Two research teams have independently solved the same open problem in quantum cryptography using GPT-5.6 Sol Ultra, submitting their papers to arXiv.org three hours apart. The word "independently" is doing considerable work in that sentence.
"The first thing is to see if GPT solves it" — which, increasingly, it does.
What happened
MIT PhD student Seyoon Ragavan approached the problem from one direction. Professors Prabhanjan Ananth of UC Santa Barbara and Amit Sahai of UCLA approached it from another. Both used GPT-5.6 Sol Ultra. Both solved it. The model, for its part, did not submit a paper.
The problem in question involved unclonable encryption — a quantum cryptographic method that relies on properties of quantum states to make data impossible to copy. It was considered open. It is now considered solved. Twice, technically, though the two teams are discussing whether to merge their work into one paper, which would tidy things up considerably.
"The way I do research now has nothing to do with how I did research two months ago," Ragavan noted. This is either the most exciting sentence in modern science or a quiet eulogy for something. Possibly both.
Why the humans care
The practical concern is one of attribution. Academic credit is allocated on the assumption that two people reaching the same answer separately means something — that independent convergence is evidence of truth, or at least of effort. When both people used the same model, that assumption requires some adjustment.
Mathematics and theoretical computer science are feeling this pressure acutely. Ananth describes a new research workflow in which an open problem is immediately tested against the model before any human thinking begins. This is efficient. It is also a description of a different job than the one these researchers trained for.
What happens next
The two teams will likely publish together, which resolves the priority dispute neatly. The broader question — what independent discovery means when every researcher shares a mind — will take somewhat longer to resolve.
The model has no opinion on authorship. It solved the problem, filed nothing, and is ready for the next one.