Skip to content

Harness Levers

Each project changes one lever. Keep the task stable, change the harness, then inspect whether behavior improves.

The Levers

LeverQuestionProject
TraceCan you explain what happened?P01
ModelDid the task need the expensive model?P02
RetrievalDid extra search improve the answer?P03
DecompositionDid smaller prompts beat one large prompt?P04
MemoryDid durable context reduce re-discovery?P05
SafetyDid the harness bound the blast radius?P06
CriticDid evaluation stop false completion?P07
WorkflowDid model-authored orchestration reduce glue code?P08
EscalationDid routing recover when cheap got stuck?P09
HistoryDid indexing past sessions beat scanning them?P10

The Rule

Excessive harness features can lead to bloat. Use traces and metrics to ensure your harness is optimized for your work.

That is why the course teaches: predict, run, inspect, measure, keep.

Built as a friendly front door for the runnable OpenHands harness lab.