Lesson 042: Eval slices that compound
Focus
Bias toward observable metrics: latency, cost, escalation rate. Token Eval slices that compound:42 keeps neighbouring lessons differentiable.
Key ideas
- Thread: Eval slices that compound · drill v2 · spin
200814. - Habit: attach a trace_id to every completion you would paste into an ops dashboard.
- Guardrail: add one RACI bullet for prompt or index changes before tomorrow's standup.
Deep dive notebook
Synthetic drill artefacts
Token CFO scratchpad
- prompt_budget: 1574
- completion_budget: 545
- cache_key: `5480`
Hypothesis: halving completions moves P95 ~7% — record actuals.
Practice
Practice Paste the worked template into an internal wiki stub and name owners. — 42 Bump 31.