gpt-6POST
Speculation: OpenAI had a super-Astra model trained by August
A timeline speculation post argues OpenAI already had a super-Astra-level model trained in August and ran a short post-training focused on Millennium Prize-level proofs around August 28.
Jai
@Laneless_Archived 2026.09.14Originally posted 2026.09.09
Jai @Laneless_Originally posted 2026.09.09
afaict, everyone's accounts up to now are consistent with the following: - OpenAI had a super-Astra-level model already trained in August - On or just before Aug 28 a relatively short post-training run specifically focused on Millennium Problem-level proofs was started
This post quotes another post: x.com/laneless_/status/2097389367656632805
x.com/laneless_/status/2097418960287473936Archived 2026.09.14 16:11