Building the AI factory for self-improving agents: What's new in Arize AX

Arize AX introduces Signal for continuous production trace review, managed agents for investigation workflows, full-agent experimentation capabilities, Harness-as-a-Judge for adaptive evaluation signals, and voice agent support with audio session observability. Co-founders Jason Lopatecki and Aparna Dhinakaran describe how the platform automates the agent improvement loop from failure detection through root cause investigation, testing fixes, and deploying enhancements. The new capabilities are designed to help teams move from manually debugging agents to systematically improving them at production scale.