Ambient Structure Discovery

Structural intelligence for the things
you don't know you don't know

A self-organizing computational architecture that detects structural anomalies in financial networks without labeled training data.

USPTO Provisional Patent #63/981,369

The system can't see
its own blind spots

$3.1T
illicit funds flow through the global financial system annually
$206B
spent on compliance globally, with 95% false positive rates
2%
of global financial crime flows are detected despite this spend

The Transparency Vacuum

In March 2025, the US Corporate Transparency Act was gutted—all domestic companies exempted from beneficial ownership reporting. Banks can no longer look up who owns an entity in a federal database. They have to figure it out themselves.

The Query Assumption

Current tools—databases, keyword matching, rule-based screening—search for what is declared. They require a query. But the query assumption is precisely what second-order ignorance violates.

You can't search for what you don't know to look for.
That's second-order ignorance.

Ambient Structure Discovery

A stigmergic mesh architecture grounded in Adaptive Resonance Theory. No labels. No predefined rules. No queries. The mesh discovers structure by processing signals it has never been trained on.

I

Eigenvalue Collapse

Loss of independent evaluation modes in the mesh's correlation matrix. When disparate entities converge on identical structural patterns—same layering, same nominees, same jurisdictions—the eigenvalues collapse. A single intermediary mass-producing shell companies becomes mathematically visible.

II

Linguistic Compression

Measurable shift from direct to hedged language. Entities engaged in concealment utilize overly broad, non-specific descriptions. "General consulting" and "international trade" replace concrete business purposes. The compression is quantifiable.

III

Action Decoupling

Discussion frequency decouples from resolution activity. Banks file thousands of suspicious activity reports while continuing to process billions in illicit transfers. The filing becomes the substitute for action. The mesh detects this ratio mathematically.

The Signal Construction Breakthrough

Raw financial data overwhelms the mesh with identity noise—names, addresses, proper nouns. The key innovation: replace text with complement-coded continuous structural vectors bounded in [0, 1]. Seven topological coefficients encode the shape of each entity's behavior, not its identity. The mesh then self-organizes by structural geometry.

D
Data Sources
PSC, GLEIF, ICIJ, Transactions
S
Structural Coefficients
DL, Fo, NP, JS, CB, VC, AD
C
Complement Coding
I = (a, 1−a) → 14-dim
M
Mesh Routing
ART resonance & BFS
A
Anomaly Detection
Gap detection & coupling

The mesh works

Three controlled experiments on the Elliptic++ Bitcoin dataset. Same mesh architecture. Same data. Different signal construction. Only complement-coded continuous vectors produce genuine structural differentiation.

Metric Text-Based
Negative Control
Structural Vectors
Complement-Coded
Improvement
Signal Dimensionality ~30 discrete terms 14-dim continuous [0,1]
Workers (self-organized) 12 → 60 12 → 60 Same expansion
Worker Differentiation All identical vocabulary Genuinely differentiated Qualitative shift
Anomalies Detected 1 59 59×
Licit Enrichment in Anomalies 2.2× (p < 0.00002) Statistically significant
Mean Familiarity Score 0.87 (saturated) 0.19–0.36 (differentiating) Real discrimination
“The mesh correctly identified that exchange and mining operations are structurally unusual in a network where illicit transaction patterns dominate the topological norm.”

This is not a failure—it is a genuine structural discovery. In the Bitcoin network, illicit-compatible activity is the norm. The labeled “licit” entities—exchanges, miners, wallet providers—are the ones with distinctive topology. The mesh found this without being told what to look for.

ASD does not detect “good” versus “bad.” It detects structural categories. The interpretation must be domain-aware.

The compliance market is
in structural failure

$6.12B
Beneficial Ownership Graph Analytics by 2033
From $1.56B in 2024. 17.2% CAGR.
$75B
Third-party AML software market by 2030
121% growth from $33.9B in 2025.
95%
False positive rate in current AML screening
Industry standard. The signal is buried in noise.

Regulatory Tailwind

US
CTA Rollback (Mar 2025)
Domestic companies exempt. Banks must internalize UBO discovery.
EU
AMLA Operational (Jul 2025)
Centralized enforcement. Mandates perpetual KYC and continuous monitoring.
UK
ECCTA (2023)
Companies House transformed from passive registry to active gatekeeper. Identity verification mandatory.

Built on public data.
Ready for institutional scale.

20.4M
nodes in the ownership graph
14.9M
ownership and control edges
110K
ICIJ-to-UK entity resolution matches
203K
Bitcoin transactions analyzed

Validated Data Sources

UK PSC Register (10M companies) ICIJ Offshore Leaks (810K entities) GLEIF Level 2 (469K relationships) Elliptic++ Bitcoin Dataset OFAC SDN List Companies House BasicCompanyData

Seeking partnerships with data exchanges and insight platforms that aggregate licensed financial transaction data. The architecture is built. The methodology is validated on public datasets. The next step is institutional-scale validation on real-world transaction flows—credit card panels, wire transfers, SWIFT messages, and correspondent banking networks. If you operate a data marketplace or insights exchange that serves financial institutions, this is a new intelligence layer for your platform.

Standing on established theory

The full theoretical framework is detailed in the ASD paper: Ambient Structure Discovery via Stigmergic Mesh (PDF).

1987

Adaptive Resonance Theory

Stable category formation with vigilance-gated plasticity. New patterns create new categories; familiar patterns reinforce existing ones. Solves the stability-plasticity dilemma.

Carpenter & Grossberg
1959

Stigmergy

Coordination through shared environment rather than direct communication. Agents leave traces; other agents respond to those traces. No central coordinator needed.

Grassé; Theraulaz & Bonabeau, 1999
1982

Crawford-Sobel Signaling

Information degradation under strategic communication. When sender bias exceeds a threshold, communication collapses. Used to detect organizational signal decay.

Crawford & Sobel
1997

Spectral Graph Analysis

Anomaly detection via Laplacian eigenvalue distribution. Structural changes surface as spectral perturbations before they become visible in metrics.

Chung
1956

Simon's Satisficing

Stop-on-first-accept routing. Bounded rationality over exhaustive optimization. Fast, stable, and biologically plausible.

Simon

Let's find what's hidden

Jeremy McEntire

Inventor & Patent Holder
Ambient Structure Discovery

Seeking data partnerships for institutional-scale validation.

If you operate a financial data exchange, insights platform, or alternative data marketplace—we want to talk. Your data is the signal. Our mesh is the intelligence layer. Together, we surface what no existing compliance tool can find: the structural anomalies that nobody knows to look for.

For financial institutions: if you're facing the post-CTA reality of building internal UBO discovery capabilities, the mesh integrates with your existing transaction data infrastructure. Consumption-based, no retraining required, no labels needed.

Start a Conversation