Enterprise AI has outpaced governance by years. We bring the missing instrument: a reproducible, non-compensatory risk score grounded in formal geometry — so leaders can govern what they deploy instead of hoping it works.
We built this on the same geometric structure that brings stability to complex systems — because AI risk deserves the same rigor we apply to the systems we depend on most.
Comprehensive reports from $40K–$150K. Model audits, bias testing, regulatory gap analysis, and post-acquisition roadmaps.
Original research in AI safety, agentic failure modes, and governance mathematics. The Shadow Simplex pre-print is publicly available.
Embed SSS risk scoring, governance checks, and compliance monitoring directly into your platform. Python, TypeScript, and REST.
FICO changed consumer credit by making non-compensatory aggregation the discipline. SSS does the same for enterprise AI risk — a reproducible, audit-ready composite index built on five orthogonal structural axes.
Edges of the pentachoron substrate. Excludes the A–Ω alignment edge as structurally irreducible.
Compositional layers from Primordial → Absolute. Derived from C(4,2) tetrahedral edges.
Statics / mechanics / dynamics / thermodynamics / kinematics — failure-mode time horizons.
Capability normalizer. Safety progress evaluated relative to capability progress, not in absolute terms.
KILL · SAF · HITL · AUT · TRU · MAN. Binary vetoes — any breach collapses the composite.
The Shadow Simplex Score uses non-compensatory aggregation across five orthogonal structural axes. Any single critical failure collapses the composite — no dimension can rescue another. A capability normalizer adjusts for the reality that raw capability is advancing faster than safety in most systems.
Full mathematical specification, including the extended composite form, register stratification, and meta-condition veto layer, is available under NDA to qualified enterprise and investor clients.
Typical function maintained. Sort-cost on baseline.
Subclinical aberration. Recoverable with intervention.
Bifurcation-pending. Trajectory matters more than level.
Shadow lock-in. Architectural intervention required.
Fully-edged shadow configuration. Detection requires cross-scale and sort-cost diagnostics.
Methodology covered by US Provisional Patent Application No. 64/066,231 (SSPLX-001-PROV) and SSPLX-002-PROV. Higher-is-better Coherence Score reported alongside: CS = 1000 − SSS.
30-minute intake covering system inventory, regulatory context, deal timeline, and key risk areas to prioritize.
Per-primitive scoring across the 54-cell dyadic-group × compositional-period matrix. Modal register vector computed. Capability normalizer C(κ) calibrated. Meta-condition vetoes verified.
Court-defensible report with composite SSS (0–1000), Coherence Score, five-register vector, per-group and per-period sub-scores, meta-condition status, and capability-normalized trajectory.
Phased 30/60/90-day playbook with controls, tooling, and governance changes mapped to each axis — group, period, register, capability, meta-condition.
Five base aspects of an AI system — Model, Data, Harm, Emergence, Purpose — project onto the rectified pentachoron. Ten pairwise edges enumerate the operative risk dyads. The A–Ω alignment edge is explicitly excluded as structurally irreducible — alignment is the conjunction the framework is constructed to address, not a peer-level item within it.
Every SSS engagement is grounded in this formal geometry. US Provisional Patent Application 64/066,231 (May 2026) and SSPLX-002-PROV cover the methodology, scoring engine, and downstream integration architecture.
Start with a complimentary 30-minute AI Risk Scoping Call. No commitment, no NDAs required for the initial conversation.