AI Solves 50-Year Math Problem After Being Told Not to Quit

A math problem that frustrated experts for 50 years just met its match in an AI that was simply told to keep trying.

OpenAI's GPT-5.6 Sol proved the "cycle double cover conjecture" last Friday, a breakthrough that shows how artificial intelligence is transforming mathematical research. The conjecture, first proposed in the 1970s, deals with graphs (networks of connected dots and lines) and whether they can be covered by loops in a specific way.

What makes this achievement remarkable isn't just that AI solved it. It's how the AI needed encouragement, just like a student tackling a tough homework problem.

The prompt that led to success included firm motivation: "Spend at least 8 hours on this before even thinking of returning or giving up." Engineers also told the AI not to dismiss the problem as too hard or assume that because humans failed, it would fail too.

"Most of that prompt is scaffolding aimed at getting the LLM to actually put in the effort needed to solve the problem," says Andrew Sutherland, a mathematician at MIT. The approach mixed affirming praise with stern directions, like a patient teacher coaching a struggling student.

The proof itself was surprisingly short and combined methods humans had tried before. It shows that almost any applicable graph can be doubly covered with no more than eight well-chosen loops.

Why This Inspires

This breakthrough reveals something unexpected about "hard" problems. Once a math puzzle gets a reputation for being difficult, experts might spend less time on it, creating a self-fulfilling prophecy of defeat.

"My guess is we will keep seeing examples of this—supposedly 'hard' problems having 'easy' solutions found by LLMs," Sutherland says. The AI didn't need revolutionary new ideas; it just needed persistence and the right encouragement to keep going.

Noga Alon, a mathematician at Princeton University, calls the achievement "yet another impressive example demonstrating that AI tools will change—and are already changing—mathematical research significantly."

The technical details matter for real-world applications too. Mathematical proofs about graphs help us understand all kinds of networks, from the internet to transportation systems.

OpenAI used 64 AI agents working together in parallel to solve the problem. Having multiple AIs communicate with each other helps prevent false proofs and made-up references, ensuring the solution was solid.

Sometimes the biggest breakthroughs come not from superhuman genius, but from refusing to give up when everyone else already has.