OpenAI's internal model solved roughly 90 problems from a set considered among the hardest open questions in mathematics. These are not textbook exercises. They are problems that have resisted professional mathematicians for years, in some cases decades.
The real story is not the score. It is the verification problem and the downstream implications. If a model can reason at this level in mathematics, that same reasoning architecture applies directly to physics, chemistry, and biology, fields where unsolved problems have direct consequences for drug discovery and materials science. The video covers how mathematicians are actually reacting, which is where the most honest signal lives.
The title is the argument: humans are now the bottleneck. Not the AI. Read the full piece to understand what that means for scientific institutions built around human-paced research cycles, and why the verification gap between AI output and human confirmation may become the defining constraint of the next decade.
[WATCH ON YOUTUBE →]