OpenAI's math breakthroughs shock mathematicians, spark identity crisis

OpenAI's recent publication of solutions to long-standing mathematical problems has detonated like a bombshell across the mathematics community, according to Robert Hart, The Verge's London-based AI reporter. In a new Decoder interview with Nilay Patel, Hart describes a discipline in 'shell-shocked' turmoil, facing a compressed version of the crisis that software engineering has already lived through. Leading mathematicians are questioning what their field even means if frontier models can simply answer the field's most elusive open questions, potentially rendering academic grants and university training programs obsolete.

Yet the situation is deeply paradoxical: AI systems remain genuinely awful at elementary arithmetic and basic counting—the classic strawberry test, counting R's, still trips them up—while simultaneously excelling at high-level abstract mathematical reasoning. This suggests the models' proficiency stems not from number-crunching but from advanced reasoning patterns embedded in academic papers. The debate has split the community: some see genuine breakthrough capability, while others suspect a marketing exercise by frontier AI labs indifferent to math's centuries-old academic legacy. Either way, the field faces urgent questions about what mathematicians should do next.