OpenAI has pushed artificial intelligence deeper into frontier mathematics, according to The Verge, which reports that the company revealed solutions to 10 long-standing problems using an advanced unreleased model called Astra. The reported results cover a broad set of mathematical areas, from abstract questions to problems with practical links to data transmission, error correction, networks, quantum game theory, and post-quantum cybersecurity techniques. The report frames the announcement as a turning point for a field that has traditionally moved slowly and prizes human proof, taste, and attribution. James Maynard, a University of Oxford professor and Fields Medal winner, told The Verge he has been doing “soul searching” over the past year as mathematicians confront what increasingly capable AI systems may mean for their work. The Verge describes Astra as a generative AI system able to draw on patterns, methods, and connections from the material it was trained on, then recombine known results and tools in new ways. In mathematics, that can mean linking methods across subfields or surfacing ideas from older academic literature that may not be obvious to a human researcher working within a narrower specialty. The reported problem set is wide. The Verge says one result involved how tightly spheres can be packed in more than three dimensions, a question connected to efficient data encoding and transmission. Another advanced work on error-correcting codes, which are used to recover information from noisy signals. Other reported results addressed the structure of complex connected networks, quantum game theory, and searches for targets inside high-dimensional grids, with The Verge noting possible implications for techniques used in post-quantum cybersecurity. One of the most closely watched examples concerns non-sofic groups: infinite mathematical structures that, in rough terms, cannot be approximated by finite ones. The Verge reports that whether such structures exist had been open for decades. But that result is also where the social and professional stakes become clearest: The Verge says there is a dispute over how much credit should go to OpenAI’s AI system and how much belongs to the human mathematicians whose recent work it built on. That credit question is not a side issue. Mathematical progress often depends on long chains of prior results, and the field’s reward system is built around identifying who supplied the crucial idea, proof, or construction. If AI systems can assemble new proofs from large bodies of prior work, mathematicians will need sharper norms for attribution, verification, and publication than the summaries here can resolve. The Verge’s reporting captures a split mood among mathematicians: excitement that AI could accelerate discovery, and anxiety about what that acceleration does to the people and institutions built around slow, individual or small-team work. The strongest version of the opportunity is that systems like Astra become discovery tools, expanding the search space and helping researchers connect distant ideas. The risk is that the field’s human pipeline, credit system, and sense of intellectual ownership are forced to adapt faster than its norms can absorb. For now, this remains a developing story because the provided material is a single reputable report and does not include independent confirmation of OpenAI’s claims, the full proofs, or the outcome of the credit dispute. What it does establish is that OpenAI is no longer merely proposing assistance for mathematics; according to The Verge, it is presenting its model as having contributed to solutions of serious open problems. Who benefits: Mathematicians and research groups may benefit if systems like Astra can surface useful connections across subfields or accelerate proof search. OpenAI also benefits from positioning its models as capable of contributing to high-status scientific work. Who's exposed: Human researchers are exposed to unresolved questions about attribution and career incentives if AI systems build on recent academic work in ways that are hard to credit cleanly. The field’s publication and verification norms may also come under pressure.