Tomás Erhlich

8 stories by Tomás Erhlich

LLM Cost Optimization Through Observability Data

Trace-level attribution reveals hidden cost drivers aggregate dashboards can't see.

Framework choices determine which agent failures become visible during debugging.

Tracing captures observable system state but cannot judge correctness or find root causes.

Separating failures by harness layer reveals where debugging actually moves the needle.

User behavior in agent sessions reveals quality problems that explicit ratings miss.

Use production traces to rank agent fixes by actual failure impact, not complaint volume.

Capture silent production failures as replayable regression tests before they compound.

Observability data only matters if engineers know which harness layer to fix.