Part 6/12:
What this reveals is that GPT-5 didn't produce new mathematical proofs or solutions; it simply acted as a remarkably effective search engine. It scoured academic literature for relevant papers and found references that had already solved these problems—papers Bloom hadn't yet encountered or hadn't linked to his database. This clears up the misconception: GPT-5 wasn't solving these problems from scratch; it was retrieving pre-existing solutions that were publicly available in the academic literature.