Check out the slides here 👉 https://docs.google.com/presentation/d/1h6EQ4nOcMk0a-1nUpiLBzDVKGSdvlW439G0Mk8rg3m0/edit?usp=sharing

I will Also give like what I think are the good parts about Intents And then eventually I'll get into like how you can use Suave as like Intent infrastructure.

That's like the rough arc The theme is largely inspired by this meme I made, if you're familiar with this format, of stop doing XYZ.

Oh, yeah, and then hello, I'd like an intent to not be front run.

I think that's also a little bit of a spoiler for the rest of the talk is that like front running is still very possible with intents.

It can move it in a meaningful direction but yeah.

Intents are probably the right abstraction.

And then, yeah, swab as intent infrastructure.

This is why I would say, this is how I would say like you know intense like aren't really that crazy or anything like substantially new there's been like 30 years of like you know UX research in web 2 and like we're suddenly bringing that now to like web 3 with intense is sort of my argument yeah so EIP 712 meta transactions those have been around for a while.

You know, if you've ever used OpenSea, those use meta transactions under the hood.

And those don't really specify a specific like execution path.

So, you know, You sort of just say, oh, I have an NFT, and I want around this much.

So yeah, giving users like, or so users giving more optionality to their executors, I think, Yeah, that would be the other part of the bear case is like, is that really where the bottleneck is right now?

Are all of our UX issues in Web 3 suddenly solved if we're like, oh, now a searcher has way more options for how to execute my stuff?

A lot of it is like users are confused, wallets, like cross-chain interaction, things like that.

And then, yeah, Cowswap is also already live.

You don't really specify your execution path.

You sort of just get thrown into a batch auction with a bunch of other people.

And then your intent is satisfied or your desire.

Yeah, and then lastly, for the bear case, would be it still reduces down to a market structure issue.

And so the intent language can have an effect on the market structure.

Like if your intent obstructs a certain party from being able to solve your intent, then yeah, that could potentially cut off parts of the market that you maybe do want.

Or maybe your intent and structure, like intent structure, could incentivize other market participants to come in and solve it.

And then I'll quickly go through what I see as intense, the progression of them.

I tried to do this on an x and y graph, but it ended up looking too much like an AMM, and I didn't want to do this.

So yeah, this is like uni v2 or maybe v1, like no router.

I need to go through this pool, and I can only satisfy x times y equals k for that single pool.

And then, yeah, OK, maybe this is like v2 plus the routers.

Like, the router gives you multiple different paths you can go to now.

And that's a huge improvement, because now you can maybe go through multiple pools that you didn't specify.

Yeah, and then I would say this is more what intents look like.

You just specify this entire space as a set of constraints.

And anyone can satisfy those constraints.

And so that's a single person intent for a single token.

But a lot of like intent infrastructure or intent like solver networks, like what they do is they take this and then they combine it with like a bunch of other people's intents, right, all like in this like 1 little box.

So yeah, I think like the intent infrastructure starter pack, it's like an ability to aggregate all of these sets of constraints into like 1 cohesive unit.

And then yeah, you sprinkle in some like crypto, like DKG, like perhaps you encrypt your, you know, set of constraints to, I don't know, like some, some, some public key that the network creates and rotates around.

I think people, I think like Furbio is an example of this DKG infrastructure on, I think, Osmosis.

Yeah, and then decentralization of compute and storage.

Like, we don't really get anywhere if you just encrypt, you know, your intent constraint into, like, my phone.

Like, we want a, like, marketplace of solvers competing for different execution on different constraints.

So that's what I would view as the bare necessity.

And how you actually encode these constraints into a way that a user can specify, I think, is where all of the intent work is.

So if anyone's working on intents, we'd love to talk about that.

But yeah, again, the language you specify to your user is going to determine the constraints that they create.

I think there's a really big argument as well that users don't know what they want.

So a really good example, I think, is Google Flights, if anyone's used that.

Like I sort of say where I want to end up in a date And you could imagine that as constraints and then it sort of comes to a bunch of solutions for me What's interesting is like the Google like search?

The Google flight search like what they'll do is like they'll actually show you the cheapest and then after that I believe they show you like Or sorry, they show you the yeah They show the cheapest first and then after that they show you the shortest flight and the cheapest And then I think they have like some other like priority ranking for like what they show you So I think there's a lot of room for platforms to make, so to choose these preferences for the users.

Because I don't really want to think about what my ordering of preferences are.

Like typically, 99% of the cases, the cheapest 1 is what I want to go with.

And then, oh, maybe it's 20 hours versus 10.

