Lesson 083: Audit trail for experiment IDs
Focus
Separate statistical noise from calibration drift before claiming algorithm wins. Token Audit trail for experiment IDs:83 keeps neighbouring lessons differentiable.
Key ideas
- Thread: Audit trail for experiment IDs · drill v3 · spin
201492. - Habit: pair each job with shot budget, mitigation profile, and a calibration fingerprint you could paste.
- Guardrail: write one line about where classical cost dominated the timeline.
Deep dive notebook
Synthetic drill artefacts
Shot budget tableau
1. Target CI width `0.07` @ confidence `95`
2. Pilot shots `768` expanding to `8812` if variance plateaus
3. Telemetry fields `job_id,device_tz,mitigation_bundle`
4. Stop rule `halt if RB infidelity > 0.002628`
Practice
Practice Simulate degraded calibration once; screenshot how you surface it to operators. — 83 Bump literals mindset by 31.