Skip to content
ExperimentalTestnet: see pageMainnet: Not availableSoftware: Core bad4a06ea · protocol pqwerai-v8-testnet/1Updated

PATE — adaptive trust execution and shadow fast finality

PATE (PASE Adaptive Trust Execution) is a research track built on top of the PASE parallel execution engine. It has three phases. Only phase 3 runs on the test network, and only in shadow mode.

Shadow finality is not finality

On testnet 20777 a transaction is irreversible only when the real Smartnode checkpoint that covers it is finalized (see Finality for applications). The shadow signal described below is provisional and measured for research. Wallets, exchanges, bridges and the fork choice must not rely on it.

At a glance

Phase What it does Status Enabled on 20777
1 — adaptive trust scheduling Node-local, advisory scheduler: transactions that are likely to conflict go straight to the serial path, so PASE wastes less work on re-execution Tested locally No (flag off by default)
2 — smart transactions Opt-in recurring payments through an explicit RecurringPermissions contract, execution templates and an adaptive TrustGuard Tested locally No (flags off; contract not deployed, needs an independent audit)
3 — shadow fast finality Validators sign an extra, provisional finality link on a short interval next to the real 32-block checkpoints, so faster finality can be measured without changing consensus Live on testnet in shadow mode Shadow switch on the seed2 validators only

The authoritative, dated status of each phase is on the Feature status page.

Phase 1 — adaptive trust scheduling

  • Purpose: reduce wasted parallel work. PASE executes independent transactions in parallel and re-executes on conflicts; PATE keeps a node-local reputation of conflict behaviour and routes predictable conflicts to the serial path first.
  • Consensus impact: none. The scheduler is advisory: block contents, ordering, state roots and validity are identical with PATE on or off. It never affects votes, weights, quorum or fork choice.
  • Configuration: a node flag, off by default. It is not set on the public seeds.
  • Limits: results are local benchmarks on one host. No throughput figure is claimed for the running network.

Phase 2 — smart transactions

  • Purpose: let a user authorise a recurring payment explicitly (amount cap, interval, expiry, revocation) instead of signing every instance.
  • Pieces: the RecurringPermissions contract, a smart-transaction engine, execution templates, the adaptive TrustGuard, and an optional wallet Recurring page.
  • Status: all flags off. The contract is not deployed on any public network; an independent contract audit is required before any deployment.

Phase 3 — shadow fast finality

Today, a transaction on 20777 becomes irreversible 32–66 blocks after inclusion (measured median about 51 blocks, roughly 8 minutes), because Casper-FFG-style finality needs two consecutive justified checkpoints 32 blocks apart.

Phase 3 studies block-level faster finality: keep GhostRider proof-of-work ordering, the existing Ed25519 Smartnode votes, the more-than-two-thirds quorum and slashing rules, but justify checkpoints on a shorter interval.

  • Transaction-level instant finality is not feasible without moving ordering authority from miners to validators, which would be a different consensus model. PATE does not do that.
  • Shadow mode (enabled on the seed2 validators since 16 November 2026): validators sign the shorter-interval links as well, and the node reports a provisional shadow finality height (ff_status) next to the real one. Nothing in block validity, fork choice or state depends on it. The first measurements showed shadow finality around 10 blocks versus about 42 blocks for real finality; these are early research samples, not a guarantee.
  • Fixed issue: a late signer could stall its shadow votes (finding FFSHADOW-LATE-SIGNER-STARVED). The fix (Core 987d727fb) runs on seed1 and seed2 since 16 November 2026; all four signers sign shadow votes.
  • Activation: turning shorter-interval finality into real consensus finality is a hard fork with a new profile commitment. It needs a decentralised validator set, the stage-aware slashing decision and an independent review. It is not scheduled.

Trust assumption

Real finality on 20777 rests on four genesis validators run by one operator. Any faster finality inherits that assumption; decentralising the validator set matters more than latency.

Source documents

  • Phase 1 design: architecture/PATE_ADAPTIVE_TRUST_EXECUTION.md
  • Phase 2 design: architecture/PATE_PHASE2_SMART_TRANSACTIONS.md
  • Phase 3 audit, design and shadow-mode prototype: architecture/PATE_PHASE3_FAST_FINALITY.md

These files are in the project repository; results and rollout records are kept with the project evidence.


Found a problem? Report a documentation issue · Build testnet-rc1+35ec5bd0b5