Unified Geometric Cognition (UGC) is a deterministic, auditable intelligence model built on CSIF (Crystal Structure Information Format) and RWIF (Resonant Wave Information Format).
UGC unifies math, logic, units, time, and contradiction geometry into a single coherent geometric reasoning substrate designed for CPU-only execution.
Unified Geometric Cognition (UGC) is a model architecture that evaluates mathematical, logical, structural, and temporal expressions through deterministic geometric transformations.
UGC is built on two foundational technologies:
CSIF is the representational ontology for:
- Crystals (valid structures)
- Anticrystals (invalid or contradictory structures)
- Units as first-class geometric objects
- Time crystals for deterministic chaotic behavior
- Phase, torsion, resonance, and trajectory
- Exact, auditable transformations
RWIF is the complete reasoning trace that records:
- Every operation
- Every phase update
- Every torsion spike
- Every contradiction event
- Every unit conversion
- Every inference step
- Every time-driven chaotic transformation
Together, CSIF and RWIF form the backbone of the UGC Model.
For the complete organized specification set (CSIF, RWIF, Semantic Layers, conformance, and implementation blueprints), see:
- Canonical specification source: this UGC-Model repository is the canonical source for CSIF, RWIF, and Semantic Layer specification edits.
- Downstream mirrors: CSIF-Guard may carry synchronized copies for operational implementation reference, but canonical changes must originate here.
- Sync versioning: every cross-repo spec sync should be tagged in commit
messages and release notes as
SPEC_SYNC_vYYYY.MM.DD.N(for example,SPEC_SYNC_v2026.05.31.1). - Sync scope: each sync must include an updated index/changelog summary listing files synced, source commit hash, destination commit hash, and effective sync version tag.
Drop-in OpenAI-style routes (/v1/models, /v1/chat/completions,
/v1/embeddings) plus deterministic CSIF-native endpoints for math,
retrieval, disambiguation, simulation, and reconciliation.
Direct command workflows for validation, migration, indexing, deterministic math evaluation, benchmarking, and local OpenAI-compatible serving.
Core operations produce replay-stable outputs with explicit audit traces so results can be re-run and verified without ambiguity.
Math, logic, units, time, contradiction geometry, and semantic disambiguation share one coherent CSIF/RWIF representation.
Contradictions are first-class objects with explicit threshold signaling, propagation, and qualified outcomes rather than hidden failures.
Deterministic token-to-sense resolution across multiple languages with cross-language alias identity and pack-scoped lexicon evidence.
Optional frame transitions and conservation policies provide deterministic projection, admissibility checks, and explicit invariant-violation diagnostics.
Branch-level what-if simulation and winner-versus-loser reconciliation expose why a trajectory wins and which alternatives are rejected.
Append-only sense trajectory logs support measurable semantic health signals, including stability, contradiction rate, ambiguity entropy, and lobe drift.
Designed for deterministic symbolic/geometric execution on standard CPU infrastructure without requiring matrix-heavy GPU inference pipelines.
UGC is organized in three conceptual layers:
Defines reasoning rules, transformation policies, and contradiction handling.
Defines crystals, anticrystals, edges, units, time crystals, and phase/torsion/resonance fields.
Captures deterministic, reproducible, inspectable, exportable execution traces.
Valid structures such as numbers, expressions, propositions, and units.
Contradictory or invalid states with full geometric traceability.
Meters, seconds, radians, degrees, joules, and related units as geometric objects.
Deterministic chaotic drivers for auditable randomness appearance.
Every evaluation step is a geometric path with phase, torsion, resonance, and causal ordering.
- Provide a reference implementation of the UGC Model
- Define CSIF and RWIF specifications
- Offer examples, tests, and demonstrations
- Enable open-source collaboration on geometric cognition
- Establish foundations for deterministic, auditable AI reasoning
- CSIF v1.0 schema
- RWIF v1.0 schema
- UGC Model definition
- Unit crystal specification
- Anticrystal lob specification
- Core crystal engine
- Phase/torsion/resonance propagation
- Unit crystal operations
- Time crystal integration
- RWIF trace generator
- Math reasoning examples
- Logical inference examples
- Unit conversion examples
- Chaotic time-driven randomness examples
- Contradiction detection examples
Contributions are welcome. Please use the contributor workflow:
Maintainer: Mogir Location: Grand Rapids, Michigan, USA
If you use this model or its specifications in research or software, cite:
Unified Geometric Cognition (UGC) Model - CSIF/RWIF Architecture Copyright (c) 2026 Mogir Jason Rofick
This repository is currently licensed under Apache-2.0.
From repository root:
cargo check --locked
cargo test --locked
cargo run -- serve-openai --port 8080- High-level technical reference: docs/TECHNICAL_REFERENCE.md
- Full specifications index: docs/specifications/README.md
- Release and sync discipline: docs/RELEASE_DISCIPLINE.md
- Change history: CHANGELOG.md