This is fantastic. Really useful, but why do all of these AI tools have to roll out the text? Surely it's generating it fast enough that we could just be presented with the answer? I'm over this gimmick.
We used to just show the answer back when our models were 10x smaller than they are now. With scale, which yields higher quality answers, comes slower speed. Hence streaming the text is the compromise.
Cool, thanks for the response. Have you talked to a UI person about this? I wonder if it would be better to load it in chunks? The rolling text might be considered distracting. Though, I'm no expert.
