Autelligence.AI Mission Control

Web-Brain Code System — Internal Dashboard

Rebuilt 2026-09-12 · numbers recomputed live from the KB by refresh-dashboard.py · INTERNAL

1,101

Files coded

across 18 active domains

12,445

Edges in the graph

3,152 curated links + 9,293 keyword kinship

Domain codes

→ 16 segment registries

1:1

Content preserved

backups hash-identical · reversible

Keep-both queued

9 Dean same-name + 4 contested

📊 Domain breakdown — by CODE PREFIX replaces 00–70 class distribution

Files per two-letter domain code. Recomputed on every nightly refresh from the segment registries (library/LIBRARY-*.md code→article rows). Sum = 1,101.

Reconciliation: LIBRARY.md declares 24 domain codes; 18 carry files today and are the 18 bars above. The other 6 — 2 private-walled (INV OS — Dean-private / Jim’s 80_ vault; never enter the shared brain) and 4 declared but not yet minted (CF AT SK WR). 24 declared = 18 active + 6 inactive. Bar counts sum to 1,101 — the same total as the Files coded KPI and the graph’s node count.

🕸 KB knowledge graph — live force-directed web RESTORED · moving

All 1,101 coded files as nodes, colored by domain-code prefix; 12,445 edges from real relationships ([[CODE]] cross-links and read_also / related fields, plus shared-keyword kinship). This panel runs the original dashboard’s graph mechanism, unchanged: the force simulation cold-starts every time you arrive at it and visibly settles — nodes drift, links pull, the web reorganises itself for several seconds. Every node carries its code label; the most connected ones also carry their article name; hover or tap any node for the full name in the panel. Drag a node to re-heat the web, scroll to zoom, click a legend chip to isolate one domain.

Hover or click a node for its article name

How the code system got here, and how every number above was counted. Nothing on this tab is a carried-forward snapshot — the figures are the same live values as the Overview tab, recomputed on the same nightly run.

🛠 Cutover state 2026-08-28

Executed: core/00 cutover ran 2026-08-28 (Jim-approved), verified and reversible. As of the 2026-09-12 refresh the coded brain holds 1,101 files routed through 16 segment registries.
Routers rewritten: LIBRARY.md → v2.0 thin code→segment router (24 codes declared); JD_MASTER_CONTEXT.md → task→code; index.md folded by augmentation (0 orphans). All PASS.
Shim: LIVE, not retired — dual-resolution (old 8-class paths + new codes) until a clean code-only session + Jim signoff. Legacy paths keep resolving through the deprecation window.
Fleet-sync: NOT run from the Cowork shell — it has no launchctl. Handed to Jim’s Main Terminal (diagnose read-only first, resume without -k).

🗂 Full code registry

Every active domain, its file count, share of the coded brain, and the segment file it routes to. These are the same live counts as the KPI row, the domain chart and the graph — they sum to 1,101. KB = governance/catalog + core files: 13 minted into codes so far; anything not yet minted still resolves through the retained index.md detail.

Code	Domain	Files	Share	Segment registry
GU	Gravitas Universe	318	28.9%	LIBRARY-GRAVITAS-GU
GG	Gravitas gameplay	248	22.5%	LIBRARY-GRAVITAS-GG
FR	Frameworks & Method (NCI)	110	10.0%	LIBRARY-FRAMEWORKS
CL	Claude / AI-ops	105	9.5%	LIBRARY-CLAUDE-OPS
BR	Briefings / logs	89	8.1%	LIBRARY-BRIEFINGS
MD	Media production	63	5.7%	LIBRARY-MEDIA
LI	LinkedIn / Bird Dog / Outreach	50	4.5%	LIBRARY-BUSINESS
SYS	System / infra / fleet	35	3.2%	LIBRARY-SYSTEM
RF	Reference / SOPs	18	1.6%	LIBRARY-MEDIA
AU	Autelligence (product/GTM/arch)	13	1.2%	LIBRARY-AUTELLIGENCE
OP	Operations (HotelStaff/Patriot)	13	1.2%	LIBRARY-BUSINESS
KB	KB standards / catalog / governance	13	1.2%	LIBRARY-SYSTEM
ST	Strategy / competitive	10	0.9%	LIBRARY-BUSINESS
SEC	Security / hardening / injection	6	0.5%	LIBRARY-SECURITY
AH	AHLA / trade shows	4	0.4%	LIBRARY-BUSINESS
WB	Web / site / GEO-SEO	3	0.3%	LIBRARY-WEB
EN	Entities / corp structure	2	0.2%	LIBRARY-BUSINESS
CX	Curiosity archive	1	0.1%	LIBRARY-CURIOSITY

🧭 Code → segment routing (the anti-bloat engine)

Three-deep router chain adopted from Dean's web-brain: two routers never grow, only the segment leaves do. Legacy 8-class paths still resolve through the crosswalk shim during deprecation.

🔀 Convergence mechanics

Edge model: internal cross-links are [[CODE##]] code-references (LOCKED). 12,445 edges counted live on 2026-09-12 — 3,152 curated link edges ([[CODE##]] refs, name wikilinks and markdown .md links, weight 3 each) + 9,293 shared-keyword kinship edges from keywords: frontmatter (weight 1 per shared keyword).
Merge-collision renumber (applied, reversible): CL62–106 → CL63–107 (+1) and GU191–320 → GU219–348 (+28), to clear Dean's live-claimed codes. One-step revert from the renumber snapshot.
Keep-both: 13 same-name/contested pairs queued to needs-jim-feed.md — never auto-picked, never merged (FR01, GU262, GU285, GU288 + 9 Dean-native candidates). Same 13 as the Keep-both queued KPI.
Periodic-merge model: Jim ⇆ Dean reconcile via periodic full-KB batch merges, operating independently between merges — not continuous live sync. No reserved code blocks; overlaps renumber at merge time.
Private walls held: INV (Dean-private) and OS / 80_ (Jim-private vault) never enter the shared brain.