And this goes back to the bear case of too much.

Say I expose 100 parameters to a user, that really, like, doesn't solve any of crypto's, like, UX issues.

Yeah, so it expands the discourse to include a user's behavior and motivation.

That's sort of what I was talking about before.

Cross-domain intents are also something that's really interesting that you can expose.

If a user simply specifies these end states, then if I can satisfy, if I can provide some type of proof that I did the required action on 1 chain and then bridged it back, then that's super fair game.

And so that's really awesome that we can use this abstraction to move to a world where the actual chain you're on doesn't matter as much.

And so I sort of refer to this cross-domain ability as a builder superpower.

That's a very ETH-centric thing, if you're not in ETH, like on the ETH world.

But yeah, you could do things like multi-transaction flash loans is 1 builder superpower you could enable.

So right now, if you do a flash loan, you have to return the liquidity by the end of your transaction execution.

But if you're the builder and you're in control of the entire block, and you're even in control of another block on another chain, you can do very interesting things.

You can maybe even mix in signals from domains that aren't even on a chain.

It's a little unclear how you verify and then show that information to a user, but all awesome possibilities.

Yeah, and then this is the strongest 1 for me.

It's way more marketable than invariants.

I think A lot of intents are really just invariants.

I think even like slippage is sort of an invariant.

Like I think we do have some basic things like this right now.

But yeah, no 1 wants to say like my invariants are XYZ.

But like, oh, my intent is to get a token.

Cool, yeah, so that's like roughly, yeah, those are my bear and bull cases for intents.

I think it's unclear where intents will go, like whether these will actually be some type of like you know someone will design some like domain specific language that like everyone uses or maybe like certain actions will have like intents that become like very standardized and then you could use them across like every single like application or like a couple of them.

If you imagine like your phone like you probably have like hundreds of apps and so like do we imagine each 1 of those is gonna have their own like domain specific language for getting things in and out that you want?

And so a very basic pattern is what you're giving and what you want.

I think this is the Anoma intent language and structure is based off of this.

So you're on Instagram or TikTok or something, and you want to buy something.

You could specify how much you're willing to put in and then you specify how much you're willing to get out, and yeah, it really doesn't matter.

And so that's what gets segues into my next point, which is you specify how much you're willing to pay in basic versions of intents, but if these intents are public, then we're actually just in the same exact place we are with the transaction supply chain in MEV.

Because if I'm willing to pay $100 USDC for 1 coin, no one's going to be like, oh, let me just give you $90 for 1 coin, actually.

Searchers, executors, whatever you call them, solvers, are gonna push you like right up to the limits that you express.

And so this reduces back to like the MEV supply chain is my opinion.

So yeah, and this is sort of where I get eventually back, like go more into SWA than like less on intents.

So yeah, the state of MEV, this is like a rough transaction supply chain, intent supply chain now, it's new.

Yeah, so yeah, when users transact, they generate more value sometimes than they like.

They generate value that's like, you know, internal or external to the system that they don't even realize.

You know, being able to combine multiple different orders can maybe move a price in a different direction that you didn't expect.

Yeah, so right now this value is mostly captured by searchers and validators.

Originally the network sort of captured it in a sense with priority gas wars, but then flashbots we introduced, you know, a way for you to like privately express these preferences.

And then now with PBS, Proposer-Builder Separation, the design philosophy that validators and maintainers of the network's consensus shouldn't be the ones responsible for building blocks.

With PBS, I think it's like 90% of the value now ends up with validators.

And that's because of the market structure where the PBS design is basically a first price auction.

So it just incentivizes everyone to sort of bid max until they win the auction.

Yeah, and then users can be negatively impacted by some strategies.

And as I mentioned before, like intents, you can still get sandwiched with an intent depending on how you structure it.

So yeah, we protected with these private channels, we protect the chain from negative externalities.

But we don't internalize any of the value.

Like, yeah, it's actually really similar.

If you think about restaurant reservations, I think that's another example of somewhere that actually could maybe potentially internalize this demand, where, let's say, a restaurant only has 10 reservations in a day, and all of them get booked for free, or maybe there's some minimum cost you have to stake, which is like cyborg resistance in a way.

Like you could actually have like secondary market like trading on that And like none of that value goes to the restaurant even though it's like their actual spots that are like generating this contention for for usage Yeah, so that's a big issue don't internalize value Yeah, so we have private channels.

That's like the main approach There are smart contract approaches, which are really interesting, and I think Intents will allow more expression into how smart contracts can internalize this value.

