One validator proposes. The network agrees. Execution keeps moving.
Follow a block from its proposer’s slot to its place in the chain. Helicon lets consensus and the EVM work concurrently, so agreeing on the next block can overlap with executing the last.
Watch the same block change color as it moves through the system. Select any block to inspect it, or step through at your own pace.
PROPOSER SELECTIONt + 0.0s
A has the next slot
The schedule is stake-weighted and deterministic.
Slot 0 · 0–5sSlot 1 · 5–10s
FOLLOW THE BLOCKSCONSENSUS ∥ EXECUTION
01 Proposed
Built & broadcast
02 Accepted
Agreed order · queued
03 Executed
EVM results available
04 Settled
Results committed later
#100
Waiting for the first proposal
Press play or step to follow a block.
Toy schedule, fictional validators, and compressed timings. Block spacing is 1s in this example; it is not a measurement or a current network setting. The 5s proposer window and 5s protocol execution-to-settlement delay are different mechanisms. Polling rounds are schematic. A slow executor scenario shows overlap, not unlimited queue growth.
02 / WHO DOES WHAT
A proposer gets a turn. It doesn’t get the final say.
A
The schedule chooses a builder
ProposerVM wraps the C-Chain VM. For a height and slot, it samples one validator using the eligible set and weights at a referenced P-Chain height. Everyone can reproduce that selection. Greater weight means greater selection probability; the same validator can win successive slots or heights.
Slots last five seconds and restart for the next height. A healthy first proposer can build much sooner. ACP-226 separately sets a dynamic minimum delay in milliseconds, influenced by validator preferences. That lower bound is not a guarantee of observed block spacing.
Snowman uses repeated, stake-weighted sampling of peers to converge on a preferred chain and accept blocks. A proposer’s signature proves eligibility, not acceptance. The small network above compresses those repeated polls; it is not a single all-validator vote.
If a scheduled proposer fails to deliver, a later slot has its own sampled proposer. Since Durango, there is no ordinary timeout that opens building to everyone. Consecutive misses are possible. Timestamp gaps suggest elapsed slots; attributing them requires the historical schedule and more evidence.
Previously, validators executed a proposed block before accepting it. Continuous Execution puts accepted work into an ordered queue while a concurrent executor computes results.
Acceptance guarantees inclusion and eventual execution, not a successful contract call. A transaction can still revert. Its outcome becomes known on execution; later settlement records those results in consensus.
Checks still cover signatures, nonces and conservative affordability. A sender must cover the worst-case gas cost and value. Queue and block bounds constrain admitted work.
Finality and settlement are different milestones.
Accepted order is final under Snowman’s consensus guarantees. Execution computes effects. Settlement commits execution results in a later accepted block. If a node falls behind on execution needed to verify a settlement, it must catch up before judging that block.
Real C-Chain observations, separate from the simulation. Refreshes every eight seconds while this tab is visible.
Latest · executed head—Waiting for RPC
Finalized · settled head—Not yet observed
Recent median spacing—Loading adjacent headers
Transactions in latest—Inclusion count, not success count
Loading public RPC observations…
What these tags mean
ACP-194 defines latest as most recently executed and safe/finalized as most recently settled. pending may map to accepted or executed, depending on the implementation. This page does not use it to claim an acceptance lead.
What this view cannot tell you
Polling does not expose proposal broadcasts, consensus votes or the executor’s queue. Head distance is not execution latency. An EVM header’s miner address does not identify the ProposerVM signer. Unsupported fields or tags stay unavailable.
State-dependent calls may wait for execution. eth_getBlockReceipts refers to the transactions included in that block, even when their receipts are committed later. RPC semantics ↗
05 / THE ROAD TO HELICON
An evolving pipeline.
These changes address different bottlenecks: who proposes, how often blocks can arrive, and when transactions execute.
APRICOT PHASE 4
Snowman++ introduces proposer pacing
ProposerVM adds scheduled production around Snowman consensus. The original design had ordered eligibility windows and a later unsigned-block fallback.
A validator-influenced minimum delay replaces the old block-gas-cost pacing mechanism. Millisecond timestamps make subsecond timing expressible without conflating it with five-second proposer windows.
Mainnet activation at 15:00 UTC, with AvalancheGo v1.15.0. Continuous Execution separates accepted order from execution and later commitments. ACP-283 also makes the minimum gas price responsive to validator preferences.
Continuous execution enables specialized execution clients and is a prerequisite for some encrypted-mempool designs. These are opportunities discussed in ACP-194, not features activated by Helicon or a promised release schedule.