🛡 Reversibility

Rollback: restore 3 routers from core-cutover-snapshot-2026-08-28/ + delete KB/library/ → full revert, zero article loss.
Backups: every class keeps a hash-identical source-backup/. That 1:1 preservation is what the Content preserved KPI reports.
Renumber revert: the merge-collision renumber is one step back from the renumber snapshot — see Convergence mechanics.
Nothing is destructive: the cutover changed routing, not content; no article was merged, rewritten or deleted by it.

🕸 KB graph — why the old one was inop, and what this shows instead

Old dashboard graph diagnosis (verified 2026-08-28 — historical, describes the retired graph only): the D3 force-graph read an inline dataset of RN = 342 nodes / RE = 905 edges. Those two figures are the retired July-era snapshot’s own numbers, not current counts — they are kept here purely as the before-picture. The dataset was structurally valid (0 missing edge endpoints, so not a dead or renamed source). The real defect: every node carried a retired 8-class domain — only 00–70/SYS keys existed, so FR showed 4 nodes. It predated the cutover, so at best it painted the wrong taxonomy: the “same stats” defect at graph level. It also depended on a ~700 KB inline D3 UMD bundle, a force sim invoked only on tab-switch, and a broken root-relative /logo.jpeg.

This dashboard’s fix (rebuilt 2026-08-28, build kbg-oldmotion-2026-08-28): the graph panel on the Overview tab runs the original dashboard’s graph mechanism verbatim — a fresh d3.forceSimulation with no pre-seeded node positions (d3 phyllotaxis start), alpha = 1 and d3’s default alphaDecay .0228 / velocityDecay .4, so the web visibly reorganises itself for several seconds on arrival; drag re-heats via alphaTarget(.3).restart(), and because the old build re-ran on every entry to the graph tab, this one re-heats on every entry to the graph card’s viewport. Descriptions are back the old way too: every node is a <g> carrying a permanent code label, the 80 most-connected also carry their article name on the canvas, and hover/tap opens the article’s full human name in the side panel. Data is the live code system: 1,101 coded nodes + 12,445 edges (3,152 curated links + 9,293 shared-keyword kinship) on the domain-code prefixes — the identical figures printed by the KPI row, the domain chart, the graph HUD and the provenance card below. FR now carries 110 nodes, against 4 in the retired snapshot. D3 v7.9.0 stays inlined (offline-safe); the baked static SVG remains underneath purely as the no-JavaScript fallback and is removed the moment D3 takes over. The previous build’s defect (fixed here): it seeded the sim from the baked settled positions at alpha(0.25) — mathematically alive but visually motionless — and put the article name in a hover-only tooltip, which does not exist on a phone. That is why Jim saw “no motion or descriptions”.

🔎 Data provenance — how each number was counted

Files per code (1,101): every CODE## row in library/LIBRARY-*.md (plus crosswalk-alias.md for anything not yet in a segment file) → dedupe by code → drop rows whose file no longer exists → count per letter prefix. Largest domains: GU 318, GG 248, FR 110, CL 105, BR 89.
Edges (12,445): 3,152 curated link edges — inline [[CODE##]] refs, [[name]] wikilinks and markdown .md links, resolved to codes and weighted 3 per reference — plus 9,293 shared-keyword kinship edges from keywords: frontmatter (weight 1 per shared keyword; a keyword appearing on more than 40 articles is treated as a topic label, not a relationship, and is skipped).
Hubs / isolated (975 / 43): computed live over THIS graph — degree ≥ 4 counts as a hub, degree 0 as isolated. These are code-graph figures, not the retired 8-class snapshot.
Codes/segments (24/16): LIBRARY.md domain registry + ls library/.
Renumber / keep-both / cutover: 00_Catalog_Governance/autelligence-migration-log.md §4–§4d + REGISTRY-RULES.md + KEEP-BOTH-LAYER.md.
Regenerated: every figure on this page is recomputed from the live KB by refresh-dashboard.py on the 2026-09-12 nightly run; nothing here is a carried-forward snapshot.

What Clark’s failures actually cost the operation, tracked so they get fixed rather than repeated (Jim, 2026-09-03). Every figure here is read straight from the human-maintained tracker ACCOUNTABILITY-TRACKER.md on the nightly run — this dashboard displays it and never writes to it.

⚖ Accountability read-only · daily

Ops impact of Clark’s failures — tracked so they get fixed rather than repeated. Read-only mirror of the human-maintained tracker; this job never writes to it.

Failures this week

across 5 tracked failure metrics

15 min

Ops time lost this week

estimated, from the tracker

Standing rules added

the improvement signal — compounding

Metric	This week	This month (2026-09)	All-time
Operator-Auth (over-deferment)	0	1	1
KB-Ref-Fail (not-looking-first)	0 ▼ 1	1	2
Cardinal Sins	0	0	8
Verify/Corrob Fail	1	1	3
Ops-time cost (min, est)	15	42	222
Focus-breaks	0 ▼ 1	1	4
Standing rules added (compounding+)	2	11	11

Standing rules added is the one line where up is good: each one is a rule banked so the same failure cannot recur. Arrows compare this week against last week from the tracker’s Weekly rollup table; green means moving the right way for that metric.
Source: ACCOUNTABILITY-TRACKER.md · parsed from the TALLY-BLOCK region on the 2026-09-12 nightly run · 7 metrics.