Teams use GPT-5.6 to solve quantum crypto problem

Two separate research teams have independently cracked an open problem in quantum cryptography using OpenAI's GPT-5.6 Sol Ultra, submitting their respective papers to arXiv.org a mere three hours apart. On one side of the breakthrough was Seyoon Ragavan, a PhD student at the Massachusetts Institute of Technology (MIT). On the other side, professors Prabhanjan Ananth from the University of California, Santa Barbara (UC Santa Barbara) and Amit Sahai from the University of California, Los Angeles (UCLA) worked together to reach the same conclusion. Both groups targeted "unclonable encryption," a security method that leverages quantum properties to prevent unauthorized copying of data.

Although both teams utilized the exact same AI model, they employed distinct methodologies to arrive at their solutions. Now, the researchers are discussing the possibility of merging their papers. This near-simultaneous discovery highlights a shifting paradigm in academic research, where the availability of powerful frontier models can lead to parallel breakthroughs. Ananth noted that checking if a model like GPT can solve an open problem has become the initial step in modern scientific inquiry.

For AI practitioners and academic researchers, this milestone signals a profound shift in how daily work is conducted. Ragavan observed that his current research methodology bears no resemblance to his workflow from just two months prior. As AI models become capable of resolving complex mathematical and cryptographic challenges, the traditional timeline of scientific discovery is compressing rapidly. Practitioners must now grapple with a new reality where the competitive edge lies not just in identifying open problems, but in how creatively they can prompt and guide AI systems to solve them. This transition is sparking a mix of intense excitement over accelerated progress and existential anxiety regarding the changing identity of human researchers.