Sources & provenance

The trust layer for generated reports and AI output: where did this come from? Opt-in; not in the core bundle.

Generated 3 sources Human-reviewed

Verified Machine-generated Stale · 14d

The migration cut p95 latency by 38%[1], but error rate doubled in the first hour[2]. A later review attributed it to the old retry path Q3 incident review.

Sources

  1. Q3 incident review, §4

    ops.internal · postmortem

    2026-06-01 09:42

    The regression was sustained for 47 minutes before rollback; root cause was the deprecated retry path.

    Verified

  2. Model output — latency summary

    gpt-x · internal tool

    2026-06-01 09:55

    Error rate doubled for the first hour after deploy, then recovered.

    Generated

  3. Capacity plan (Q1)

    wiki · planning

    2026-01-14

    Superseded by the Q2 revision — kept for context.

    Stale