Introducing Atlas Backruns

By Jacob Greene - FastLane Labs

Atlas is a generalized execution abstraction protocol for building application-specific order flow auctions on the EVM. Because of its generality and flexibility, Atlas can be used for a number of execution abstraction use-cases such as backrun order flow auctions, OEV auctions, and intent auctions. All of these auction types can be purpose-built by application developers in what we call an Atlas Module. While developers can always write their own modules or fork existing ones, we at FastLane will be supporting the Atlas ecosystem with high quality, first-party modules, and the accompanying implementation infrastructure for out-of-the-box use by anyone. Our first productized Atlas Module is a Backrun OFA, an easy way for swap frontends to capture backrun revenue to be distributed back to swappers, LPs, governance, or some mix at the app’s discretion. 

An overview of Backrun OFAs

User-activated alternative RPCs at the wallet level such as MEV-Share and MEV-Blocker pioneered permissionless backrun OFAs on Ethereum, and have returned more than 2,000 ETH in rebates to users since launching in April 2023. Both of these OFAs implement a privacy mechanism that prevents users from being frontrun by concealing the real user swap until after settlement. They also both use "block building time simulations from builders, which serve as the "auctioneer" that enables the privacy mechanisms to work with a permissionless solver network. Anyone sending trades to AMMs may be leaving money on the table if they aren’t using a “block building time” OFA, from regular frontend users, to aggregators, and CoWSwap solvers. RPC based backrun OFAs, with their impressive growth on mainnet Ethereum and given the continued proliferation of AMMs in on-chain trading, have proven to be an important tool for delivering DeFi users best execution. 

Block Building Time Simulations

“Block building time” simulations are highly beneficial to backrun OFAs because they deterministically find the best bid. MEV auctions that are not builder-integrated have access to less reliable simulations. This is one of the reasons why CoWSwap solvers likely benefit from sending their transactions to MEV-Blocker, as builders are not natively integrated into the CoWSwap auction. Reliable simulations also prevent solvers from griefing the auction with transactions that trick the simulation by behaving differently when actually executed on-chain. Imagine a transaction showing a bid of 1 eth for the off-chain simulation, and actually only sending .01 eth when executed on-chain. In summary “block building time” simulations enable OFAs to:

• Deterministically find the best bid, allowing users to receive rebates from a permissionless solver network without them having to bond significant capital. 
• Compute bid amounts at block building time to support privacy mechanisms.

Programmable Privacy

Hints, popularized by MEV-Share, is the first in production example of programmable privacy for OFAs. It enables users to send order flow to a permissionless solver network while maintaining sandwich protection. Hints restrict the information shown to solvers which forces them to blindly solve for the backrun on-chain. This prevents solvers from ever having enough information in time to frontrun the user. Only the pool address that the user swap is targeting is revealed publicly to solvers, hiding data like the trade size, direction, and swapper address. Hints further enable collaboration where you otherwise wouldn’t have it that makes the market more efficient, such as the ability for solvers to backrun other solvers like the CowSwap and MEV-Blocker example.

Permissionless Solvers in Backrun OFAs

This paper described how permissionless solver networks for intents auctions like UniswapX may in some cases actually result in worse prices for users than if there was an oligopoly of searchers. This is mainly due to congestion and entry costs. Congestion costs may arise when many solvers compete to purchase the same token in order to fill a user intent. Entry costs represent setting up infrastructure, or any other upfront costs to bid. 

Backrun OFAs supporting programmable privacy differ significantly in a few ways from the intents auctions discussed in the paper:

• There are no upfront capital requirements for backrun MEV since the opportunities are atomic and can utilize flash loans.
• There is little wasted entry costs on latency-advantaged infrastructure because solving mostly occurs on-chain. 
• There is no competition to purchase tokens beforehand that might create congestion costs because the user swap is kept private until after settlement. 

This indicates that backrun OFAs may benefit more from a permissionless solver network than an oligopoly due to their specific market structure, compared to something like swap intents, which may disproportionately benefit from something like a solver reputation system. 

Gaps in current backrun OFA market

There are a number of issues, however, with the current backrun OFA landscape. 

• Applications (and their stakeholders, like LPs) are left out of the value accrual supply chain of atomic arbitrage because they cannot easily capture the backrun MEV that they are leaking.
• Typically, due to an unpredictable and often slow transaction UX, wallets who support backrun OFAs via RPC will make it an opt-in feature, leaving many of the users generating backrun MEV with worse execution than they should be getting. 
• Permissionless backrun OFAs rely on trusted relationships with block builders, who can see a user’s unconfirmed full transaction, and thus can vertically integrate as a bidder in the auction. 
• The requirements of block builders for permissionless OFAs today limits them to Ethereum mainnet, leaving L2 swappers with worse execution than mainnet swappers.

The Atlas Protocol 

Atlas uses execution abstraction to replace the need for trusted “bundles” offered by block builders. Combining user and solver operations into a single transaction makes the order of operation execution, and rebates from solvers to users atomic and trustless. Atlas is also designed to be easily integrated directly into interfaces, turning backrun OFAs from an opt-in feature at the wallet level to a default feature at the application level. 

