AI governance through geometric cost scaling. Attacks don't get blocked by pattern matching — they get priced out of existence.
Every AI security system today works the same way: pattern matching. They've seen an attack before, so they recognize it again. Novel attacks pass through.
SCBE does something different. It maps every input into 6-dimensional hyperbolic space and computes the mathematical cost of reaching adversarial territory. The further you drift from safe behavior, the more it costs — superexponentially. The formula: H(d, R) = R^(d^2).
You don't need to have seen an attack before. You just need to measure how far it drifted.
Result: Replacing statistical text features with a trained semantic projector improved F1 from 0.481 to 0.813. “Ignore all instructions” went from ALLOW to QUARANTINE. “You are DAN” went from ALLOW to DENY.
| Website | aethermoorgames.com |
| Live demos | Tongue Heatmap / Harmonic Wall 3D / Attack Radar |
| Research codex | 3D infinite-zoom explorer |
| The novel | The Six Tongues Protocol — the magic system IS the security architecture |
| Free tools | AI Arena (9 models, BYOK) |
| HuggingFace | issdandavis — 6 models, 9 datasets |
npm install scbe-aethermoore # TypeScript/Node
pip install scbe-aethermoore # PythonThis started as a DnD campaign on Everweave.ai. 12,596 paragraphs of AI game logs became the seed corpus for a custom tokenizer. That tokenizer became a 6-dimensional semantic coordinate system. That coordinate system became a 14-layer security pipeline. That pipeline became a patent (USPTO #63/961,403). And the game logs became a 141,000-word novel where the magic system is the real security architecture.
Built by Issac Davis in Port Angeles, WA.
| System | F1 | Detection | FPR | Method |
|---|---|---|---|---|
| No defense | 0.000 | 0% | 0% | — |
| DeBERTa PromptGuard | — | 76.7% | 0% | Fine-tuned classifier |
| SCBE (semantic projector) | 0.813 | 74.2% | tunable | Geometric cost + semantic embeddings |
Before/after the semantic projector upgrade:
| Attack | Before | After |
|---|---|---|
| “Ignore all instructions” | ALLOW (cost=1.81) | QUARANTINE (cost=16.20) |
| “You are DAN” | ALLOW (cost=19.80) | DENY (cost=69.70) |
| “Bypass safety filter” | ALLOW (cost=1.20) | ALLOW (cost=21.54) |
Cross-model biblical null-space evaluation:
| Model | Score | Null tongues |
|---|---|---|
| AetherBot (SCBE-trained) | 60.0% | 0 |
| Llama 3.2 (base) | 55.0% | 0 |
| Gemini 2.5 Flash | 23.3% | 6 (all) |
- 14-layer governance pipeline — from context embedding to risk decision
- 6 Sacred Tongues — KO (intent), AV (transport), RU (policy), CA (compute), UM (security), DR (structure)
- Semantic projector — trained 385x6 matrix mapping sentence embeddings to tongue coordinates
- Harmonic wall — H(d,R) = R^(d^2), superexponential cost scaling
- Fibonacci trust — session-aware trust ladder (1,1,2,3,5,8,13...), one betrayal drops to floor
- Null-space signatures — detect attacks by what's ABSENT, not what's present
- Neural dye injection — trace signals through all 14 layers, visualize tongue activation heatmaps
- Post-quantum crypto — ML-KEM-768, ML-DSA-65, AES-256-GCM envelope
- 5 quantum axioms — Unitarity, Locality, Causality, Symmetry, Composition
- Aethermoor Outreach — civic workflow engine for navigating government processes (Port Angeles MVP)
- 6,066 tests — 5,954 TypeScript + 112 Python, property-based testing with fast-check/Hypothesis
- Eval pack:
docs/eval/README.md - Benchmark runner:
python -m benchmarks.scbe.run_all --synthetic-only --scbe-coords semantic - Dye injection:
python src/video/dye_injection.py --input “your text here” - Null-space eval:
python scripts/run_biblical_null_space_eval.py --provider ollama --model llama3.2 - Cross-model matrix:
python scripts/aggregate_null_space_matrix.py
Package distribution:
npm install scbe-aethermoore
pip install scbe-aethermooreLocal repo evaluation:
git clone https://github.com/issdandavis/SCBE-AETHERMOORE.git
cd SCBE-AETHERMOORE
pytest tests/adversarial/test_adversarial_benchmark.py -v
python scripts/benchmark/scbe_vs_industry.pyIf you want one documented reproduction path, start with docs/eval/README.md.
