Context is such a curious problem for LLMs.

Context takes up vram/ram. So to have infinite context, you'd have to have infinite vram? The whitepaper link cant be clicked?