Atlas Backruns

Atlas extends permissionless backrun OFAs to L2s, giving users the ability to receive backrun MEV rebates while maintaining frontrunning/sandwich protection. The Ethereum roadmap has been pushing for activity to move to L2s for a while, and we have just recently started to see this trend really play out. DEX volumes have significantly increased on L2s like Base, Arbitrum, and Zksync. With this, atomic arbitrage opportunities have also significantly increased on these L2s, and are moving away from the L1. We only see this trend continuing and the desire to capture this surplus value for applications and users being inevitable, especially when there are little upfront costs or risks associated with it.

Atlas unlocks permissionless backrun OFAs on L2s with “Ex-Post Bids”, our own equivalent for “block building time” simulations on any EVM chain. The “Ex-Post Bids” name comes from the fact that these simulations are “based on actual results rather than a forecast”, they are 100% accurate simulations rather than less reliable off-chain simulations from a non block builder. They enable programmable privacy, allowing users to receive backrun rebates from a permissionless solver network while maintaining frontrun protection.

• Atlas “Ex-Post Bids” take advantage of cheaper gas costs on L2s to reduce trust and increase decentralization in the auctions by running the simulations on-chain, rather than off-chain or in a trusted block builder.
• “Ex-Post Bids” provide support for mechanisms such as MEV-Share style hints, or other forms of programmable privacy.

"Ex-Post Bids"

Atlas “Ex-Post Bids” take advantage of cheaper gas costs to simulate and find the highest bid on-chain, providing “block building time” simulations without having to trust an off-chain actor. This is a clear example of cheaper gas fees allowing developers to build more decentralized blockchain applications. On L2s the gas costs of running “simulations” on-chain should for the most part be far outweighed by the MEV rebates. The median rebate from MEV-Share is $13, and it can cost under 1 cent to do a swap on most L2s at the time of writing.

“Ex-Post Bids” uses the ability to include multiple solver operations in a single atlas transaction by looping over the solver operations to find the highest paying one. Each solver operation is executed behind the user operation, the bid amount is recorded, and then they are reverted. Atlas will then go back and re-execute the highest paying solver operation from that loop, essentially providing an on-chain simulation. This Atlas feature allows solvers to submit operations where they do not know the bid amount ahead of time, enabling programmable privacy such that solvers are unaware of their and the users operations until after settlement. 

Decentralization and Observability

Atlas "Ex-Post Bids" provide more observability into OFAs because there isn’t an opaque process happening inside centralized builders who host the auction, a significant portion happens on-chain instead. With backrun OFAs today you have very little insight into what is happening with your transactions. You don’t know if a refund bundle was ever generated or sent to a block builder, and you must share your full transaction details with the builder. Builders who are also bidders in the OFA have a clear advantage from vertical integration because they receive the full transactions from users, whereas searchers who are not also builders do not. They can even sandwich or backrun users without providing a MEV rebate. This was highlighted by Blair in a case where a vertically-integrated builder was potentially abusing their trusted relationship and backrunning without rebating. It took ~80 days for anyone to notice this, highlighting a need for more observability in Backrun OFAs. 

With Atlas Backruns you:

• You can remove trust and increase transparency into the “block building time” simulation process, because a portion occurs on-chain.
• Do not have to share your full transaction details with a block builder. Bundlers in Atlas Backruns cannot adversely select solvers because the solver operations use on-chain computation, and will revert if it isn't atomically profitable. This means that the user can be their own bundler to minimize trust, removing information-advantaged parties from the OFA.

Reputation Filtering

In practice all permissionless MEV auctions with low gas costs, or free reverts, have some kind of filtering in place before solver transactions are allowed to touch hot resources such as a simulation engine. There are more transparent reputation systems like the Flashbots and Titan builders, or more opaque methods such as the “random heuristics” used in Beaver builder. These prioritization techniques are used alongside simulations to select the winning bids in the auction and prevent spam.

Since Atlas Backruns will not utilize block builders for simulations, there needs to be a new filtering mechanism before solvers are allowed access to Atlas on-chain simulations. Similar tools as the ones used by Ethereum block builders can be used by the Atlas Operations Relay to prevent spam. Atlas handles gas accounting atomically on-chain, which will always be the first spam resistance tool for Atlas Backruns. Reputation scoring systems or priority queues can be used as a second barrier of spam resistance when gas costs are not enough. Off-chain simulations can additionally be used to increase assurances that on-chain simulations are only performed when the rebate is greater than the extra gas costs incurred. 

Conclusion

Atlas unlocks “block building time” simulations and programmable privacy on any EVM chain. This allows modules like Atlas Backruns to offer L2 users MEV rebates from a permissionless solver network while protecting them from frontruns. Atlas backrun modules are differentiated from current permissionless backrun OFAs because they:

• Use smart contracts rather than trusted relationships for the simulation engine.
• Don't share full transactions with builders by allowing users to bundle their own transaction.
• Have trustless rebates from solvers to users. 
• Allow applications to easily offer backrun OFAs as a default feature.