What would be the cost of achieving a significant breakthrough in an open problem that has confounded the academic community for decades? OpenAI's response: approximately $2,000 to solve ten such problems. Yesterday afternoon, OpenAI announced remarkable progress, revealing that its internal iteration of the next-generation flagship model, Astra, has yielded ten novel research insights across diverse fields such as high-dimensional geometry, coding theory, and group theory—some of which had seen no advancement for over a decade. These solutions were generated by Astra, subsequently compiled into research papers by human researchers, and the model further converted the solutions into Lean proof certificates to undergo formal verification.