GPT-6 Astra strengthens old paper's theorem, Pangram scores it 100% human

Dominik Peters @DominikPetersOriginally posted 2026.09.15

Open original on X

Had GPT-6 Astra strengthen a theorem in a pre-AI paper of mine and @DelemazureTheo. For the write-up, I told it to only use words already in the original paper (or provide a justification for each new word) and to minimize new bigrams as well. Result: 100% human on Pangram.