How would you decide what to cut when a GenAI project at Meta has too many requirements for the available timeline?
