AI agent evaluation misses broken products | VentureBeat

LangChain, Conviva and CoreWeave leaders say scoring AI agent conversations one at a time hides broken products — and explain the shift to cohort analysis.

Raccontata daventurebeat.com