OpenAI reports that its AI agents solve a long-unsettled mathematics problem that dates back about 90 years. According to the coverage, the effort involves large-scale prompting and computation, with the agents processing an estimated 130 billion tokens before producing a solution.
The outlets describe the work as a test of the agents’ ability to tackle difficult, previously unsolved questions rather than smaller or routine tasks. Business Insider and Yahoo News both focus on the scale of the experiment and the age of the problem, presenting it as evidence of improved problem-solving capability. While neither source emphasizes conflicting technical details, the reporting mainly differs in framing—one highlights the token count and the “cracking” narrative, while the other presents the same development as a headline item—rather than providing divergent accounts of the method or results.
