The razor between rigorous science and the unknown.
The world's first dual-node, cryptographically verifiable paranormal investigation platform. SPECTER doesn't just measure anomalies — it benchmarks them against certified Quantum Random Number Generator (QRNG) streams using Bell-like statistical tests, cryptographically signs every data block at acquisition, and requires OSF preregistration before investigation begins. Every reported anomaly must survive adversarial scrutiny before it's called one.
Most paranormal investigation apps use basic EMF meters with no scientific framework and no controls. SPECTER applies real experimental methodology:
| Feature | What It Does |
|---|---|
| Dual-Node Architecture | Active measurement node + Null control node running in parallel — eliminates investigator-site contamination |
| HSM Cryptographic Signing | Every data block signed with ECC P-256 via Microchip ATECC608 — tamper-evident, forensically verifiable |
| OSF Preregistration | Hypotheses, thresholds, and analysis pipelines locked before investigation begins — prevents post-hoc bias |
| Bell Inequality Testing | CHSH-style statistical tests on sensor triplets — flags non-classical correlations with Monte Carlo p-values |
| QRNG Benchmarking | Compares environmental entropy against certified quantum random streams (ANU, IDQ Quantis) |
| EVP Validation Pipeline | Blinded segmentation, phonotactic likelihood testing, adversarial transforms — rejects audio pareidolia |
| Environmental Impairment Index | CO + infrasound monitoring auto-pauses investigation when perceptual impairment risk is elevated |
| Automatic Null Publication | Null results published alongside positive findings — eliminates publication bias |
The claim is defensible: "We report what the statistics cannot explain — after eliminating everything we can."
[Active Node A] ---(BLE/WiFi)---> [Mobile App] <---(BLE/WiFi)--- [Null Node B]
\ | /
\ [Echo AI Analysis] /
\-----------(optional cloud backup)------------------/
|
[PostgreSQL + TimescaleDB]
[Immutable session records]
[Public null dataset]
Time sync: GPS PPS + PTP — inter-node skew ≤10 ms
Security: Microchip ATECC608 HSM — ECC P-256 per block — immutable hash chain per session
| Screen | Function |
|---|---|
| HOME | Quantum HUD — live sensor readings, Bell S-value, QRNG deviation, anomaly severity |
| SCAN | 200-particle quantum visualizer, Bell inequality meter, environmental sensor grid, dual-node differential |
| SPIRIT | FM sweep (87.5–108 MHz) with full Web Audio API engine — bandpass filter tracking, sawtooth voice bursts on contact |
| ECHO | Llama 3.3 70B AI — full session visibility, metric interpretations, cross-modal correlations. Never asserts paranormal causation. |
| EVP | Full scientific validation pipeline: blinded segmentation, phonotactic testing, entropy/compressibility metrics, adversarial transforms |
| HUD | Comprehensive quantum metrics: Bell S-value, QRNG z-score, LR score, p-value, dual-node differential |
| LOG | Complete session event history with HSM signatures and OSF session IDs |
Most apps play back audio and call patterns "voices." SPECTER's EVP pipeline:
- Blind segmentation — 2–3s windows mixed with controls, labeled without investigator input
- Phonotactic likelihood testing — phoneme probability vs shuffled audio, white noise, QRNG-modulated baselines (p < 0.01)
- Entropy & compressibility — Lempel-Ziv, sample entropy, spectral flatness vs noise distribution
- Dual-node differential — candidates must appear exclusively/significantly on Active node vs Null node
- Multimodal time-locking — ±200 ms correlation with EMF, RF, vibration events
- Adversarial transforms — reversal, pitch shift, time-stretch — semantic persistence under transformation reduces confidence
- Pre-registered constraints — phrase length, phoneme classes, thresholds locked before investigation
Every completed investigation produces a signed bundle:
{
"session_id": "SPECTER-2026-001",
"osf_id": "preregistered",
"lr_score": 4.72,
"p_value": 0.003,
"bell_s_value": 1.87,
"qrng_z_score": 2.14,
"calibration": "NIST-traceable",
"signature": "ECC-P256:verified"
}Echo is powered by Llama 3.3 70B running on Groq with a live system prompt built from your actual session data every message:
- Current EMF, temperature, pressure, motion readings
- Bell S-value and anomaly count by severity
- Spirit Box sweep status and all logged contacts (word, frequency, entropy %, timestamp)
- Full conversation history (last 18 messages)
Echo reasons across all screens simultaneously — correlating Spirit Box contacts with simultaneous EMF spikes and explaining what the quantum statistics actually say.
Echo never asserts paranormal causation. All interpretations are probabilistic and grounded in statistical pipeline outputs.
API key management: The Groq API key is stored securely server-side. Never exposed to the client, never in localStorage, never in this repository.
- Theme: Deep void black with animated purple/teal nebula gradients
- Effects: CRT scanline overlay, neon glow system (teal nominal · purple quantum · red anomalies · amber warnings)
- Cards: Glass morphism with light streaks
- Typography: Orbitron (brand) · Share Tech Mono (data) · Rajdhani (labels)
- Mobile-first: 70px bottom navigation, 48px minimum touch targets, keyboard-safe layouts
Frontend: React 18 + TypeScript
Audio: Web Audio API (OscillatorNode, BiquadFilterNode, GainNode, LFO)
AI: Llama 3.3 70B via Groq API (server-side proxied)
Database: PostgreSQL + TimescaleDB (Phase 2)
Security: ATECC608 HSM + ECC P-256 (hardware tier)
Visuals: Canvas API, CSS animations
Icons: Lucide React
- Quantum benchmarking engine (Bell test, QRNG comparison)
- Echo AI with full session context
- Spirit Box with Web Audio API
- EVP entropy analysis
- SPECTER white paper published
- Dual-node hardware reference implementation
- HSM signing integration (ATECC608)
- OSF preregistration API integration
- Real QRNG API integration (ANU, IDQ Quantis)
- GPS-tagged sessions with heatmap export
- Multi-device team investigation sync
- Exportable PDF investigation reports
SPECTER: Ethereal Insights — Technical Design & White Paper
Full specification: dual-node architecture, HSM security model, complete statistical pipeline, EVP scientific validation layer, threat model, and roadmap.
Released under the Or4cl3 Open Model License (OOML) v1.0 — open like MIT, protective like GPL.
Part of the Or4cl3 AI Solutions ecosystem — 270+ repositories, 6 published books, pioneering intrinsic AI ethics.
- 🌐 or4cl3-ai-1.github.io/consulting — AI Governance & EU AI Act consulting
- 📚 github.com/or4cl3-ai-1/or4cl3-research — Complete IP archive
- 🏠 github.com/or4cl3-ai-1 — Full ecosystem (43 repositories)
"Code is not just logic; it is a performance." — Dustin Groves, Or4cl3 AI Solutions