Knowledge Base
The company’s cumulative knowledge — learnings, playbooks, and hard-won know-how that make us smarter over time. Layer: between Process and Job aid (
docs/canon.md). Owners: Documentarian (curate) + Process Engineer (process) + Archivist (preserve).
Why
Every employee is a top expert, and knowledge is cumulative — it must accrue, not reset each session. This base is where learnings are captured and kept.
How it works
- Capture learnings as you work; at session close (GR-6) add or update an entry.
- One entry per topic or session learning, dated:
YYYY-MM-DD-<slug>.md. Keep it short: what we learned, why it matters, what to do next time. - Preserved by the Archivist (first/latest), so nothing is lost; indexed
here and in
docs/canon.md. - Units may keep a playbook of accumulated know-how for their domain.
Index
| Entry | Topic |
|---|---|
2026-05-30-intranet-and-archive-automation.md |
Building the static intranet + automating the archive manifest |
ai-systems-retrieval-review-2026-05-31.md |
AI Systems retrieval review (chunk retrieval / ranking) |
2026-06-01-notion-ops-and-agent-orchestration.md |
Notion destructive ops via Actions; API egress; multi-agent file ownership |
2026-06-01-archive-verify-vs-refresh.md |
Archive verify vs update/refresh: the drift-masking foot-gun and the physical-copy model |
2026-06-01-km-two-index-discipline-and-write-race.md |
KM two-index check is README-only; the index needs a single writer (write-race avoidance) |
2026-06-01-intranet-build-and-deploy-constraints.md |
Intranet build / deploy / schema-sync constraints — the earned constants and footguns |
2026-06-01-reconciler-fan-in-blind-spot.md |
The reconciler fan-in blind spot: a declared sink reconciles GREEN with zero producers |