AI agents still struggle with real outages
A new benchmark tests AI agents on real production incident response, and top models still fail more than a third of the time.
A new benchmark tests AI agents on real production incident response, and top models still fail more than a third of the time.