According to CoinMeta, OpenAI recently provided its new model with 4,000 unsolved mathematical problems. Researchers found that the model performed exceptionally well in solving these problems, generating 722 manuscripts that were categorized into 372 relevant results. OpenAI indicates that each result required an average of about three hours of computational reasoning. The company warns that some of the unverified results may have issues, which could lead to new bottlenecks for humans when reviewing the research. Although current financial benchmarks show that cutting-edge AI still faces challenges in complex investment research, its improved analytical capabilities could change the way researchers, engineers, or analysts work.