Problem

The diff view compares two sessions, but it does not explain clearly why one run is slower, costlier, or lower quality. The user has to infer the answer from raw metrics.

Current screenshot:

Expected direction

• Add a one-line verdict: slower because of duration, tools, failures, context, or model change.
• Show deltas for cost, tokens, duration, turns, tools, fail count, and health.
• Use direction-aware styling for better/worse values.
• Keep the side-by-side metrics, but make the explanation primary.

Validation notes

Captured from a real local run, not demo data.