Topic

Agent Evals & Observability

If you cannot see the tool calls, you cannot trust the agent. This hub is about traces as ground truth, evals that match on-call reality, local models scored against production LangSmith runs, and observability that scales past hero engineers — including how OpenTelemetry fits an AgentOps stack.

The 11th-Hour Pivot: Lessons from Scaling AI Observability in the Enterprise
An in-the-trenches story of how an intern-led tracing prototype forced an 11th-hour pivot to LangSmith—and why observability, not model choice, became the true north for production agentic workflows.
Mar 12, 2026 · Agent Evals & Observability

The 11th-Hour Pivot: Lessons from Scaling AI Observability in the Enterprise