Skip to main content
ParalleliX Docs

// Section 1.2 · Concepts

The Two-Layer Model

4 min1.2Concepts
// 4 of 4 · first principles

Compute and settlement

// 1.2 · compute + settlement

The network and the token are easy to conflate. Two layers, two jobs, two failure modes. Reading this section once makes the rest of the document stop sounding contradictory.

// Two-layer model

// layer 01 · compute

Compute network

Operators · coordinator · task lifecycle · scheduling · validation

// paycredits + stake → contract
// proofPoE commitment → on-chain anchor

// layer 02 · settlement

$PRLX settlement layer

ERC-20 on Ethereum · AI credits · operator rewards · stake

0x93ff39…f893

// Cross-layer seams · exactly two

everything else · off-chain

  • // seam 01
    Node stake + rewards

    An operator locks $PRLX on-chain in NodeRegistryLocker to register a node, and claims uptime rewards from the same contract. Registration is permissionless; principal is always returned on unstake. There is no slashing.

  • // seam 02
    ParalleliX AI credits

    A ParalleliX AI user deposits $PRLX on-chain once to fund a credit balance, then spends it off-chain per inference request. This is the demand side at launch.

Note

The two layers connect at exactly two points. Every other interaction is off-chain.

Comparable architectures

// 1.2 · comparable architectures

Two reference points orient the reader. ParalleliX shares the two-layer pattern with both and diverges on two axes: what the network does (general compute, not storage or rendering only) and how work is routed (request-level fan-out across the node pool, with whole AI inference requests served by one node and parallelizable workloads split into sub-tasks, not single-machine assignment only).

  • // ref 01$FIL

    Filecoin

    Off-chain storage providers · stake FIL as collateral

  • // ref 02$RNDR

    Render Network

    Off-chain GPU render nodes · render-only workloads

Full 5-project comparison