Not really a bug.
but the way @ and Command+K works in jupyter right now is that it include everything which means cell type and execution count and all those jupyter related info.
perhaps if it just include the input and output and doesn’t include the other stuff, it would use less token and the quality of output will be much better.
