What finality means
Deterministic vs. probabilistic finality. On chains like Ethereum, finality is probabilistic: the longer you wait for confirmations, the lower (but nonzero) the chance of a reorg. You choose a confirmation count (6, 12, 32) based on your risk tolerance. On the Grid, finality is deterministic. Once an activation reaches thesuccess status, the network has reached Byzantine consensus that it will never be reverted, regardless of how many blocks have passed since. You do not count blocks. You do not have reorg risk.
Two settlement modes. the Grid offers two paths:
- Optimistic acknowledgement: under ~50 ms. The leader cluster receives and accepts your activation for immediate execution. This fits inside a normal HTTP round-trip, so an agent can submit a payment and continue in the same request.
- Deterministic finality: 400–800 ms. The activation has completed execution, replicated to follower clusters, passed Byzantine consensus, and is final.
How the Grid delivers it
Grid finality is built on two layers: GridBFT (consensus) and GridSMR (replication and execution).GridBFT: Consensus
GridBFT is a leader-based Byzantine fault tolerance protocol. Blocks are proposed every 10 milliseconds. The protocol is pipelined: multiple blocks move through different stages of consensus simultaneously, so block proposal is not gated on the previous block’s full finalization. This keeps the pipeline full and latency low. Finality is global, not per-shard. An activation is final when the entire network reaches consensus, not when a single shard agrees.GridSMR: Execution and replication
The leader cluster executes activations immediately (at execution speed, not consensus speed) and propagates resulting cross-shard continuations. It produces an optimistic execution view of the next block and sends that proposal to follower clusters. Followers independently reproduce and verify that execution. Only after verification does GridBFT finalize the replicated state. Why this matters: consensus is removed from the critical path of execution. The system reacts at execution speed and only commits what replicates correctly. Causal compression. When execution on one shard produces a continuation for another shard, the entire cross-shard causal chain finalizes in a single consensus round. Not one round per shard hop. A multi-hop workflow stays as fast as a single-shard one.What this means for your client
Waiting for finality
There are two ways to wait for an activation:-
Poll for deterministic finality:
Repeat until
statusissuccess(final and applied) orfailed(final and rejected). This is the settlement-grade path. - Use optimistic acknowledgement: The activation submission RPC returns once the leader cluster has accepted the activation. Use this if your flow can tolerate the small risk that the activation might not finalize (rare but possible). Deterministic finality follows shortly after.
Status values
pending: the activation is in the leader’s queue or is executing.success: the activation has executed successfully and is deterministically final.failed: the activation has been rejected and is deterministically final. Check the error details to see why (e.g., balance too low, delegation exceeded).
No confirmation-count pattern
Do not wait N blocks. Do not count blocks. Poll forstatus: success or failed.
Next steps
- Activations: the unit of work you submit.
- RPC overview: how to call the API.
- CLI reference: activations: CLI polling and status.
- The Grid for agent builders: the higher-level context.

