Skip to content
BitcoinAI — The AI Blockchain for Bitcoin
GitHub (opens in a new tab)White Paper
BitcoinAI — The AI Blockchain for Bitcoin

AI

AI architecture

The deeper execution and routing model behind decentralized AI jobs.

AI for people and applications

Use AI services without building around one central provider.

BitcoinAI is being designed so applications can request AI work, coordinate providers, verify evidence and settle services through open infrastructure while the heavy model execution stays where compute is efficient.

Temporary Unsplash placeholder · Photo by Fiqih Alfarish

Scale without replicating the model

Do not make every validator run every AI model.

Planned

Wrong architecture

Every validator runs every model

Every validator would process

  • Large prompts
  • Model weights
  • Images
  • Video
  • Tensors
  • Embeddings
  • Retrieval data
  • Large outputs
  • Model logs
  • Consensus would become expensive
  • Validators would need specialized GPUs
  • Hardware requirements would rise dramatically
  • Throughput would collapse
  • Proprietary model APIs could not be reproduced locally

BitcoinAI architecture

Providers execute → validators verify compact evidence

  1. 01AI compute
  2. 02BatteryAGI / provider network
  3. 03Compact result + proof receipt
  4. 04BitcoinAI verification

BitcoinAI validators do not need GPUs unless the same operator also chooses to provide AI services.

Pay before execution — release after verification

Escrow separates user authorization from provider settlement.

Planned
  1. 01User / agent
  2. 02AI request
  3. 03BatteryAGI quote
  4. 04BATT-denominated service cost
  5. 05BTCAI / supported-asset equivalent
  6. 06Escrow lock
  7. 07AI_JOB_ID
  • The service may be priced internally in BATT.
  • The user can fund through BTCAI or another supported payment asset where implemented.
  • The user locks value rather than transferring it immediately.
  • The job receives a unique AI_JOB_ID.
  • Successful verification triggers settlement.
  • Failure or timeout can trigger refund or cancellation under final rules.
Pricing oracle / liquidity routing — implementation-dependentEngineering

No exchange-rate oracle mechanics are specified here.

Multi-model intelligence

Fusion AI combines specialized strengths instead of relying on one model.

Planned

Parallel Consensus: Several models answer independently, then the answers are compared, scored and synthesized. Nodes: Model A, Model B, Model C. Steps: Compare, then Score, then Synthesize.

General Fusion AI pipeline
  1. Prompt
  2. ├─ Model A
  3. ├─ Model B
  4. ├─ Specialist C
  5. └─ Retriever
  6. ↓ Evaluation / critic
  7. ↓ Fusion / synthesis
  8. ↓ Final result
Fusion AI is not simple averaging.
AI infrastructure supports researchers and engineers working with real scientific and technical systems.Photo by ThisisEngineering · Unsplash

AI execution layer

Request on BitcoinAI. Execute on BatteryAGI. Verify on BitcoinAI. Settle in BATT.

  1. 1

    BitcoinAI

    AI request

    An app, user or agent submits a job with BTCAI-escrowed terms.

  2. 2

    BatteryAGI

    Model routing / Fusion AI

    BatteryAGI routes the job to the right model or ensemble.

  3. 3

    BatteryAGI → BitcoinAI

    Proof-of-AI receipt

    A signed receipt commits to inputs, model and output.

  4. 4

    BitcoinAI · BATT

    Verify + settle

    The chain verifies the receipt and settles the service in BATT.

Model-agnostic

  • OpenAI-class APIs
  • DeepSeek
  • Open-weight models
  • Local enterprise models
  • Specialist models
  • Future providers

Model-agnostic by design. Named providers illustrate compatibility; integrations are announced only when live.

Qualifying BATT AI-service fees

Canonical
90%AI Providers
10%Validators + Delegators

Real AI usage can strengthen validator economics without requiring additional BTCAI inflation.

Explore AI & BatteryAGI