Content Audit
Deep audit of old vs new repo content across docs, research, context, and archived work logs.
This is the deeper content pass across the repo, not only the docs app. The repository currently contains four overlapping documentation corpora, and the confusion comes from treating them as if they were one.
Inventory
| Corpus | Markdown / MDX files | What it is |
|---|---|---|
apps/docs/content/docs | 91 | Reader-facing docs app |
docs | 100 | Internal architecture, research, plans, briefings |
context | 119 | Extracted notebook and source-summary corpus |
archive/work-logs | 129 | Historical session and operational memory |
| Total | 439 | Before PDFs, HTML, and images |
The Core Problem
The repo has:
- current truth
- theory and synthesis
- extracted archaeology
- historical logs
The old docs UX flattened those layers together. That is why the reader kept seeing a mix of current, old, speculative, and archival material.
Current Canonical Sources
docs/architecture/unified-architecture-state.mddocs/design/BRAND_COLOR_SEMANTICS.mddocs/design/BRAND_FINALIZATION_2026-03-03.mddocs/research/RESEARCH_INVENTORY.mddocs/research/analysis/MASTER-SYNTHESIS.md- live code in
apps/*andpackages/*
High-Signal Findings
February vs March truth drift
docs/research/analysis/APPLICATION-STATE-OF-PLAY.md still describes the pre-CES gap. docs/architecture/unified-architecture-state.md and current code show that the CES alignment work later landed.
User-reachable stale docs existed
This pass found and fixed representative stale reader-facing nodes:
- legacy top-level network duplication
bootstrapUserexample drift in the protocols docs- old AI SDK naming in live communication/protocol pages
context/ is not optional anymore
The extracted notebook corpus now contains foundational material, especially:
notebook-17for business model and network projectionnotebook-26for network architecturenotebook-27for token transaction architecture
archive/work-logs is valuable but should stay archival
The archive still contains important provenance, especially for docs migration, deployment fixes, GraphQL cleanup, and production operations. It should inform docs maintenance, but it should not drive reader-facing copy directly.
If two repo documents disagree, prefer live code and the latest architecture tracker. Treat work logs and older research audits as context, not final truth.
Recommended Status Model
| Status | Meaning |
|---|---|
| Canonical | Safe to cite as current truth |
| Current reference | Maintained, but subordinate to canonical sources |
| Research | Analytical input that still needs code verification |
| Extracted notes | High-value context, not polished docs |
| Archive | Historical only |
What To Do Next
- sweep all user-reachable docs for old stack names and dead operations
- add explicit historical labeling to legacy/showcase pages
- create a first-class index for
context/ - thin and group
archive/work-logsby topic, not only date/session