Quasi-Voxel Ternary Storage
90KB technical specification on non-binary data storage in hyperbolic security manifolds.
View Spec →Public research layer
Keep the product page lean. Put the proof here.
This domain layer is for benchmark summaries, mathematical framing, and active experiment tracks. It separates what is public, what is replicated, and what is still exploratory.
Benchmark board
Current public signals
The board below is the readable surface. The repo remains the audit surface. Keep the run timestamps visible so newer evidence does not get mixed with older summary claims.
03 / 31
March 31 evidence snapshot
Updated with overnight benchmark suite: spectral sweep (8 configs), null-space ablation (A/B/C), industry comparison (vs ProtectAI), and 233K multi-view training dataset.
85.7% detection, 0% FP (E4 config)
Spectral sweep best-balanced: semantic tongue coordinates + triple-weight remainder on 91-attack corpus.
Full methodology
91 / 91 vs ProtectAI 62 / 91
Industry comparison: SCBE blocked all attacks with zero false positives.
F1: 0.481 → 0.813
Semantic projector orientation improved F1 by 69% on 260-sample adversarial benchmark. Anchor result for pump architecture.
504-bit trichromatic state
IR + Visible + UV hidden-band governance. 5/5 forgeries caught even when all visible signals matched.
233K multi-view SFT pairs
Binary-first training stack: L0 substrate (9%) + L1 coordination (11%) + L2 orientation (31%) + L3 expression (49%).
HuggingFace
Confirmed output
Strict isolation training run
Latest Kaggle plus local SFT pipeline snapshot, with the blind holdout left outside training.
Overlap harness
Hybrid scorecard
Shows what the classifier, gate, and trichromatic layer each contribute on the same attack set.
Structural packet
Trichromatic governance
IR, visible, and UV hidden-band diagnostics presented as measured structure, not finished storefront claims.
Interaction map
Where the variables couple
Explains the benchmark split and the real decision path through L3, L7, L12, and L13.
Measured runs
Public entry points for the newest evidence
These are the March 30 pages that turn local logs and JSON artifacts into readable public research surfaces without hiding the weak spots.
Strict isolation training run
Kaggle plus local SFT plus benign docs plus a blind 20-category holdout, kept separate on purpose so the benchmark still means something.
Hybrid overlap scorecard
Classifier, RuntimeGate, and trichromatic compared on the same benchmark set, with misses and overlap cells left visible.
Trichromatic governance
IR, visible, and UV bands modeled as one structural signal, with measured energy separation and a 504-bit combinatorial packet.
Case studies
Applied research with measured results
These pages apply the 14-layer pipeline to real multi-agent architectures and adversarial scenarios, showing where the math provides measurable protection.
Deep Archive
The Hard Math of Aethermoor.
Ground-level specifications, ternary logic, and architectural proofs.
Null-Space Hypothesis
Bridging the gap between the Seven Sacred Codices and triadic harmonic security math.
Read Hypothesis →DARPA Security Standards
SCBE's alignment with 2026 DARPA AI security programs and global governance benchmarks.
See Standards →Active tracks
Exploratory tracks still worth growing
Keep these separated from the measured pages above. They are still useful, but they are research tracks rather than the latest public evidence packet.
History + Bible theory packet
Bounded training packet built from the local theory corpus around witness, covenant, genesis control, memory, and sanctuary-building. Research language stays separated from implemented claims.
Context embedding lab
Dual-space retrieval, tongue compression, and 21D state comparison. Best next step is replacing synthetic embeddings with a real encoder baseline.
Flux trail
A-to-Z trajectories instead of A-to-B endpoints. Good for showing where embeddings oscillate, reverse, and settle.
Spectral telescope
Frequency, amplitude, coherence, spin, tongue dominance, and settling as six views on the same underlying signal.
Math stack
Public entry points into the harder material
Link people from summaries into the real documents rather than trying to make the landing page carry the full weight.
AetherBrain / PHDM framework
Two-layer intelligence model, deterministic control shell, and hyperbolic permission space.
Langues weighting system
The six sacred tongues, phi-weighting, and semantic decomposition.
Spectral map
Good bridge document between symbolic language, signal view, and experimental output framing.
Skill vault
Sphere Grid skill map
A 3D navigation surface for the Codex skill library. Click spheres to open the vault notes and then copy an invocation prompt for local Llama (Ollama) or Hugging Face inference.