# 4337 mev supply chain ![](https://i.imgur.com/WQ6L5aj.jpg) :::info :bulb: TLDR; Once widely adopted, ERC4337 will create alternate "user operation" mempools, fragmenting public sources of MEV opportunities. OFAs will also most likely provide support for user operations. 2 new actors are added to the supply chain, "bundlers" and "paymasters", both of which have functionality that may be subsumed by searchers, builders, and OFAs. ::: ## intro The main innovation 4337 brings is one of programmable control, specifically over the core identity in ethereum, the EOA. For the primary actors in the MEV supply chain, the end result will be a new parallel mempool for meta-transactions known as user operations.(And some other cool stuff) If you're looking for an explainer of what 4337 is and actually means, [check out frontier tech's write up](https://frontier.tech/unpacking-erc-4337) or [blocknative's](https://www.blocknative.com/blog/account-abstraction-erc-4337-guide). ## 🔀 evolution of tx venues In the beginning, there was only one public venue for txs; the mempool, accessible by participating in the eth p2p network. From there private venues popped up to support scenarios of state contention in tx level preferences. Post `mev-geth` and `mev-boost`, builders started keeping their own private mempools, and our supply chain looked like this. Typically mempool's and nodes aren't shown in the supply chain but we do so to highlight the differences on the horizon. ![](https://i.imgur.com/WXG21Oc.png) If we add in OFAs it might look something like this: ![](https://i.imgur.com/HBFpDKq.png) ### ☀️ enter erc4337 Instead of txs being the only kid in town, user operations are a new and exciting way for interacting with the chain. From the user's POV the process looks [entirely similar to signing a meta txn](https://eips.ethereum.org/EIPS/eip-4337), but one key difference is that user operations are not propagated in the standard p2p network that ethereum nodes run. Each "bundler" participates in an alternative mempool p2p network. For the happy contention free path our 4337 modified supply chain looks like this: ![](https://i.imgur.com/U6Ndm3j.png) ### 📦 but what's a bundler? The main purpose of a bundler is to act as a userop relayer and proxy EOA - it does that by participating in the alt mempool p2p network, aggregating user operations, and submitting bundles of user operations as a tx to the public tx mempool. Alternatively they can also be sent directly to a builder. ![](https://hackmd.io/_uploads/By4hbUT1h.png) There is alot of overlap between the functionality of bundlers, searchers, and builders. For example, searchers maintain very live mempools to find MEV opportunities. Builders do as well to ensure they have the most competitive blocks because private orderflow still must compete with public orderflow. On top of this, builders are also familiar with the combinatoric games of ordering and simulating transactions. Thus it seems very likely that ***searchers will run bundlers*** to extract MEV from user operations, ***AND*** ***builders will run bundlers*** to build the most competitive blocks possible. ***It only takes one searcher or builder to gain an edge from incorporating this stream of orderflow before it becomes table stakes for playing at large***. ## 💔 the divorce of mempools 💔 Assuming that searchers and builders converge on running bundlers, the key difference that 4337 will create for the MEV supplychain is splitting the old public tx venue, "the mempool", into two. One for txs and one for meta-transactions. Not only does the 4337 spec create 1 new public mempool, it allows for the creation and participation in an arbitrary number of alternate mempools denoted by their `Mempool ID`. This opens the door for arbitrary user operation mempools with potentially even different rulesets. It would be interesting to see if OFAs could run over these alternate mempool p2p networks. Thus our new supply chain looks like this: ![](https://hackmd.io/_uploads/B1N62Rkg3.png) In the old regime where txpools were heavily linked to an execution client, anyone who was interested in keeping sync'ed chain data was also by default running a mempool. With the decoupling of txpools and execution clients, the initial bundler network will likely be much less populated than the current. If the bundler network is primarily run by searchers, it's possible we could see delays on user operation inclusion becoming a feasible strategy to gain an edge. ![](https://i.imgur.com/v5tvSmO.png) ## 💸 paymasters Paymasters are another role in the erc4337 system which allow 3rd parties to pay for your user operation at execution time. In order to sponsor your user operation, it must be sent to a paymaster API. On approval, Paymasters will respond with `paymasterAndData`, data which must be included in the signed user op before broadcasting. ![](https://hackmd.io/_uploads/HykJ601e3.png) Paymasters, if used, will technically have the earliest access to user op flow in this new environment. One thing to note is that paymasters are initially intended to be dapps or wallets looking to provide a differentiated and gasless experiences to their users. But we could also see OFAs guranteeing a rebate up front for a users user op orderflow as a strategy. Once we end up with hundreds of different venues to have your txn processed, a big question will become how to abstract this experience from the user, as well as still provide lively quoting. ## 🐢 mempools all the way down Bundlers may be a trojan horse for bringing alternative mempools to ethereum. In many ways alternative-mempools-as-standalone-nodes is a superior solution to the status quo, because it allows for upgradability and experimentation in mempool design. With minimal spec changes, someone could create and support an encrypted mempool. There is still some work to be done on the [censorship front](https://notes.ethereum.org/@yoav/unified-erc-4337-mempool) as well. ![](https://i.imgur.com/3S9CvTz.png) If we assume that the role of the bundler will be subsumed by searchers and builders, along with OFAs continuing to proliferate, our MEV supply chain looks more like this: ![](https://i.imgur.com/VFtzsGv.png) ## 🪄 enter: the intent layer Is a global intent layer for counter party matching the missing solution? We can think of services like OpenSea, CowSwap, 1nch, etc as off-chain counter party matching services. 4337 alt mempools may be laying the groundwork for a global intent gossiping layer, which if equipped with sufficient cryptography, could allow for general purposes solvers to match a users intent to a relevant counter party intent before even hitting the chain. With this frame of mind we can think of 4337 alt mempools as setting the framework for an intent gossiping network. ![](https://hackmd.io/_uploads/BJR9hWZgh.png) source : https://anoma.net/blog/agents-of-anoma-intent-gossip/ ## 🥪 food for thought - How can we usher in a safe year of experimentation with OFAs and private mempools? - userops once placed into a tx vessel can be stolen by any other bundler (or EOA), so bundlers will most likely send directly to builders - if user ops can be valid on multiple chains, wen multichain bundler? - how do context aware "smart" transactions play into all of this?