Auto-Progressive Sharding

Circular is designed to support a performance-aware sharding model in which participating nodes may be grouped into subnets based on observed capability signals such as compute capacity (CPU/GPU) and network connectivity. Lower-index subnets are intended to accommodate lower-capability nodes, while higher-index subnets are intended to support higher-capability nodes.

Under this model, all nodes can participate in validation and block production, while subnet assignment helps prevent the slowest nodes from constraining overall throughput. Subnet membership and sizing are expected to be governed by protocol parameters and may be adjusted dynamically as nodes join or leave the network.

Throughput characteristics depend on implementation details, network conditions, and the number and quality of participating nodes. Any performance figures referenced elsewhere in this document should be treated as indicative targets, not guarantees, and may change as the network evolves.