Intents qualm

Just some loose thoughts that might open up some interesting discussion since it seems like SUAVE is taking a different perspective:

Intents won’t be that useful for MEV searching because if there is value to capture, the solvers would just keep it for themselves. Even if the user placing the intent, offered to pay the solver on some utility curve for the quality of execution of their strategy, they won’t be able to pay enough for it to +EV for both the solver and the “intender”. This is because MEV searchers are probably the most unbias, rational actors in all of crypto - they care about profit above all else.

1 Like

I suspect we can find some edge cases where this is not true, but I agree in general that intents alone (as defined by 4337) don’t do too much.

If your definition of an intent is a bit broader than what EIP4337 implements and the intent looks more like a bundle (which can predicate about things outside the information available in the EVM), I do think that intents alone are quite powerful.

Think of:


The definition I’m using for intents

From this definition CoWswap “intents” are not submitted by searchers but by retail and the solvers are the searcher/builder equivalent.

The questions of whether 1) retail will use intents and 2) searchers will help solve coincidence of wants or directly fill intents is a separate one - and both are probably yes. My original point was more about whether searchers will use intents.

@Quintus thoughts?

Also I don’t really see Cowswap intents as a viable example of intents, because there are pretty antithetical trust assumptions to what we consider DeFi (although the Cowswap team are great people!)

I’m not 100% sure what the question is here.

Weird things i thought of:

  • searchers could take 4337s and add a backrun into effectively another 4337 and then pass this on if they know that there is another searcher with a complementary/non-exclusionary strat on similar opportunities if 4337 would lead to more gas effective execution and stronger unbundling guarantees.

  • users could submit 4337s asking for a price that’s barely attainable and you would need almost optimal routing for. Then if the block pans out in a way that the trade is actually possible, a builder could route it for them. Not sure how practical this is

I think changing tx types is definitely interesting for MEV but feels as thought 4337 is more of a UX thing than massively impactful for MEV


I’m not 100% sure what the question is here.

Needing two participants in a tx is a new tx format (what you’re describing 4337 as being used for).

My original point was that the “solver” (other participant) has an incentive to always min max the person submitting the tx so intents won’t be that useful for MEV searching (at least w/o credible solvers).

You mentioned intents are quite powerful and cited Cowswap and a paper on Verifiable DEX design.

But Cowswap has a major trust assumption re: the solver auction being not censoring and run properly so it doesn’t seem like a viable counterexample for intents working for MEV searchers in practice (let alone retail users).

I was interested to see if you still held your initial opinion / why?


Oh I see!

Yes, I think you’re getting at one of the main concerns for the “MEV supply chain” now. Good execution currently depends on reputation/trust which can be loosely enforced over the repeated game. As soon as you make something permissionless you get rekt because of exactly what you’re describing.

The whole idea with SUAVE is to remove the necessity of trusting things like the CoWswap or MEV share servers, but one intermediate step you could take is to make the trusted aspect easily verifiable.

Using your example, if I’m sending intents to builders requiring good routing. It’s much better to specify that routing must be at least as good as X algo can provide.

1 Like