Where are you getting that? Gpt-5-codex has absolutely blown my mind at how good it is even compared to GPT-5. One year ago we were at Sonnet 3.5, which needed a ton more handholding to complete tasks. The difference between say 3.5 and 4.5 is massive.

Ha that feels like the butt of this whole joke: “no we’re literally creating a god, what don’t you people understand? But this one will be a benevolent and wise god, not like a shitty one. Trust me.”