On: OpenAI is well positioned to fast-follow Jev

[dgellow]

> autoregressive LLMs are indeed generating “probabilities”

I find it slightly more helpful to say they generate plausibility