Berkeley researchers argue that as LLM inference gets cheap, AI agents — not one-off calls — become the dominant workload for data systems.
Continue to AI University →