NVIDIA published several pieces showing how AI agents can be used to investigate production issues and run AI factories, moving agents toward operational practice.
NVIDIA's developer blog explains how its Factory Operations Blueprint coordinates multiple AI agents to investigate production problems. In the same week it also covered streamlining AI factory operations and using NeMo to turn agent failures into tests and harness fixes.
The pieces cover a full loop of investigating, testing and fixing agent behavior, shown through NVIDIA's own use and tooling rather than a single feature. Google Research's work on helping humans define gray areas appears alongside it, pointing to human-in-the-loop designs. Note: the scale of real deployments is not confirmed in the articles.
Watch for adopters of the blueprint outside NVIDIA, and whether the NeMo failure-to-test approach spreads into ordinary agent development workflows.