AI PROVIDERS
Provides models, inference or AI execution.
Users define the workload and required resources. Providers compete to execute it.
01 · WHAT THIS PROVIDER DOES
What this provider does
An AI provider supplies an executed output. A compute provider supplies the capacity it runs on.
GPU computation, rendering and specialized computing hardware are the compute provider's, not this one's.
02 · WHAT IT SUPPLIES
What it supplies
Possible workloads
- inference
- agent execution
- model serving
- batch processing
- embeddings
- data processing
03 · REQUIREMENTSSpecification pending
Requirements
Indicative until specified. These are the categories an AI provider is expected to satisfy, not the values it must meet.
- Models
- Which models or capabilities are offered and how they are described.
- Availability
- The uptime a provider commits to for the duration of an order.
- Result verification
- How an output is shown to be the output ordered.
- Collateral
- What is staked against the delivery a provider accepts.
Requirements will be published with the provider framework.
04 · HOW IT IS PAID
How it is paid
AI execution fees are one of the protocol's fee sources. A provider is paid for the execution it performs, settled against the verified result.
05 · HOW IT IS MEASURED
How it is measured
Compute settles against the work performed. AI settles against the result.
| Market | Settled against |
|---|---|
| Compute | completed workload / verified result |
| AI | execution receipt / result verification |
VerifiedIllustrative
order 0x9c31…a70f
provider 0x2a91…8e03
work delivered 1,000,000 tokens · ai-inference
proof execution-receipt · verified
settled 9.40 USDTFig. 01 — execution receipt
Beyond settlement, provider reputation is built from measurable information such as
- successful executions
- response time
- delivery accuracy
- dispute history
- execution quality
Applications can automatically choose providers according to their own requirements.
06 · RISKS AND OBLIGATIONSSpecification pending
Risks and obligations
A provider commits model availability for the duration of an order and collateral against the delivery it accepts. Settlement follows the verified result, and a provider whose result fails verification is exposed to economic penalties where appropriate.
The penalty mechanism is not designed.
- model availability committed
- collateral at stake
- penalties where appropriate
07 · ROADMAP DEPENDENCY
Roadmap dependency
Phase 4
Compute & AI Marketplanned
create an open marketplace for AI and computational infrastructure
- GPU providers
- compute orders
- workload routing
- AI inference market
- execution receipts
- provider benchmarking
- compute APIs
AI inference market and execution receipts are named deliverables of that phase.
Related