There are numerous errors in the coverage of #2: not all 9 experts verified the dis/proof (some limited their remarks to larger implications); none of the 9 experts were critics of the conjecture (most human experts assumed the conjecture was true, and that assumption was one reason the problem remained unsolved for decades); the linked paper is a different paper from the companion remarks that are found on the OpenAI page (in fact, Sawin may not even have used GPT for that solo paper, although he used GPT's discovery as the starting point for this linked preprint).
There are numerous errors in the coverage of #2: not all 9 experts verified the dis/proof (some limited their remarks to larger implications); none of the 9 experts were critics of the conjecture (most human experts assumed the conjecture was true, and that assumption was one reason the problem remained unsolved for decades); the linked paper is a different paper from the companion remarks that are found on the OpenAI page (in fact, Sawin may not even have used GPT for that solo paper, although he used GPT's discovery as the starting point for this linked preprint).
Thanks for the note. Had a few papers and notes on the actual OpenAI report I wanted to highlight. It was a mix-up. It should be fixed now.