UNDERSTORY holds the picture level.
The base beneath the primes, instrumented: what open sources can see of the lower tier today, what stays dark, and which data join lights the next layer.
Six numbers that level the picture
The whole console, compressed for a senior decision-maker: how much of the lower tier is still dark, what the funded portfolio looks like through an adversary’s eyes, whether any prohibited source appears by name, and where scarce vetting effort pays first. Every tile is computed live from the sections below it — tap any number to see the working.
Every dot is one firm
The illumination gap, drawn rather than stated. Each dot is one member of the public roster. A lit dot is a firm whose sector can be read from its name; a dark dot is one that cannot. Pick a filter to see where the non-US domicile candidates and the government-funded vendors sit, then hover a dot for the firm or click it for the dossier.
Sector and domicile are name-string heuristics — MODELED — and are not vetting, exclusion, or FOCI determinations on any listed company. Lit dots are grouped by sector, largest first; the dark tier fills the rest.
The threat isn't at the top of the DIB. It's underneath it.
The console's central metric is its own limit. The Opacity Ratio below is the share of the roster that resists even basic sector characterization from public name-strings alone — a direct, measurable proxy for the illumination gap DoW faces at scale across its 200,000+ supplier network.
All figures derive from the embedded public roster. Sector tags and domicile flags are keyword heuristics on the company name string only — MODELED — and are not vetting determinations. See Method & Sources.
Ask the console
Two modes. Question chips return pre-written analyst notes — SYNTHETIC, deterministic, keyed to this roster. The free-text field calls the live model through the duty-officer Netlify function (Anthropic API, AskRipple / CHINA DESK pattern) — responses arrive chipped LIVE. If the function or its API key is not deployed, the console says so plainly and falls back to the chips.
The geography of the problem
Four layers, each independently toggleable, each carrying its provenance on hover and click. ■ DEPENDENCY IN — material exposure flowing into the US base (USGS / EPRS / BIS aggregates). ■ US ARMS EXPORTS OUT — the other half of the ledger: top recipients by share of US major-arms exports 2020–24 (SIPRI Arms Transfers Database, free download). ■ ALLIED ROSTER — allied-domiciled firms recruited into this roster, weighted by live evidence-token counts. ● SINGLE-SOURCE ANCHORS — pulsing rings mark documented single- or limited-source production sites (State of Competition 2022; GAO shipbuilding & readiness reports): the choke-point section, placed on the map.
The globe draws entirely in the browser from country geometry bundled with this page — no external map service, library, or network call in the render path. Red dependency arcs carry national-level import shares from published USGS and EPRS aggregates — curated figures, never per-firm claims. Gold alliance arcs are weighted by a live count of roster names carrying that country’s evidence token; the count is real, the country attribution is heuristic and marked MODELED.
Six pressure indicators on the lower tier
Indicator pressure is a structured qualitative judgment — MODELED — expressed in ICD 203 estimative language and pinned to the public sources listed on each card. Levels reflect analytic reading of cited reporting, not computed telemetry.
Four fronts, twelve seams, one roster
The full I&S problem set for the DIB: the three macro-fronts from threat reporting (cyber, supply chain, human layer) plus a fourth front the reporting implies but statute makes explicit — the NDAA/DCSA regulatory seams (covered-source prohibitions, counterfeit parts, conflict-of-interest deconfliction). Three seams per front. Every cell carries: the threat pattern (from cited public reporting), a ROSTER LENS computed live from this dataset where the data permits, what to watch, and sources. Where a seam maps to a stress scenario, run it in place. Per-firm exposure classes from this framework appear in every member dossier.
Exposure classes are structural (population-level pattern matches), not incident data — MODELED, and labeled BASELINE where a seam applies sector-wide. Pattern descriptions stay at the level of the cited public reporting.
