But also, given that we agree that o3’s prowess was overreported, and given that we didn’t see any really clean test of OOD challenges, and given that we don’t know what was in the data augmentation, why are you confident that o3 is an important step towards AGI? Another

→ View original post on X — @garymarcus,