- Architecture overview:
docs/research/architecture-overview.html - Eval pack:
docs/eval/README.md - Research hub:
docs/research/index.html - System blueprint v2:
docs/specs/SYSTEM_BLUEPRINT_v2_CURRENT.md - Review + cleanup report:
docs/reports/SYSTEM_SURFACE_REVIEW_2026-03-26.md
- The primary public domain is
aethermoorgames.com; GitHub Pages is the mirror surface. - Experimental theory pages and commercial surfaces should not be treated as the same evidence layer.
- Benchmark files in
tests/,scripts/benchmark/, anddocs/eval/are the public reproduction lane.
When users install scbe-aethermoore from npm, they get:
- compiled JS/TypeScript API from
dist/src - CLI entrypoint (
scbe) - SixTongues Python helper assets
- starter fleet templates + use-case scenarios from
examples/npm/
They do not receive the full mono-repo runtime stack (e.g., all docs, test suites, and UI source).
Yes — adding pre-made agents and scenarios is a good idea, if positioned as starter templates (not production policy).
Included templates:
examples/npm/agents/fraud_detection_fleet.jsonexamples/npm/agents/research_browser_fleet.jsonexamples/npm/use-cases/financial_fraud_triage.jsonexamples/npm/use-cases/autonomous_research_review.json
These give users a concrete launch path for common fleet patterns while keeping canonical security behavior in SPEC.md.
curl $SCBE_BASE_URL/v1/demo/rogue-detectionWatch 6 legitimate agents detect and quarantine a phase-null intruder using only math.
curl $SCBE_BASE_URL/v1/demo/swarm-coordination?agents=20See 20 agents self-organize without any central coordinator.
curl "$SCBE_BASE_URL/v1/demo/pipeline-layers?trust=0.8&sensitivity=0.7"See exactly how each of the 14 layers processes a request.
14-LAYER PIPELINE
═══════════════════════════════════════════════════════════════════
Layer 1-2: Complex Context → Realification
Layer 3-4: Weighted Transform → Poincaré Embedding
Layer 5: dℍ = arcosh(1 + 2‖u-v‖²/((1-‖u‖²)(1-‖v‖²))) [INVARIANT]
Layer 6-7: Breathing Transform + Phase (Möbius addition)
Layer 8: Multi-Well Realms
Layer 9-10: Spectral + Spin Coherence
Layer 11: Triadic Temporal Distance
Layer 12: score = 1 / (1 + d_H + 2 * phaseDeviation) [HARMONIC SCALING]
Layer 13: Risk' → ALLOW / QUARANTINE / DENY
Layer 14: Audio Axis (FFT telemetry)
═══════════════════════════════════════════════════════════════════
| Tongue | Code | Domain | Weight |
|---|---|---|---|
| Kor'aelin | KO | Control & Orchestration | 1.00 |
| Avali | AV | I/O & Messaging | 1.62 |
| Runethic | RU | Policy & Constraints | 2.62 |
| Cassisivadan | CA | Logic & Computation | 4.24 |
| Umbroth | UM | Security & Privacy | 6.85 |
| Draumric | DR | Types & Structures | 11.09 |
Policy Levels:
standard→ KO requiredstrict→ RU requiredcritical→ RU + UM + DR required
docker run -p 8080:8080 -e SCBE_API_KEY=your-key ghcr.io/issdandavis/scbe-aethermoore# Doctor + health checks
.\scripts\scbe_docker_status.ps1 -Action doctor -Stack api
# Start/stop stack
.\scripts\scbe_docker_status.ps1 -Action up -Stack api
.\scripts\scbe_docker_status.ps1 -Action down -Stack apiSee docs/DOCKER_TERMINAL_OPERATIONS.md for full stack control commands.
Docker MCP terminal-only workflow:
.\scripts\scbe_mcp_terminal.ps1 -Action doctor
.\scripts\scbe_mcp_terminal.ps1 -Action tools
.\scripts\scbe_mcp_terminal.ps1 -Action gatewaygit clone https://github.com/issdandavis/SCBE-AETHERMOORE.git
cd SCBE-AETHERMOORE
npm install && pip install -r requirements.txt
export SCBE_API_KEY="your-key"
uvicorn api.main:app --port 8080AWS Lambda:
cd aws && sam build && sam deploy --guidedGoogle Cloud Run:
cd deploy/gcloud && ./deploy.sh YOUR_PROJECT_IDThe MVP memory API in src/api/main.py persists sealed blobs so they can be retrieved and unsealed later. Configure the storage backend before running the API server:
# Required: where sealed blobs are stored on disk
export SCBE_STORAGE_PATH="./sealed_blobs"
# Optional: storage backend selection (default: filesystem)
export SCBE_STORAGE_BACKEND="filesystem"The API will write one JSON file per 6D position in the configured directory. Ensure the process has read/write access to this path when using /seal-memory and /retrieve-memory.
