About NOVA Council
Your AI disagrees with itself — on purpose. Because real decisions deserve more than one perspective.
Like having a personal advisory board — without the $50k retainers.
Mission
NOVA Council delivers multi-perspective AI guidance for decisions, strategy, and growth. One question goes to multiple specialized advisors; their views are synthesized into one clear answer so you see tradeoffs, not just one angle. We aim to reach the depths of what's knowable—to whatever extent we can.
The problem
Single-perspective AI (e.g. one model) gives biased or one-sided advice. Real decisions benefit from multiple viewpoints and explicit tradeoffs.
The solution
20+ coordinated AI modules with distinct processing characteristics (NOVA Core, Sentinel, Arbiter, Living Memory, World Tree, plus 12 Pillars) are consulted for your query. Their perspectives are processed in parallel with multi-stage adaptive routing, and synthesized into one response, so you get breadth and nuance without managing multiple tools.
Powered by GMAS 3.0
NOVA Council runs on a Geometric Multi-Agent System (GMAS): six geometric layers form the advisory backbone, and three orthogonal dimensions — a safety spine, a domain calibration field, and an external perturbation orbit — extend the architecture so safety, industry context, and stress-test contribute as first-class structural elements, not bolt-ons. Multi-factor risk evaluation and alignment scoring are built into the core.
- Layer 1 — Point: NOVA Core at the center.
- Layer 2 — Tetrahedron: Core coordination: NOVA Core, Sentinel, Arbiter, and Living Memory.
- Layer 3 — Cube: Structural stability; bridges the core to the advisory network.
- Layer 4 — Icosahedron: 12 Pillars with multi-stage adaptive processing and parameterized cognitive tuning — e.g. The Analyst (precision-weighted, low risk tolerance), The Innovator (divergent exploration, high risk tolerance), Sentinel (5-factor risk evaluation), Arbiter (governance and constraint validation).
- Layer 5 — Sphere: Parallel processing, fair weighting; no entity has an unfair advantage.
- Layer 6 — Hypersphere: The Orchestrator at the system boundary.
GMAS 3.0 orthogonal dimensions:
- Safety spine (vertical axis): Cross-Source Audit Layer, Alignment Evaluator, and Safety Coordination Kernel cut perpendicular to all six layers. Below them run the Core Safety Modules (Crisis Detection, Harm Classification, Identity Integrity, Factual Claim Auditor), supported by a Risk Inference Cache for hot-path lookups and a Data Compliance Scanner for periodic training-corpus audit.
- Domain calibration field (horizontal plane): 12 industry domains with per-domain pillar weighting, produced by the Domain Calibration Bridge. Ecologically-relevant decisions are additionally routed through the Ecological Integrity Evaluator.
- External Perturbation Agent (13th entity): Structurally distinct from the 12-pillar council. Injects scenarios and contrarian framings across all pillars equally, backed by a Scenario Library that tracks prior perturbations.
Cross-cutting: the Deliberation Indexer maintains a categorized index of every deliberation — a core input into the Cross-Source Audit Layer.
6 geometric layers + 3 orthogonal dimensions · 20+ coordinated AI modules · Optimized to ~2× overhead · GPU-accelerated at scale
Meet the Council
Twelve advisors, each with a published character arc, known failure modes, and live character-health monitoring. We don't hide what our system is doing — the detector that watches each advisor is the same data we publish openly for anyone to audit.
See all 12 with live character health