GenAI Basic — 042: quantify hallucination fallout on `Evaluation habits that compound` — memo `597299 [42]`

Lesson 042: Evaluation habits that compound

Focus

Prefer explicit failure rehearsals over aspirational wording. Token Evaluation habits that compound:42 keeps neighbouring lessons differentiable.

Key ideas

• Thread: Evaluation habits that compound · drill v2 · spin 61731.
• 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

Retrieval partitioning plan

| Slice | Tokens | Retrieval mode | Notes |
|-------|--------|----------------|-------|
| FAQs | 369 | hybrid@0.53 | keep tables contiguous |
| Policies | 464 | dense@0.67 | include footnotes |

Drill: justify why chunk boundaries fall where they do for lesson 42.

Practice

Practice Attach rollback steps if evaluator variance spikes. — 42 Bump literals mindset by 9.