OpenTelemetry in the age of AI

The discussion centers on the evolution of observability in the context of AI, particularly generative and agentic AI. Austin Parker, a core contributor to the OpenTelemetry project, explains Observability 2.0 as a shift towards treating telemetry data—logs, metrics, and traces—as an interconnected stream, enhanced by context. This approach is crucial for understanding AI systems, where traditional monitoring tools fall short due to the variable nature of AI outputs. OpenTelemetry's semantic conventions aim to standardize data for AI, benefiting both model builders and developers integrating AI. Parker also touches on the debate around instrumentation, advocating for a hybrid approach that balances built-in telemetry with specialized, high-performance tools.