Two research teams reached the same result on an open quantum cryptography problem with help from the same AI model. Their papers were submitted to arXiv.org just three hours apart, according to Scientific American, turning a technical result into a broader signal about how research itself is changing.
The Same Problem, Two Paths
The work centered on "unclonable encryption," a method based on quantum properties. MIT PhD student Seyoon Ragavan worked on one solution. In a separate effort, professors Prabhanjan Ananth of UC Santa Barbara and Amit Sahai of UCLA worked on another.
Both teams used OpenAI's GPT-5.6 Sol Ultra. The source article says they took different approaches, even though they were pursuing the same open problem and relying on the same AI model. Their arXiv.org submissions arrived only three hours apart.
That timing matters because it makes the episode more than a story about one solved problem. It shows how AI tools can compress research timelines and bring separate groups to similar destinations almost simultaneously.
Why The Timing Stands Out
In science, independent discovery is not new. Researchers have often worked on the same questions at the same time, especially when a field is ready for a breakthrough. What is different here is the role of a shared AI system in helping both sides move toward a solution.
The source article frames the case around a larger question: when every researcher can use the same models, what still counts as an independent discovery? That question is especially sharp when the same model is available to multiple teams, pointed at the same open problem, and used in close succession.
The teams are now considering merging their papers. That possibility reflects the unusual nature of the overlap. The work was not identical in approach, but the convergence was close enough that collaboration may make more sense than treating the papers as fully separate milestones.
How Researchers Are Using AI Differently
The article describes a shift in research habits, not just a single technical achievement. Ananth describes a new first move when an open problem comes up:
If someone mentions an open problem, the first thing is to see if GPT solves it
That statement captures a practical change. Instead of beginning only with manual reasoning, literature review, or informal discussion, researchers can now test whether an AI model can help make progress immediately.
Ragavan describes the change in even more personal terms:
The way I do research now has nothing to do with how I did research two months ago.
Those comments suggest that AI is not merely being added to an existing workflow. For some researchers, it is changing the order of operations: what gets tried first, how quickly ideas are explored, and how a researcher moves from an open question to a possible proof or paper.
Math Feels The Pressure
The source article says math is especially feeling AI's impact. That makes sense within the facts of this case: a problem in quantum cryptography was solved with help from a model, and the result appeared in near-real time from two different directions.
The reactions described are mixed. Some researchers see new possibilities. Others feel a loss of identity. Both responses follow logically from the same change: when AI becomes a capable research partner, it can expand what people attempt while also challenging older ideas about expertise and authorship.
For fields built around hard problems, the emotional stakes are real. If a researcher once measured progress through slow, personal struggle with a question, a model that can rapidly assist with that process changes the experience of doing the work.
What This Means For Scientific Credit
The central issue is not whether the problem matters technically. The source makes clear that an open quantum cryptography problem was solved. The bigger issue is how the research community should understand credit when multiple teams use the same powerful tool and arrive at related results almost at once.
Several questions follow from this case:
- How should researchers describe the role of an AI model in a discovery?
- When two teams use the same model, how should overlapping results be compared?
- Does near-simultaneous submission strengthen the claim that the result was ready to be found?
- Should merged papers become more common when AI-assisted work converges quickly?
The article does not answer those questions, and it should not be stretched beyond what it says. But the episode clearly points to a new research environment. OpenAI's GPT-5.6 Sol Ultra was used by both teams. The papers went to arXiv.org three hours apart. The researchers involved are already thinking about combining their work.
That is enough to show that AI is changing how science gets done. It is accelerating attempts, changing research habits, and making independent discovery harder to define in familiar terms.