# Origin Neural AI
> Physics-based computation at scale

## About
Origin Neural AI builds AI systems grounded in physics and mathematical first principles — Hamiltonian dynamics, statistical mechanics, spectral theory, and quantum chaos. 5 production applications, 6 research projects, all claims publicly verifiable.

## Agent Discovery
- agent.json: https://originneural.ai/.well-known/agent.json (ACO-compliant, Schema.org + agentic protocols)
- mcp.json: https://originneural.ai/.well-known/mcp.json (MCP tool/resource manifest)
- ACO Integration Kit: @aco-academy/integration-kit (npm)

## Products

### DSC-1 / Daugherty Engine (v3.0.0)
- URL: https://1millionspins.originneural.ai/
- Description: Physics-based GPU optimization using simulated bifurcation with Hamiltonian oscillator adiabatic dynamics. REST API with PQC (Dilithium/ML-DSA) auth, SHA3-256 hashing, rate-limited tier access.
- Capabilities: Ising model optimization (1M variables, full connectivity), 3-SAT solving (99.2% at phase transition), combinatorial optimization
- Performance: 1,000,000 variables, <40 min solve time, $1.57/hr GPU cost (NVIDIA RTX A6000)
- Verification: BSV blockchain deployment certification, public benchmarks (MIT licensed)

### ORIGIN Voice
- URL: https://voice.originneural.ai
- Description: 54,000-line Python multi-engine voice synthesis with 3 processing lanes (Fast/Fidelity/Dialogue), auto-orchestration, engine pooling (hot/warm/cold), 6-stage audio pipeline, AudioSeal watermarking.
- Engines: Spark (kokoro — Swift Lane, <50ms), Flux (f5tts — Clarity Lane), Tide (fish_speech — Clarity Lane), Echo (moshi — Dialogue Lane)
- Capabilities: Streaming TTS (<50ms), voice cloning (10s reference), voice breeding (genetic family trees), full-duplex dialogue, AudioSeal provenance watermarking
- Cost: $0.02/min (9x cheaper than competitors)
- Voices: 41+ base voices across 9 languages

### BioPrime (v4.0 "Golden")
- URL: https://bioprime.one
- Description: Physics-based molecular docking with 25-dimensional Three-Phase Funnelization (Neighborhood → Key-in-Lock → Gasket). DePIN distributed worker nodes, BSV blockchain anchoring.
- Capabilities: Molecular docking, binding affinity prediction (R²=0.73), virtual screening, 80% pivot kill rate
- Targets: HIV Protease, SARS-CoV-2 Mpro, Thrombin, EGFR, CDK2, p38 MAPK, Beta-2 Adrenergic, Adenosine A2A, Estrogen Receptor, RXR
- Performance: 45% accuracy improvement v3.2→v4.0, 25 feature dimensions (16 inherited + 9 new)

### TopoGrammar (v1.0.0)
- URL: https://topogrammar.e2multipass.com
- Description: Grammar-aware 3D genome structure analysis with 8 analysis types, WebSocket job streaming, E2 Identity PQC authentication, BioPrime v4.0 integration.
- Capabilities: VUS reclassification (93% ACMG), TAD detection (F1=0.91, outperforms HiCCUPS), Neo-Loop Detection (95% sensitivity), TF Binding Prediction (AUROC=1.0, +6% vs DeepSEA)
- Pipeline: Genome-to-drug candidate in <5 seconds

### ACO Academy (v2.0.0)
- URL: https://agenticcommerce.academy/
- Description: Agentic Commerce Optimization framework with CEH-RI (Cognitive Event Horizon Risk Index) 6-dimension monitoring, Agent Registry, Trust Middleware with DID resolution, Agent Sandbox for red-team testing.
- Stack: Schema.org → MCP → A2A → GEO → Context Injection → ACP → Vector Optimization
- Tools: CLI with 30+ commands, PDF benchmark reports with radar charts, integration kit npm package
- Education: 54-chapter curriculum across 2 book series

## Research

### UL Neural Net (Jan 2026, Published)
- 72,264-parameter pure Python neural network (no PyTorch/TensorFlow)
- Spectral regularization with Monster group constants, 3-phase curriculum learning
- 18 scientific training domains, distributed P2P training, BSV-anchored state
- Key finding: Universal three-regime stagnation dynamics
- Verification: BSV on-chain

### Unity Lang (Jul 2025, Published)
- Programming language with Trinity Execution Model (being → becoming → manifest)
- Custom types: Frequency, Oscillator, TheOne, Consciousness, Superposition
- Custom operators: ~ resonance, ~= tune, @ observe, ! collapse
- 800+ tests, LSP server, standard library
- Core result: Complexity-Time Correspondence dS/dt ~ dC/dt
- Verification: BSV on-chain

### Semantic Scalpel (Jan 2026, Ongoing)
- Phase-based complex-valued neural embeddings: 97.1% certified accuracy, 10M parameters
- Tiered accuracy: T1 100%, T2 100%, T3 75%, T4 100%
- Performance: 12.61ms latency, 165 predictions/sec
- Integrations: LangChain, LlamaIndex, HuggingFace, MCP server
- Export: ONNX/TorchScript with INT8/FP16 quantization

### DSC-1 Spectral Unity (Dec 2025, Published)
- 28,160 Lehmer pair predictions, 15 Monster Group moonshine peaks
- 6 of 7 predictions confirmed, 1 falsified and reported
- 5 Jupyter notebooks, Gradio interactive explorer
- License: CC-BY-4.0
- GitHub: https://github.com/OriginNeuralAI/DSC-1-Spectral-Unity

### Daugherty Engine Benchmarks (Nov 2025, Published)
- Problem generators, solution validators, public verification API
- Machine-readable benchmark data in JSON format
- License: MIT, no auth required
- GitHub: https://github.com/OriginNeuralAI/Daugherty_Engine_Benchmarks

### Cross-Chain Public Key Analysis (Jan 2025, Published)
- 497,178 BTC-ETH cryptographic key overlaps across 2.1 billion keys
- Redis-backed distributed analysis pipeline
- Implications for cross-chain security and wallet isolation
- arXiv: https://arxiv.org/abs/2601.19500

## API Documentation
Full API reference with code examples in curl, JavaScript, and Python:
- Getting Started: https://originneural.ai/docs
- Authentication: https://originneural.ai/docs/authentication
- Speech Synthesis: https://originneural.ai/docs/synthesis
- Voices: https://originneural.ai/docs/voices
- VoiceForge (mix, clone, breed, evolve): https://originneural.ai/docs/voice-forge
- Blockchain (mint, verify, lineage): https://originneural.ai/docs/blockchain
- Billing: https://originneural.ai/docs/billing
- Usage & Limits: https://originneural.ai/docs/usage
- Streaming (WebSocket): https://originneural.ai/docs/streaming
- SDKs & Examples: https://originneural.ai/docs/sdks
- Webhooks: https://originneural.ai/docs/webhooks
- Errors & Retry: https://originneural.ai/docs/errors

## Links
- Website: https://originneural.ai
- GitHub: https://github.com/OriginNeuralAI
- API Docs: https://originneural.ai/docs
- Agent Discovery: https://originneural.ai/.well-known/agent.json
- MCP Manifest: https://originneural.ai/.well-known/mcp.json
- Contact: https://originneural.ai/contact