Has AI finally become a mathematician, or not?

The author explores the current capabilities of artificial intelligence in mathematics, prompted by recent high-profile announcements from OpenAI and Anthropic. These companies have been frequently reporting new records and discoveries made by their models in solving complex mathematical problems. The article examines whether modern neural networks can be considered true mathematicians capable of independent scientific research, or if their successes are limited to statistical data processing and pattern recognition within training sets. The author attempts to distinguish between marketing hype and genuine breakthroughs in automated theorem proving and competitive mathematics. The piece invites readers to discuss the boundaries of AI's cognitive abilities and how the role of the mathematician is evolving in the era of large language models.