GenAI Basic — 047: pair eval slices on `Evaluation habits that compound` — memo `82148 [47]`

Lesson 047: Evaluation habits that compound

Focus

Prefer explicit failure rehearsals over aspirational wording. Token Evaluation habits that compound:47 keeps neighbouring lessons differentiable.

Key ideas

• Thread: Evaluation habits that compound · drill v7 · spin 687353.
• Habit: pair every model utterance with a trace_id you could paste into Grafana.
• Guardrail: write one RACI bullet referencing this lesson tomorrow.

Deep dive notebook

Synthetic drill artefacts

Prompt scaffold

ROLE: Incident analyst cohort 47
INPUTS:
- excerpts tagged [chunk_id ...]
- guardrails referencing policy_bundle_3
TASK:
  1) Summarize deltas with citations
  2) Confidence label LOW|MED|HIGH + evidence
  3) If facts missing emit MISSING_FACTS list
USER_SEED_QUESTION >>> What changed between rollout 32 and 22?

Practice

Practice Attach rollback steps if evaluator variance spikes. — 47 Bump literals mindset by 29.