RADIX WikiRADIX Wiki

Overview

Radix's consensus mechanism has evolved through multiple generations, each building on lessons learned:

eMunie (2013–2017)

Dan Hughes' earliest experiments with distributed ledger design. Explored various approaches to achieving consensus without a single chain.

Tempo (2017–2019)

Tempo was a DAG-based consensus mechanism that used logical clocks and gossip protocols. While innovative, Tempo had limitations in providing the strong finality guarantees needed for DeFi.

Cerberus (2020–present)

Cerberus replaced Tempo with a formal BFT protocol using braided cross-shard consensus. The whitepaper was published in 2020 and validated through the Hyperscale tests. In practice, the live Babylon mainnet runs Cerberus in a simplified unsharded form (a single shard group), a variant of the original HotStuff BFT; the fully sharded, multi-shard Cerberus design has not yet shipped.

Xi'an (2027 target)

Xi'an is Radix's sharded mainnet upgrade, intended to deliver the full Radix Engine sharding and linear scalability that has been Radix's goal since inception. Its original consensus design was the fully sharded Cerberus protocol, but the current production candidate (the community hyperscale-rs implementation, subject of the April 2026 Xi'an RFC) does not implement Cerberus, using a HotStuff-2–derived per-shard consensus instead – the later two-chain commit variant, distinct from the original HotStuff that Babylon runs today.

Two clarifications on the lineage. First, braiding has never run in production. Babylon runs Cerberus unsharded, where there is nothing to braid; the Radix Foundation's Hyperscale implementation, according to the lead developer of the Xi'an candidate, "never really used Cerberus"; and on 2 August 2026 he added that the original Java Hyperscale did not braid either — Dan Hughes having replaced it with an execution process asynchronous to consensus, which hyperscale-rs continues. The stated objection is liveness contagion: braiding makes shards co-dependent for liveness, so a single stalled shard can impede others, whereas asynchronous execution lets a down counterparty shard time out only the transactions that touch it.

Second, Xi'an's execution layer is changing along with its consensus. On 1 August 2026 the lead developer confirmed that a purpose-built virtual machine is underway to replace the Radix Engine, on the reasoning that the sharding adjustments required "would require touching everything" and that a VM able to resolve a transaction's data dependencies deterministically in advance is a precondition for the cross-shard commit protocol. Existing dApps are expected to need a recompile at best and a source-level transpiler at worst.