Cortex Docs Map

At-a-glance map of what lives where in the Cortex docs and why.


On this page
  1. By Kind
  2. By Question
  3. Canonical Docs Versus Dated Artifacts
  4. Related

Cortex Docs Map

A one-page orientation to the documentation tree. When in doubt about where a new doc belongs, start here.

By Kind

KindWhereLifetimeDated?
Usage guideUsage/Long-term canonNo
Canonical architecture chapterArchitecture/Long-term canonNo
Normative specificationReference/Long-term canonNo
Code style guideStyle.mdLong-term canonNo
Architecture decision recordADRs/PermanentNo
Downstream consumer bindingConsumers/{consumer}.mdLong-term per consumerNo
Logos reasoning libraryConsumers/Logos/Long-term downstream library canonNo
Active roadmap planRoadmap/Plans/Active until completed or supersededNo
Active roadmap epicRoadmap/Epics/Active while retainedNo
Completed initiativeRoadmap/Completed/Historical, retained selectivelyNo
Archived ideaRoadmap/Archive/Historical, retained selectivelyNo
Paper manuscript and figuresPublications/paper-N-*/Published-onceNo
Paper notes and ideasPublications/Notes/WorkingNo
Research or synthesis memoResearch-notes/{scope}/Historical, retained selectivelyYes
Controlled experimentExperiments/{scope}/Epic-controlled, retained selectivelyYes
TemplateTemplates/Long-term canonNo

By Question

Where do I document a settled architecture decision? Use ADRs/ with the next NNNN- number.

Where do I write up the Wire grammar surface? Use Reference/Wire/. Canonical language rules live in reference docs, not research notes.

Where do I explain how to use Cortex? Use Usage/. Usage pages should lead with working paths, commands, and decision points. Link to Architecture/ for rationale and Reference/ for exact rules.

Where do I document runtime architecture? Use Architecture/06-pulse-runtime.md for the stable model and Architecture/07-rewrites-and-materialization.md for topology evolution.

Where do I document LLM or model-mediated reasoning workflows? Use Consumers/Logos/. Cortex architecture should describe the substrate those workflows run on, not the reasoning library itself.

Where does downstream-specific design belong? Start with Consumers/{consumer}.md. Consumer docs may contain concrete product bindings; Cortex architecture should stay consumer-independent. Create a consumer subdirectory only when a public binding grows beyond one page.

Where does an implementation plan go? Use Roadmap/Plans/ while it is active. Move or delete it when it ships, is superseded, or stops explaining the current plan.

Where do dated synthesis notes go? Use Research-notes/ under the right scope. Keep only notes that still explain the current canon or a live research thread.

Canonical Docs Versus Dated Artifacts

Canonical docs are edited to remain current. They should frame Cortex as an independent upstream substrate and use downstream products only as examples.

Dated artifacts preserve reasoning from a point in time. They are retained only when they still help explain the current system. Stale transition notes, issue-specific migration plans, and superseded research snapshots should be removed instead of preserved by default.