The foundation work that makes the pipeline trustworthy long-term: canonical taxonomy, a verification ledger schema, and three architectural findings that need fixing before Phase 6 runs.
TAXONOMY.md is now the canonical source for city and region values across the AICV content collection. It replaces the deprecated "zones" system, which had drifted into inconsistency across the 80-node library. Every node, brief, and structured data surface should now reference TAXONOMY.md for the correct values — not whatever happened to be in the frontmatter when the content was first written.
This matters for citation quality. An LLM querying the AICV intelligence layer needs consistent city field values to route queries correctly. Inconsistent taxonomy means inconsistent answers, which means lost citations.
Drafted in aicv-playbook/ with a two-shape design for org/ — one shape for node verification records, one for brief verification records. The scaffold itself is deferred: the ledger queue currently sits at approximately 17 items from Phase 6, and the right move is to let the queue grow to a meaningful batch size before running the scaffold rather than building infrastructure for a handful of items.
The schema draft is stable. What's waiting is execution timing, not design.
The session surfaced three issues that need resolution before the next major content pass. First: a silent 78-vs-80 node drift on org/, caused by the retired Mintlify docs.json fetch that was still being called and returning stale data. The site was displaying 78 nodes when 80 exist. The computed stats pipeline fixes this going forward, but the root cause — the orphaned fetch — needs to be fully retired.
Second: city field drift across the 80-node library. Some nodes are using deprecated zone values, some are using informal city names, some are correctly using the TAXONOMY.md canonical values. A content pass to normalize all 80 nodes to the taxonomy is now scoped as part of Phase 6.
Third: residual {node.zone} template renders appearing in at least one location — a template variable that was never replaced with actual content. Small surface area, easy fix, but worth flagging explicitly so it doesn't get missed in the Phase 6 batch.
Phase 6 ledger queue is at approximately 17 items. The scaffold runs when the queue justifies it. The city field normalization pass across 80 nodes is the most load-bearing item — that's what makes the taxonomy work real rather than aspirational. The orphaned Mintlify fetch needs to be cleanly retired. Then the verification ledger scaffold can run against a clean foundation.