343 investments · 259 vendors · the government’s own bets, risk-scored
The Industrial Base Policy Investment Intelligence Center publishes what OUSW(A&S)/IBP is funding to fix the base — $8.07B across 343 investments (IBAS & DPA Title III). UNDERSTORY ingests that portfolio and scores each vendor for the security risk an adversary (China foremost, per NDS/NSS) could pose to it and its people, then renders a baseball card per vendor. The scoring is transparent and roster-cross-referenced: a funded critical-materials vendor sitting on a single published site, cross-listed in the DIBC lower tier, scores higher than a diversified prime.
Risk score is a transparent MODELED composite (materials/kinetic/micro sector weight + single-site fragility + DIBC lower-tier cross-listing + funding-visibility exposure), 0–100, shown with its factor breakdown on every card. It estimates adversary interest & targetability from open signals — not any company’s actual security posture, and never an allegation against a vendor. Investment facts are LIVE from the published portfolio; unresolved records are flagged, not fabricated.
Investment data LIVE from the IBP Investment Intelligence Center portfolio. Risk score MODELED (adversary-interest estimate from open signals), not a posture assessment or allegation. Company site: LIVE where inherited from the DIBC roster, CURATED for widely published official domains, otherwise omitted rather than guessed.
Both rosters, screened against the Chinese Military Companies list
DoW’s Section 1260H list of “Chinese military companies” (91 FR 35189) names ~70 parent entities and their US-operating subsidiaries. UNDERSTORY screens every name in the DIBC roster and the IBP investment portfolio against it, plus the NDAA §889/§5949/battery covered-source tokens. This is a word-boundary name screen — the necessary first gate, never a determination: a clean pass is not clearance, and the real exposure lives in ownership chains and resellers that only the SAM/beneficial-ownership joins reach.
List source: the published Federal Register notice (Defense Department, 2026-06-10, Docket DOD-2026-OS-1288). Matches are name-string only and must be verified against the notice; allied and coincidental name collisions are possible and are flagged for human review, not asserted.
Who’s funding the vendors DoW is funding?
DIB investment isn’t only DoW’s. Behind DIBC members and IBP-backed vendors sits private capital — and some of it traces to adversary states, sanctioned parties, or opaque offshore structures. This screen is a typology of nefarious/risky funding patterns drawn from public GAO/Treasury/CFIUS reporting, each with the signatures to look for and the public filings that would reveal them. It is deliberately not a per-vendor tool: cap-table beneficial ownership is exactly the data open sources do not provide, so this screen stays at the level of categories — what the patterns are, what their signatures look like, and which public filings would surface them.
This section is methodological reference material. The patterns, signatures, and examples summarize categories of risk described in published U.S. government reporting (GAO, Treasury under E.O. 14105, CFIUS annual reports, CRS, EPRS); risk tiers reflect the prominence of each category in that reporting, nothing more. Gimble Technologies makes no assertion, express or implied, about the funding sources, investors, ownership, or conduct of any company named anywhere in this console, and no such inference should be drawn from a company’s appearance here. Ownership determinations belong to the responsible authorities (CFIUS, DCSA, FinCEN) with data this console does not hold.
How the pieces connect
An interactive relationship map computed live from the two datasets on this page. Dots are companies; rings are sector hubs; lines connect each company to its sectors. Pick a dot or a name and the panel beside the graph lists what it connects to. A map shows what a list can’t: where government investment and structural opacity pile up in the same places.
Lenses and controls
Four lenses: ELEVATED VENDORS (highest adversary-interest scores), DIBC CROSS-LISTED (funded vendors that also appear in the small-supplier roster), 1260H / COVERED FLAGS (any name-screen matches; empty when the screen is clean), and CAPITAL PATTERNS (amber rings for the funding-risk patterns from the section above, connected to sectors where public reporting documents them — companies appear only through their sector membership, never linked to a pattern directly).
Click a dot or a name to select it and step from node to node in the panel · drag a dot to place it (it pins; RELEASE PINS frees them all) · drag empty space to pan · Ctrl + scroll or +/−/FIT to zoom · double-click a company for its full card · Esc clears the selection.
