According to Greg Brockman (@gdb) and Terence Tao, GPT-5.2 Pro has reached a significant milestone by independently solving an Erdos problem—a first for large language models (LLMs). This achievement demonstrates the increasing capability of AI tools in tackling complex, unsolved mathematical problems, highlighting practical applications in scientific research and advanced mathematics. The success of GPT-5.2 Pro in this domain signals major business opportunities for AI-driven research platforms and could accelerate the development of AI-powered discovery tools for mathematics and STEM fields. As noted on Twitter, this breakthrough is expected to drive further momentum and investment in AI solutions for scientific and academic markets (source: https://twitter.com/gdb/status/2009739147402244381).