AI Helps Crack Cryptography Puzzle in Just Two Weeks

A graduate student in Massachusetts and two professors in California just proved that artificial intelligence can supercharge scientific discovery in ways we've never seen before.

Last Thursday, MIT student Seyoon Ragavan sent his friend Yao-Ting Lin an exciting message about solving a tricky quantum cryptography problem. Lin couldn't respond right away because he was in a meeting where his professor was announcing they'd solved the exact same problem. "We are definitely living in strange times," Lin wrote back.

Both teams had used OpenAI's GPT-5.6 Sol Ultra to crack a puzzle that had stumped researchers for years. They submitted their papers to the same scientific website just over three hours apart, neither knowing the other was so close behind.

The problem they solved deals with unclonable encryption, a way to use quantum physics to protect messages so they can't be copied. It's specialized work, but it represents a real leap forward in how we might secure information in the future.

What makes this moment special isn't just the solution. It's how quickly it happened. Ragavan had tried to solve this problem years ago without success. This time, with AI assistance, he cracked it in just over two weeks.

Ragavan worked hands-on with the AI, guiding it through two-hour sessions and redirecting it when needed. The California team took a different approach, using a custom system that let the AI propose and critique its own solutions. Same AI model, different methods, identical breakthrough.

Why This Inspires

This story shows us a future where stubborn problems that have blocked progress for years might finally get solved. Professor Prabhanjan Ananth says researchers now have a new first step when they hear about an unsolved problem: see if AI can crack it.

The overlap raised important questions about how we credit scientific discoveries. When everyone has access to the same powerful tools, do simultaneous breakthroughs become more common? The answer seems to be yes, and these researchers are handling it with grace. They're now working together to combine their papers into one submission.

This isn't AI replacing human scientists. It's AI helping them think faster and explore paths they might have missed. The researchers still had to verify the proofs, refine the ideas, and understand what the AI had discovered.

Other old math problems have already fallen to AI assistance this year, including puzzles that went unsolved for 50 and even 80 years. Theoretical computer scientists are just beginning to explore what's possible when they team up with these tools.

The timeline really is remarkable. From hearing the problem posed at a talk to submitting proven solutions took just two weeks and a day. Problems that stumped brilliant minds for years are now yielding to human creativity amplified by artificial intelligence.

These researchers turned a potential conflict into collaboration, showing us that even as technology changes how we make discoveries, human connection and generosity still matter most.