The eval that finally caught a bad regression for me was not the one in CI. CI had run twelve hours earlier, against a frozen test set, and passed. The regression was sitting in live production traffi
Markdown supported · ⌘↵ to post
No responses yet.