Programming using unreliable language models could finally justify PL research's pervasive focus on reliability and verification.
Realistically, though, everyone is just going to ask the AI to write unit tests.