Skip to main content

// Section 1.1 · Concepts

What ParalleliX Is

1 min1.1Concepts
// 2 of 4 · first principles

Orientation

// 1.1 · orientation

// Definition

A decentralised parallel-compute network.

Idle GPU and CPU capacity from independent operators worldwide, pooled. The live demand is ParalleliX AI: each inference request is dispatched whole to a capable node, with many requests running in parallel across the pool. Genuinely parallelizable workloads can also be segmented into sub-tasks and aggregated. That is the whole verb.

// Queue vs fan-out · two models

// centralised cloud

Queue model

One job lands on one box. Throughput is capped at the rented instance. Cost scales with rental time, not work done.

// one job · one box · one ceiling

// parallelix

Fan-out model

Each request routes to a capable node and runs whole. Many requests dispatch in parallel across the mesh. Throughput scales with the count of online nodes, not the instance.

// request · mesh · ceiling moves

The point lives in the contrast. The unit of work is the request, not the box, so capacity grows by adding nodes rather than by renting a larger instance.