tl;dr: Amit Sahai argues that as AI systems begin producing genuinely novel mathematical ideas that outpace human researchers, mathematicians shouldn't abandon the field but instead form collaborative groups dedicated to understanding AI-generated results. He warns that consequential decisions—like approving a novel AI-designed fusion plant—require human communities capable of independently understanding the underlying reasoning, not just trusting AI outputs. This "deployable intellectual reserve" of mathematically sophisticated humans is essential for preserving meaningful human agency, and will require significantly expanding, not shrinking, the mathematician workforce.