Tbh, I haven't read the paper, but I think it's pretty self-evident that large contexts aren't cheap - the AI has to comb through every word of the context for

Tbh, I haven't read the paper, but I think it's pretty self-evident that large contexts aren't cheap - the AI has to comb through every word of the context for each successive generated token at least once, so it's going to be at least linear.