How deep does the light reach?
Commercial n-tier mapping practice works as a cascade: each documented tier is asked to reveal the next, until raw-material origin — and documentation completeness is tracked as a managed metric, not an aspiration. UNDERSTORY applies the same discipline to the DIB lower tier and reports, per tier, exactly where open-source coverage ends.
Illumination Completeness Scorecard
Single source of truth: dibc-data.js. Every dimension below is computed from it at page load — audit readiness as a number, per commercial best practice.
Sense → Adapt → Realign
Scenario management is the capability analysts rate as most differentiating in planning platforms, and the strongest commercial pattern is a sandbox that runs the same rules as production. This sandbox does exactly that: each scenario applies a deterministic filter over the live roster — the rule is printed with the result, glass-box — and returns roster-relative exposure with a sense/adapt/realign read.
Exposure counts are roster-relative and heuristic (name-classification only). They quantify visibility of exposure, not the exposure itself — no fabricated impact metrics.
1,745 members of the DIBC, searchable
Click any row or card to open that member’s dossier — per-firm illumination status, scenario membership, and public-record search links.
| Company | Status | Class | Sector (modeled) |
|---|
Where the bench gets thin
Sector counts within this roster, name-classification only. Thin benches (red: <20 identified firms; amber: <50) mark subsectors where, within this consortium's visible membership, redundancy is lowest — the same fragility pattern the 2022 State of Competition report documented at the prime level (tactical-missile suppliers 13→3; one remaining tank manufacturer), replicated one tier down.
Counts are roster-relative and heuristic; a thin bar means "few firms visible by this method," not "few firms exist." That distinction — visibility versus existence — is itself the illumination problem.
Non-US domicile indicators, name-string only
roster entries carry a corporate-suffix or name-string indicator of possible non-US domicile (Pty, Ltd., Limited, Canada, Australian, etc.). This is the shallowest possible FOCI aperture — and that is the point. Adjudicated FOCI determinations require beneficial-ownership data DCSA collects under Section 847; none of that is public. The gap between this list and a real determination is the visibility deficit GAO-25-107283 documents.
Screening best practice presents each candidate match with its evidence for a human decision — the Evidence column shows exactly which name-string pattern fired. Allied-nation domicile is not adverse. Many flagged entries are AUKUS/FVEY-allied critical-minerals firms deliberately recruited into the DIBC. The flag marks where beneficial-ownership questions begin, not where they end.
| Company | Evidence | Status | Class | Sector (modeled) |
|---|
What every number rests on
Each dataset the console draws on, how it is labeled, and which views it feeds. Nothing scraped, private, or export-controlled enters the build; raw workbooks never enter the repo.
