Lattice Physics · Neuromorphic Architecture

The geometry
determines the answer.

Calera Computing derives cognitive architecture from mathematical first principles — not statistical approximation. We build systems where every output is provable, traceable, and deterministic.

Truth first.
Always.

Calera Computing, Inc. is a Delaware C-Corporation building the next generation of deterministic cognitive systems. Our architecture is organized around algebraic constants that emerge from lattice geometry — the same constants that govern energy propagation in simplicial structures across dimensions.

We don't approximate. We don't predict. We derive. Every architectural decision traces back to a mathematical proof. That's not a philosophy — it's a structural constraint.

σ
Lattice Physics
Architecture organized around the xd = x + 1 polynomial hierarchy — including the 4D σ-constant (σ ≈ 1.22074) — representing geometry-native decay bases across dimensions.

Our architecture utilizes the polynomial family xd = x + 1, which generates geometry-native organizing constants for simplicial lattices across dimensions.

This includes the 4D σ-constant (σ ≈ 1.22074), which governs energy propagation on the A₄ root lattice, alongside φ (2D) and ρ (3D).

Our newest paper B-02 provides cross-dimensional spectral validation on Ad root lattices, proving that these constants are uniquely derived from Laplacian spectral properties across dimensions with zero free parameters.

Key Insight: B-02 proves that the d-dimensional hyperbox subdivides into exactly C(2d−1, d−1) proportional types, extending Dom Hans van der Laan's architectural system to four dimensions and beyond.
Deterministic Recall
Generation is recall, not prediction. If the system lacks a confident pathway, it returns nothing — by construction.

Traditional generative systems are trained to always produce output — even when uncertain, leading to hallucinations.

Our architecture treats generation as geometric recall: the system navigates a lattice of stored patterns and retrieves the nearest match.

If no pattern exists within the confidence threshold defined by σ-decay, the system returns ∅ (null) — silence is a valid, designed output. Hallucination is architecturally impossible because there is no generation pathway without a geometric address.

Key Insight: The ∅-return property is not a safety feature bolted on — it's an inherent mathematical consequence of recall-based architecture.
⟨⟩
Geometric Provenance
Every output traces to a specific geometric coordinate. The address is the audit trail.

Every stored pattern lives at a specific coordinate in the simplicial lattice — a geometric address expressed as a tuple ⟨layer, simplex, vertex⟩.

When the system recalls a pattern, it returns both the content and its lattice address. This address is immutable and deterministic: the same input always maps to the same coordinate.

Unlike attention-based systems where the "reasoning path" is a statistical artifact, geometric provenance is a mathematical proof of origin.

Key Insight: Auditability is not a feature we added — it's a consequence of the geometry. The lattice is the audit trail.

Open Access Research

Our research establishes the mathematical foundations of the xd = x + 1 hierarchy — the family of algebraic constants that organize energy propagation on simplicial lattices across dimensions.

P-05 · Zenodo · Open Access NEW

Betti Numbers as Cognitive Integrity Metrics: Topological Health Diagnostics for Deterministic Memory Systems

Casey Lee Race
August 31, 2026
DOI: 10.5281/zenodo.22205195
Read Abstract

We introduce the Cognitive Integrity Score (CIS), an operational, real-time topological health diagnostic for deterministic simplicial associative memory networks. We demonstrate that the sequence of Betti numbers (β0, β1, β2, β3, β4), computed via harmonic eigendecomposition of combinatorial Hodge Laplacians Δk = ∂kTk + ∂k+1k+1T, directly characterizes fundamental structural failure modes: β0 > 1 identifies concept space fragmentation; β1 >> 0 detects circular reasoning loops and associative trapping; and β2 >> 0 reveals relational voids (missing factual bridges). Monotonically gated with algebraic spectral connectivity (λgap), CIS achieves strong rank correlation (ρ = 0.9710, p < 10-37) with multi-hop retrieval accuracy, while boundary nilpotency (∂2 ≡ 0) enforces exact algebraic query refusal without statistical hallucination.

B-04 · Zenodo · Open Access

The Lattice Octave: A Characteristic-Delay Model and Growth-Rate Partition on Simplicial Lattices

Casey Lee Race
August 30, 2026
DOI: 10.5281/zenodo.22181883
Read Abstract

We prove that the algebraic polynomial family xd = x + 1 (for integer d ≥ 2), whose roots are known as generalized golden ratios, characterizes dual-channel boundary-bulk energy transport on d-dimensional simplicial lattices under a dominant characteristic-delay model. We formulate and prove the Face-Poset Delay Decomposition Theorem, establishing that the combinatorial face-poset topology of a regular d-simplex admits exactly two maximal transit channel classes: a boundary facet transit channel of characteristic delay d−1 hops, and a bulk interior transit channel of characteristic delay d hops. Under the dominant-delay approximation, the unique positive real root cd > 1 of xd = x + 1 induces the growth-rate partition identity cd−(d−1) + cd−d = 1, balancing asymptotic energy transport between boundary and bulk.

B-02 · Zenodo · Open Access

The xd = x + 1 Hierarchy: Cross-Dimensional Spectral Validation on Ad Root Lattices

Casey Lee Race
June 14, 2026
DOI: 10.5281/zenodo.20692936
Read Abstract

We prove that the polynomial family xd = x + 1 generates geometry-native organizing constants for Ad root lattices across dimensions — unifying the golden ratio (2D), plastic constant (3D), and σ-constant (4D) into a single hierarchy. Four independent lines of evidence: infinite-lattice Fourier analysis, algebraic recurrence, higher-order asymptotics with exact coefficients via Lagrange inversion, and a proof that the d-dimensional hyperbox subdivides into exactly C(2d−1, d−1) proportional types — extending Van der Laan's architectural system to four dimensions and beyond.

B-01 · Zenodo · Open Access

The σ-Constant: A Universal Algebraic Invariant for Energy Propagation in d-Dimensional Simplicial Lattices

Casey Lee Race
May 22, 2026
DOI: 10.5281/zenodo.20350425
Read Abstract

We derive σ ≈ 1.22074, the unique positive real root of x⁴ − x − 1 = 0, and demonstrate that it is the geometry-native organizing constant for energy propagation on the A₄ root lattice. The result is validated through two independent methods: Hopfield recall optimization and spectral Laplacian analysis. Zero free parameters. The lattice geometry determines the constant.

The Dimensional Hierarchy

Dimension Equation Constant Name Geometry
2 x² = x + 1 φ ≈ 1.618 Golden Ratio Triangle, Pentagon
3 x³ = x + 1 ρ ≈ 1.325 Plastic Constant Tetrahedron
4 x⁴ = x + 1 σ ≈ 1.221 σ-Constant Pentatope, A₄ Lattice
5 x⁵ = x + 1 ≈ 1.167 5-Simplex
→ 1.0 Isotropic Limit

The Lattice Letters

Mathematical research, historical ironies, theological reflections, and corporate position papers from our editorial pipeline.