Introducing SIP-6: Upgrade Sei to Autobahn Consensus

SIP-6 proposes swapping Sei's consensus for Autobahn. It will be the biggest piece of the Giga upgrade. Autobahn will introduce Multiple Concurrent Proposer design to Sei, raising the chain's performance to 200,000 transactions per second and 250ms finality, as well as introducing properties that will make it substantially better for onchain trading than other venues.

On Sei today (as well as essentially every other blockchain), each block is built by one validator, who has a complete monopoly over the contents of each block it assembles. This single proposer design, while commonplace, is the cause of most of the shortcomings of today's blockchains, such as rampant MEV, censorship risk, slow throughput and more.

SIP-6 proposes ending that. Under Autobahn, each of Sei's 40 validators will continuously stream transactions to the network as they are received. A deterministic formula will then order transactions across all 40 lanes such that every validator will compute the same result. The Sei community is invited to read SIP-6 and leave feedback in the GitHub discussion.

Giga is Sei's plan to rebuild all three layers of the chain: execution (how transactions run), storage (how the chain keeps track of balances and history) and consensus (how validators agree on the order of transactions, and the outcome of execution). The first stages of the execution and storage work, Ares and Eidos, are already live on mainnet. SIP-6 will start the consensus upgrade, the single most important part of Giga.

From a user perspective, nothing will change on the chain until Autobahn goes live on mainnet.

Why Autobahn is good for trading

Minimal censorship risk: On a single proposer chain, the proposer has a complete monopoly over the contents of the block it creates. It can leave transactions out, or reorder them to extract as much value as possible for itself. On Autobahn, no validator will hold that monopoly. All 40 validators will stream transactions out at the same time, each in its own lane. Autobahn is designed so you'll be able to send your trade to more than one validator, which means no single validator will be able to keep it off the chain. A validator acting maliciously will only be able to affect its own lane, not the whole network. This removes the very real, and costly, risk that a validator could exclude your transaction for its own profit.

Deterministic ordering: On a single proposer chain the proposer decides who gets those spots, and can sell them or take them itself, which is how toxic MEV such as sandwich attacks succeed. Autobahn will replace that decision with a deterministic ordering rule that every validator will run, putting transactions with the highest priority fees first. No proposer will be able to reorder your trade to squeeze more value out of the block. This has the major effect of internalizing MEV. Whereas on other chains MEV is targeted at users and value is extracted out of the ecosystem, on Sei much of the value that would otherwise have been extracted as MEV will be internalized as revenue, internalizing much of the MEV aimed at ordinary users on other chains. Sedna, a planned future upgrade, is designed to amplify this effect by introducing privacy to Sei, making sure no single validator gets enough of your trade to read it early.

Protected cancels: Autobahn will also directly benefit market makers on Sei. On a single proposer chain a cancel can be held up by one slow or hostile validator, forcing market makers to protect themselves with wider prices. On Autobahn a cancel will get the same protection from censorship and reordering as any other trade, allowing market makers to offer tighter quotes.

Extremely high-performance: A single proposer caps the chain at what one machine can send. Autobahn will put all 40 validators to work at once, so the chain will be able to carry far more trades. That's how Giga plans to hit its targets of 200,000 transactions per second and finality under 250 milliseconds.

Autobahn will enable centralized-exchange-quality trading, on a fully decentralized L1. That will open the door to trading infrastructure that has mostly lived on centralized venues: central limit order books, prop AMMs, and similar systems, all built natively on-chain.

The final steps towards Giga

SIP-6 initiates the final track of the Giga upgrade. It should be noted that there is considerable work to be done before Giga goes live on mainnet.

1. SIP-6 has to be accepted under Sei's SIP process.
2. Autobahn needs a public testnet. So far it has only run on Sei Labs' internal test networks. A public testnet will let validators and app teams run it and try to break it, and anyone curious will be able to join in.
3. The code needs extensive audits, and auditing a multiple concurrent proposer chain is highly specialized work. SIP-6 requires several independent firms to audit the release, and every critical and high-severity finding has to be fixed before it can activate. The upgrade also has to be rehearsed against a copy of Sei's real chain state, and validators need published hardware and bandwidth requirements.
4. Sei's Cosmos retirement, SIP-3, has to be completed first.

Autobahn will only be the first of Giga's consensus upgrades with Hermes planned to follow. Hermes is designed to sit on top of Autobahn, accelerating finality even further.

What it means for you

Nothing changes until Autobahn switches on. Until then, Sei keeps running the way it does today, with one validator building each block.

If your SEI is in an EVM wallet, that wallet will carry over as it is. SIP-6 won't change how you stake either, and it won't change who receives fees or how staking rewards work.

Still holding SEI in an older Cosmos-style wallet, with an address starting sei1? Follow the SIP-3 migration guide before Autobahn switches on.

Get involved

Read the full proposal and leave feedback in the GitHub discussion. Follow @SeiNetwork to hear when the public Autobahn testnet opens.

Disclaimer: Forward-looking statements are subject to change based on development progress, testing results, network conditions, governance processes, market feedback, and other factors. Actual timelines, performance, figures, and outcomes may vary.