Key takeaways
- AggLayer uses zero-knowledge technology to enable low-latency interoperability between blockchain chains while maintaining safety and preventing destination chains from being rugged.
- The aggregation layer aims to unify liquidity and state across multiple chains, solving the fragmentation problem that creates poor user experience and higher costs through slippage.
- Polygon's synthesis approach combines monolithic and modular blockchain philosophies by supporting a multi-chain ecosystem with mechanisms to unify state and liquidity across chains.
- Bridge and call functionality allows users to move funds between chains and execute smart contracts in a single interaction without understanding underlying zero-knowledge technology.
- Fungibility of wrapped assets across chains is a key ingredient for unifying liquidity, requiring liquidity providers to bridge synthetic versions of assets to native versions.
Why Polygon built the AggLayer
Ashton Addison: What is the AggLayer, and what problem is it meant to solve?▶ 00:45
Brendan Farmer: The aggregation layer is a big one for us. We've been thinking about these underlying problems for a long time. Right now the debate over blockchain scaling is stuck between the monolithic side and the modular side. The monolithic side says that scaling in crypto really means scaling access to liquidity and shared state, and maintaining composability. If we create a bunch of chains that can't access liquidity where it's available and useful, we haven't scaled in a meaningful sense. The modular side says we'll never bring internet-scale crypto to a single chain. We'll always run into bottlenecks and contention for shared resources, so we need a multichain approach. Both sides get something right. Fragmented liquidity creates a bad user experience and complexity, and it costs users through slippage and worse execution. But we also need horizontal scaling, so we can add block space and avoid intractable limits. We need some way of taking fragmented block space and unifying it so users can access shared liquidity and shared state. That's what we've been thinking about with the aggregation layer: how do we take a modular, multichain ecosystem and unify liquidity and state across those chains?
Ashton: What was the design thinking behind it, and how do chains interoperate quickly without losing safety?▶ 04:16
Brendan: The main design principle is simplicity and minimalism. We want a protocol that can support internet-scale crypto adoption, so it has to be minimal and not opinionated. It needs to accommodate a range of chains running on different execution environments, using different tokens and different sequencing mechanisms. The foundation of this vision is ZK tech. Right now, for two L2s to interoperate, you go from your initial L2 to Ethereum and then to your destination chain. If I want to move tokens from chain A to chain B, I withdraw from chain A and wait until the proof of my withdrawal hits Ethereum and that block is finalized. That's about 25 minutes. Then I deposit into chain B, which is another 12 minutes of latency. So the question was how to create low-latency interactions between chains in a way that's safe and trustless. It turns out we can do that with ZK tech. Chains can interoperate however they want. They can run full nodes for each other, use a shared sequencer, or rely on builders or relays. That coordination is out of scope for us. What we guarantee is this: if chain B receives a message from chain A at lower latency than Ethereum, and two seconds later wants to start building blocks on that message, chain B can commit to the state of chain A that produced it. We can use ZK tech to ensure chain B will never be finalized and accepted by Ethereum unless the chain A block it relies on is valid and also finalized on Ethereum. So the destination chain will never be rugged by anything chain A does.
Nothing changes for users, validators or full nodes. We just turn on a prover and start generating proofs for that chain.
Unifying liquidity and state
Ashton: How will liquidity actually be unified across chains?▶ 08:18
Brendan: One key ingredient is fungibility. Think about how third-party bridges work today. If I want to bridge from Polygon to Arbitrum via Wormhole, I'm locking my ETH and minting Wormhole ETH on Arbitrum. The problem is that Wormhole ETH is not fungible with regular ETH. I need a liquidity provider to get me from the wrapped, synthetic version of an asset to the native version. The unified bridge and the AggLayer let us ensure that L1-native assets and L2-native assets are fungible across chains. I won't be minting zkEVM ETH. I'm just getting ETH on the destination chain. That's a very powerful thing for unifying liquidity. What the world will look like beyond that, whether a single super-deep pool or a bunch of pools that rebalance automatically through arbitrage, is an open question. With a single pool, getting best execution and priority becomes really expensive, and normal users could be priced out. You can imagine those users batching their transactions on low-cost chains, with the batch executed on the pool with the deepest liquidity, so the cost is amortized across everyone batching. That has similarities to the model of many pools with orders routed between them. I think both are uniquely enabled by the AggLayer, and it will be interesting to see which dominates.