| Dataset | What it is | Provenance | Feeds |
|---|---|---|---|
| Public DIBC membership roster | 1,745 unique records after de-duplication: company name, non-traditional status, business class, and the company link where the roster carried one. Snapshot 2026-08-11. dibconsortium.org/current-members | LIVE DATA | Dashboard · Roster · Graph · Duty Officer context |
| Sector tags and non-US-domicile indicator | Two deterministic keyword heuristics run over the company name string only. Inspectable in the header of dibc-data.js; the share that resists classification is reported as the Opacity Ratio. | MODELED | Dashboard · Roster · FOCI aperture · Sandbox |
| IBP investment portfolio | 343 investments across 259 vendors, $8.07B published (IBAS and DPA Title III), from the businessdefense.gov Investment Intelligence Center; workbook as of 2026-08-15. The adversary-interest score per vendor is computed here. | LIVE DATA RISK MODELED | IBP 343 · Dashboard · Graph |
| Section 1260H Chinese Military Companies list | The published Federal Register notice, 91 FR 35189 (2026-06-10, Docket DOD-2026-OS-1288), plus statutory covered-source tokens. Matching is word-boundary on names only: a first gate, never a determination. | LIVE LIST NAME-MATCH | 1260H · Dashboard · Graph |
| Dependency and alliance globe layers | Material dependency shares from USGS, EPRS and BIS aggregates; US major-arms export shares 2020–24 from the SIPRI Arms Transfers Database; anchor sites from GAO and CRS reporting. The allied layer is derived from roster domicile evidence. | CURATED LIVE-DERIVED MODELED | Intro |
| Warning board and threat framework | Analyst judgments in ICD 203 estimative language, each pinned to the public sources listed on its card. Levels are a reading of cited reporting, not computed telemetry. | CURATED MODELED | I&W |
| Capital-source patterns | A typology of risky funding patterns drawn from public GAO, Treasury and CFIUS reporting. Deliberately not per-vendor: cap-table ownership is data open sources do not provide. | CURATED TYPOLOGY MODELED | Capital · Graph |
| What-if scenarios | Deterministic rules run over the same roster records the explorer shows, with the rule printed beside every result. | MODELED | M&S |
| Duty Officer | Question chips return pre-written analyst notes. Free-text questions go to a live model only when the function is configured; the console says which is in effect. | SYNTHETIC CHECKING… | Intro |
Key citations
The references cited across the console, by ID and authority tier. Filter by tier or kind, or search a title.
Authority tiers: A1 primary government · A2 official research/PME · B1 FFRDC · B2 academic · C1 think tank · C2 industry association. Source names appear as published.
What lights the next layer
Verified free sources for future globe/map layers — the build queue for turning CURATED aggregates into LIVE flows. Order of attack: 1 USAspending place-of-performance join · 2 QCEW NAICS-336x county choropleth · 3 congressional oversight layer over the anchors.
- SHELF-01 · A1USAspending API + monthly award archives — recipient & place-of-performance geography; the join that lights Tier 2 — https://api.usaspending.gov/
- SHELF-02 · A1SAM.gov entity & exclusions public extracts — registration + debarment screening — https://sam.gov/data-services
- SHELF-03 · A1Census International Trade API — imports by HS code × country; commodity-specific dependency arcs — https://www.census.gov/data/developers/data-sets/international-trade.html
- SHELF-04 · A1USGS Mineral Commodity Summaries data releases (ScienceBase) — machine-readable import reliance — https://www.sciencebase.gov/catalog/
- SHELF-05 · A1BLS QCEW flat files — employment by county × NAICS; defense-manufacturing density choropleth — https://www.bls.gov/cew/downloadable-data-files.htm
- SHELF-06 · A1BTS Freight Analysis Framework — domestic freight flows by region × commodity — https://www.bts.gov/faf
- SHELF-07 · C1SIPRI Arms Transfers Database — free transfer-level download 1950–2024; already powering the exports layer — https://www.sipri.org/databases/armstransfers
- SHELF-08 · A1UN Comtrade public API — bilateral trade flows by HS code — https://comtradeplus.un.org/
- SHELF-09 · PDgithub.com/unitedstates/congress-legislators (+ committee-membership-current.yaml) — HASC/SASC members → oversight layer over the anchor sites — https://github.com/unitedstates/congress-legislators
- SHELF-10 · PDgithub.com/unitedstates/districts — congressional district GeoJSON — https://github.com/unitedstates/districts
- SHELF-11 · PDNatural Earth vectors + topojson/world-atlas — boundary geometry (public-domain standard) — https://github.com/nvkelso/natural-earth-vector
Glass-box method
Limitations
Name-string classification cannot see subsidiaries, ownership chains, DBA relationships, or actual contract activity; USAspending/SAM.gov joins are the designed next increment (DAT-001/002/003). Indicator pressure levels are analyst judgments, not computed indices. The roster is one consortium — large and unusually non-traditional-heavy, but not the whole lower tier. Nothing here constitutes a vetting, exclusion, or FOCI determination on any listed company.