@RokoMijic i think we can go further w efficiency @ inference time, getting better model performance from smaller models that run locally ... and that large context performance can improve more ... and that wrappers & platforms & ecosystem integrations will get data flowing