Ashton: Beyond speed and privacy, what does ZK technology add to the AggLayer?▶ 11:54
Brendan: ZK tech is the fundamental ingredient of the AggLayer. We need to be able to prove to Ethereum that this giant multichain ecosystem, with thousands or tens of thousands of chains, is operating on consistent views of the same state. You can't have chain A telling chain B one thing while double spending on chain C. Proof aggregation is a really important part of that. It does a few things. It guarantees that everything happening across all chains is valid, and that chains are relying on consistent views of other chains' states. It also provides a safety guarantee. If a single chain is compromised, attackers cannot drain the entire ecosystem of funds, because we enforce that no chain can withdraw more than the tokens deposited in that chain, even if that chain's prover is unsound.
The Type 1 Prover and onboarding EVM chains
Ashton: Many EVM chains have different levels of ZK technology. How do they join the AggLayer?▶ 13:58
Brendan: It's a timely question, because we just announced our Type 1 Prover. To participate in the ZK-powered ecosystem that runs on the AggLayer, chains must include ZK proofs. We need a guarantee from each chain that everything that happens on it is valid. Today there are lots of EVM chains, including optimistic rollups, Polygon PoS and alt-L1s that use the EVM. There's a ton of liquidity, state and huge user bases locked in those chains, and they can't interoperate the way we're describing because they don't support ZK tech. What we released last week is the Type 1 zkEVM prover. It's a really cool piece of technology because it lets us take any existing EVM chain, whether that's an optimistic rollup, a sidechain like Polygon PoS or an alt-L1, and we can instantly start generating proofs for that chain. That lets those chains onboard into the AggLayer. It's a really important component of the overall vision, which is to take all of this fragmented liquidity, state and user bases and unify them so we can deliver a better experience for users.
Ashton: How hard is it for a chain to adopt this? Do they have to rebuild everything?▶ 16:03
Brendan: No. There are no client changes required. Nothing changes for users, validators or full nodes. We just turn on a prover and start generating proofs for that chain. If a chain is an L2 and wants to migrate its L1-native assets to get fungibility with L1-native assets on the AggLayer, it would have to do a bridge migration. If it doesn't want to do that, it can start generating proofs immediately.
No chain can withdraw more than the tokens deposited in that chain, even if that chain's prover is unsound.
Costs and what comes next
Ashton: How will this affect transaction costs on layer twos?▶ 17:46
Brendan: It's important to note that L2 costs are generally dominated by the cost of posting data to Ethereum. Every rollup transaction has to post some data, and that has a cost. This Ethereum data availability cost is what most of the transaction costs you see on L2s are based on. But there are other costs. Optimistic rollups have a downside: if you want to withdraw via the native bridge, you have to lock your funds for seven days, the challenge period, during which anyone can submit a fraud proof if the operator is misbehaving. That window imposes a real cost on users. No one wants to lock up highly volatile crypto assets for seven days and take on duration risk, so they rely on third-party bridges that charge fees. In aggregate that's millions of dollars. The nice thing about the Type 1 Prover is that we can convert these optimistic rollups into ZK rollups and replace the duration risk and bridge fees with a proving cost. Users cover the cost of generating zero-knowledge proofs for the chain, and they get instant withdrawals. The proving cost is really low. Right now we're seeing two to three tenths of a cent to prove Ethereum mainnet blocks. In the next few weeks we'll release an upgrade that takes that to between one and two tenths of a cent. In the medium term we expect a further 30 to 50x reduction, to something like four thousandths of a cent. That's orders of magnitude less than the capital and efficiency costs imposed by optimistic rollups. It doesn't solve the data availability cost, but it does solve these other rollup costs that users still bear.
Ashton: What is the rollout plan for the next month or two?▶ 22:00
Brendan: The AggLayer will be released in stages. This month comes V1, which is effectively a unified bridge. It provides the fungibility property I described, so you can move assets across chains, but it won't have the low-latency property yet. That will come in a few months. So in the near term it's about providing fungibility and getting chains onto a unified bridge. In the medium term we'll release further upgrades that enable the low-latency mode for cross-chain composability and interaction.
This interview has been edited for length and clarity. Recorded February 2024.

