P2b shipped on 2026-06-05 (PR #60): the approver rubric (ADR-0007) and the quarterly atom audit playbook. The playbook's three thresholds are v0 defaults marked '⚖️ calibrate:' — they were NOT data-derived. P2a telemetry has been accumulating since 2026-05-30, so it's time to run the first real audit and calibrate.
On the host machine, follow the playbook (apps/docs/docs/guides/atom-audit-playbook.md):
- Run
pmk gateway atoms telemetry (and --json for detail).
- Apply the four review flags with the documented precedence (Questioned -> Dead-weight -> Stale -> Load-bearing -> Unflagged), action each, and record an audit-log entry.
- RECALIBRATE the three v0 thresholds against what you actually observe, and adjust them in the playbook:
- Dead-weight maturity window (v0: ~one quarter / 90 days)
- Questioned threshold (v0: questionedCount >= 2 OR (reuseCount >= 5 AND questionedCount/reuseCount >= 0.3))
- Stale window (v0: lastRetrievedAt older than two quarters with low reuse)
- Sanity-check that the telemetry is now REAL organic signal, not demo/test pollution (the AcmeAds demo atoms were unseeded on 2026-06-05).
Reference docs:
- Playbook: apps/docs/docs/guides/atom-audit-playbook.md
- Rubric: apps/docs/docs/adr/0007-atom-approval-rubric.md
- Priorities plan P2: apps/docs/docs/plans/2026-05-product-priorities-plan.md
The whole point of Path B was to defer these numbers to real data — now is when they get calibrated.
P2b shipped on 2026-06-05 (PR #60): the approver rubric (ADR-0007) and the quarterly atom audit playbook. The playbook's three thresholds are v0 defaults marked '⚖️ calibrate:' — they were NOT data-derived. P2a telemetry has been accumulating since 2026-05-30, so it's time to run the first real audit and calibrate.
On the host machine, follow the playbook (apps/docs/docs/guides/atom-audit-playbook.md):
pmk gateway atoms telemetry(and--jsonfor detail).Reference docs:
The whole point of Path B was to defer these numbers to real data — now is when they get calibrated.