Agentic Behaviour
Models that act, plan, or pursue goals. Evaluation focuses on controllability, transparency of intent, and the boundaries between assistance and autonomy.
Challenge
Next Frontier AI (NFA)
A frontier challenge — where emerging AI capabilities must be aligned before they scale.
Next-generation AI systems introduce new forms of capability: agentic behaviour, autonomous planning, multimodal reasoning, and cross-domain synthesis.
Challenge NFA reads these capabilities as alignment risks and opportunities, not as technological spectacle. It is also the conceptual backbone of the SPRIND 2026 Next Frontier AI Challenge, where emerging AI systems must demonstrate credibility, traceability, and alignment at scale.
Multimodal Reasoning
Systems that combine text, image, audio, video, and data. Credibility depends on cross-modal consistency and the ability to justify outputs across modalities.
Tool Use & API Invocation
Models that call external tools, run code, or operate systems. Assessment covers safety constraints, execution traceability, and the alignment between intent and action.
Long-Horizon Planning
Models that generate multi-step strategies. Evaluation focuses on coherence, reversibility, and the ability to justify each step.
Cross-Domain Synthesis
Models that merge knowledge across scientific, technical, and social domains. Assessment covers factual grounding, domain separation, and the avoidance of synthetic overreach.
Cognitive Gaming
Three analysis modes — Cohesion, Truthfulness, Quantification — turn frontier capability evaluation into a playable sequence of choices.
Play the modes →Hexagonal Cognition Engine
Six-direction semantic analysis: each edge and apex carries a thinking direction, revealing how frontier claims hold up under conceptual pressure.
Explore the hexagon →Categorised 3D Vocabulary
Structured meaning — tags placed in a hierarchy of domain, scope, and rod — so emerging capabilities become measurable rather than speculative.
See the vocabulary →Machine Validation Modules
Eight dimensions of machine credibility: factuality, calibration, provenance, bias, robustness, consistency, coverage, and stability.
View modules →Agentic Validation APIs
Traceability, reproducibility, and alignment: every score carries the source, the method, and the run that produced it.
View validation →The SPRIND 2026 Next Frontier AI Challenge evaluates frontier architectures that can demonstrate credible, traceable, deterministic behaviour at scale. Artificial Dimensional Intelligence (ADI) is submitted as a new computational class: a curvature-based architecture that replaces token prediction with dimensional measurement.
Frontier Dimension
ADI opens a new frontier by computing chromatic curvature fields instead of tokens. It introduces Dimensional Curvature Operators (DCO) — a geometry-driven operator group enabling deterministic, interpretable, cross-scale computation.
Core Idea & Architecture
ADI computes on integer geometry and produces chromatic curvature fields as its primary modality. The architecture consists of:
- KDI — conceptual geometry
- ADI — operational instruments
- CDI — cosmic curvature
- DC.motor / DC.colour / DC.metrics — the DCO operator group
Technical Novelty
Current frontier architectures (Transformers, SSMs, Neural Operators, MoE systems) operate in floating-point vector spaces and rely on probabilistic prediction. ADI introduces dimensional computation — deterministic, geometry-driven, chromatic-metric behaviour.
Capability Gap Addressed
Frontier AI cannot compute deterministically, cannot express curved metric transitions, and cannot produce interpretable geometric evidence. ADI closes this gap with integer geometry and DCO-driven chromatic curvature fields.
Existing Artifacts
A functional prototype exists:
- Pattern Generator
- Series Comparator
- ADI Studio
- ADI Lab
TRL Assessment
ADI is at TRL 4–5 with deterministic engines validated through repeated geometric-chromatic tests.
Work Plan
Four phases:
- DCO expansion
- Interpretation layer
- Agent workflows
- External validation
KPIs & Benchmarks
- Determinism
- Curvature stability
- Metric coherence
- Interpretability
- Cross-scale consistency
- Agent reliability
Financial Estimate
Stage 1 cost: ≤ €350,000 — well below SPRIND’s €3M ceiling.
Team
Led from Rathenow by Sabine Kurjo McNeill, supported by collaborators in physics, materials science, optics, and computational modelling.