AI Supplies Key Proof Step

Scott Aaronson says an advanced LLM (GPT5‑Thinking) contributed a crucial technical step to a new paper proving limits on black‑box amplification in the quantum class QMA. This is presented as his first paper where AI provided a substantive proof insight, not just boilerplate help. It suggests LLMs are now participating in genuine theoretical discovery. — If AI can generate novel proof steps in frontier theory, norms for credit, peer review, and verification in science will need to adapt.