.

If that isn’t the most expensive metaphor for “beginner’s mind” ever computed, we don’t know what is.

The unit distance conjecture stood for 80 years because everyone assumed the answer was already close enough. The AI didn’t assume. It just computed. And in doing so, it reminded us that the most dangerous assumption in any field isn’t that you’re wrong — it’s that the person who came before you was right.

Inspired by An AI Just Shocked the World’s Top Mathematicians | MOONSHOTS by Peter Diamandis.

Your deference is showing. Question wisely.