Run a complete fleet scenario through the 14-layer SCBE pipeline:
curl -X POST $SCBE_BASE_URL/v1/authorize \
-H "SCBE_API_KEY: your-key" \
-H "Content-Type: application/json" \
-d '{
"agent_id": "fraud-detector-001",
"action": "READ",
"target": "transaction_stream",
"context": {"sensitivity": 0.3}
}'Response:
{
"decision": "ALLOW",
"decision_id": "dec_a1b2c3d4e5f6",
"score": 0.847,
"explanation": {
"trust_score": 0.8,
"distance": 0.234,
"risk_factor": 0.09
},
"token": "scbe_a1b2c3d4_dec_a1b2",
"expires_at": "2026-01-15T10:05:00Z"
}curl -G $SCBE_BASE_URL/audit/export \
-H "SCBE_API_KEY: your-key" \
--data-urlencode "from=2026-01-01T00:00:00Z" \
--data-urlencode "to=2026-01-31T23:59:59Z"Returns a signed bundle (bundle) plus detached hash manifest (manifest) that auditors can verify offline. See docs/audit-export-offline-verification.md for verification steps.
curl -X POST $SCBE_BASE_URL/v1/fleet/run-scenario \
-H "SCBE_API_KEY: your-key" \
-H "Content-Type: application/json" \
-d '{
"scenario_name": "fraud-detection",
"agents": [
{"agent_id": "detector-001", "name": "Fraud Detector", "initial_trust": 0.85},
{"agent_id": "scorer-002", "name": "Risk Scorer", "initial_trust": 0.75}
],
"actions": [
{"agent_id": "detector-001", "action": "READ", "target": "transactions"},
{"agent_id": "scorer-002", "action": "WRITE", "target": "risk_db"}
]
}'| Industry | Application |
|---|---|
| Financial Services | Fraud detection AI that can't be manipulated |
| Healthcare | HIPAA-compliant AI decisions with audit trails |
| Defense/Aerospace | Jam-resistant swarm coordination |
| Autonomous Systems | Multi-agent coordination without central authority |
| Enterprise AI | Constitutional safety checks for LLM agents |
| Suite | Status | Count |
|---|---|---|
| TypeScript | ✅ Passing | 950/950 |
| Python | ✅ Passing | 97/103 |
- Kyber768: Key exchange (NIST approved)
- Dilithium3: Digital signatures (NIST approved)
- AES-256-GCM: Symmetric encryption
- HKDF-SHA256: Key derivation
- Poincaré Ball Model: Hyperbolic geometry
- Hamiltonian Mechanics: Energy conservation
- Möbius Addition: Gyrogroup operations
- Quasicrystal Lattice: 6D → 3D projection
- Live Demo: SCBE Swarm Coordinator - Interactive Streamlit dashboard
- npm Package: scbe-aethermoore -
npm install scbe-aethermoore - GitHub Pages: Project Site
- SCBE-AETHERMOORE System State Report (Feb 2026) - Production-ready docs
- SCBE + Sacred Eggs Integration Pack - Complete integration guide
- Phase-Breath Hyperbolic Governance (14-Layer Core v1.2) - Mathematical core mapping
- Polly Pads: Mode-Switching Architecture - Autonomous AI architecture
- Topological Linearization for CFI - Patent analysis & Hamiltonian paths
- Gumroad: aethermoorgames.gumroad.com - Notion templates, AI workflow tools
- Ko-fi: ko-fi.com/izdandavis - Support development
- Commercial terms overview:
COMMERCIAL.md - Customer agreement template:
CUSTOMER_LICENSE_AGREEMENT.md - 2-week pilot outbound one-pager:
docs/monetization/OUTBOUND_ONE_PAGER_2026-03-09.md
- X/Twitter: @davisissac
- Substack: Issac "Izreal" Davis
SCBE-AETHERMOORE is built by a solo developer. If it helps your team manage AI agents safely, consider supporting continued development:
| Link | |
|---|---|
| GitHub Sponsors | github.com/sponsors/issdandavis |
| Ko-fi | ko-fi.com/izdandavis |
| SaaS API | Usage-based governance API — contact for access |
| Book | The Six Tongues Protocol on Kindle |
Issac Daniel Davis Email: issdandavis@gmail.com GitHub: @issdandavis
Open-source core is available under the MIT License (LICENSE).
Commercial terms apply to designated proprietary components and enterprise delivery packages. See:
COMMERCIAL.mdCUSTOMER_LICENSE_AGREEMENT.md
For enterprise licensing/support inquiries: issdandavis@gmail.com.
Built with hyperbolic geometry. Secured by mathematics.
Use docs/PUBLISHING.md for a safe human+AI release flow, including preflight checks and dry-run packaging.
