Technology

How Aethelred works.

Deterministic settlement meets enclave-native inference. Two planes, one protocol, continuous evidence.

Architecture

Four layers. Independent scaling.

Layer 4Settlement

CometBFT consensus with deterministic finality. Proof references, Digital Seals, and token transfers settle with cryptographic binding to block height and validator set. No probabilistic confirmation windows.

Layer 3Verification

Hybrid TEE attestation combined with zkML proofs. Five proof systems (Groth16, PLONK, EZKL, Halo2, STARK) verified through EVM precompiles. Mismatched evidence fails the operation. No silent normalization.

Layer 2Execution

Confidential compute inside attested hardware enclaves. Multi-platform support: Intel SGX, AWS Nitro, AMD SEV-SNP, Azure Confidential VMs, and Google Confidential VMs. Each produces attestation quotes bound to the workload.

Layer 1Consensus

Proof-of-Useful-Work. AI inference jobs are submitted, VRF-scheduled to validators, executed in TEE backends, and evidence is aggregated via ABCI++ Vote Extensions. Rewards scale with utility category.

Verification

How inference is verified.

Hybrid TEE attestation combined with zero-knowledge proofs. Dual evidence required per job — no silent normalization.

Zero-knowledge machine learning verification
Consensus Proof of Useful Work

Validator rewards tied to real AI inference, not wasteful hash computation. VRF-scheduled across the validator set.

Verification TEE + zkML

Hardware enclave attestation paired with zero-knowledge proof systems. Dual evidence required per job.

Security posture Post-quantum from genesis

ML-DSA-65 signatures (FIPS 204) and ML-KEM-768 key exchange (FIPS 203). No migration required.

Node Operations

Validator admission standards.

Institutional-grade infrastructure requirements for consensus participation and inference execution.

Server hardware for validator operations
RequirementBaselineNotes
Minimum stake100,000 AETHEL (design target)Subject to governance adjustment
TEE hardwareSGX DCAP, Nitro, or SEV-SNPAttestation quotes required per job
Key custodyHSM (FIPS 140-2 L3 path)Thales Luna or Fortanix DSM
GPU classH100, MI300, or equivalentWorker node inference execution
Uptime target99.5%Slashing for extended downtime
Sentry nodes2+ recommendedPublic ingress boundary, no signing keys
Resources

Go deeper.

Protocol specifications, developer tooling, and operator documentation.

Hardware infrastructure detail

Whitepaper

Full protocol specification including consensus design, verification architecture, and security posture.

Developer documentation

API reference, SDK quickstarts, testnet configuration, and deployment guides for building on Aethelred.

Node operations

Hardware requirements, node roles, and operator onboarding for validators, workers, and sentry nodes.