Yeah, some quick examples are like liquidations in the Maker Auction, or the Liquidations Maker chose a Dutch auction to sort of remove some of this, like the negative externalities of everyone just like, you know, sending in tons of bids.

Now with the Dutch auction, it's just like, oh, that's the price I want, I'm taking it auction over.

Depending on if you can get on the chain, which Max had a lot about.

Yeah, NFTs, gradual Dutch auctions, that's another approach.

Yeah, and then AMMs, like order flow discrimination, has mixed effects.

Yeah, that seems to be 1 approach as well to internalize some of this demand.

Yeah, so those are current solutions, but they're not really good enough is what our view is.

So we introduced and have been working on like Suave, which is a system to like handle all of these things.

And so at a very high level this is like the different users of Suave and like how they would specifically interact with it.

But you can think of it as like, I'll start with a concrete example, so like MevBlocker or MevShare, if you guys are familiar, it's sort of like, you can submit a transaction and then, so with MevShare, we will like release a subset of the information that searchers can then search on and send a transaction back in which will back run it.

Medblocker does something similar, except they just like reveal the full information.

And then they, in order to stop like probabilistic sandwiching, they introduce like fake transactions.

So with SWAB, what you can actually do is you can encode that logic for your order flow auction as a smart contract.

And then the key here is that the user will then encrypt their transaction to the like Suave network and then they'll send it into a Suave node and then what happens is anyone can now run the auction and so you have like a verified program that they're running on top of like confidential bids.

And so the other cool thing that like smart contract enabled OFAs give you is that you can actually control the privacy model.

So like if you want to leak some of the information as in in my share like you can actually choose to do that and the user specifically allows certain contracts that they send their bids to so if I you know I can write some contract that will just like leak all of your information and give you nothing.

And as the user, you're free to click on that and use it.

But we think you can also submit to some really shitty AMM as well.

And people don't do that right now because there's various network of trust types of scenarios, people verifying.

And then we also envision people, like searchers, builders, etc.

And so they're incentivized to try and get as much order flow.

Another cool thing that this solves in the builder markets that I had personally experienced working on another builder is a lot of the iteration on block building mechanisms is most heavily constrained based on the order flow that you have.

If you only have 50% of the order flow and say you want to compete with bundle merging with Jared from Subway or something, the infamous sandwicher.

You can't really, like, meaningfully iterate there if you don't have their flow.

And so this stops a lot of builders from, like, being able to iterate.

And so we hope that in the base case, Swab will allow, lower the barriers to entry of the markets to be a block builder because you can sort of, based on some tenuous trust constraints in the beginning, you can get access to the entire order flow of everyone submitting through swab.

Yeah, so these are the takeaways, if none of that made sense.

Replace trust in operators with TEs, maybe eventually like threshold NPCs, some type of cryptography.

I think like if we had the exact same block building infrastructure as we have today, but it's on TEEs, like I think we're in a radically better place.

Like this isn't the, you know, 5 to 10 year like timeline of like what this like market will actually look like.

But we think it's way better than you just trusting some random server that says they won't leak your data or something like that.

And then many people have opinions on TEs.

We have a lot of approaches and very experienced SGX engineers on how to mitigate exploits and things of that nature.

I think the quick answer is if there is an exploit and 3 blocks are unbundled, that's really bad, yes.

But typically, it's not as bad as Ethereum going down.

And on top of that, like, sort of a lot of the people finding these exploits are, like, very sophisticated labs, like, doing it for academic clout.

They're not necessarily, like, some 16 year old, like, in their bedroom.

Which it could, that, I mean, that would also be cool.

Yeah, make the operation of the service decentralized geographically and administratively like ideally 1 person shouldn't be able to shut down the system.

And then geographic decentralization is like a much harder concern and that's that's a little bit longer term.

And then, yeah, user programmable based on smart contracts.

Yeah, the real vision is a marketplace for mechanism.

Any DGEN sort of AMM designer or smart contract dev should be able to just like read a tutorial deploy their own OFA on top of like Uniswap and like you could actually like meaningfully compete with flow and you Don't need to be like Associated with Uniswap or anyone?

Yeah And then okay these next few slides are a little technical.

If anyone's, like, familiar with, like, how an Ethereum node works, like, you have, like, the EVM.

You have, like, the state DB is, like, the main place you store things.

And then, like, some configs and contexts.

And then here's like, so over here is all traditional EVM.

And so yeah, you have this runtime, which ends up like at the interpreter, which is what translates your bytecode and actually runs the execution.

All that the MEVM is is it's a new runtime.

And it just plugs right into the interpreter.

And so now at execution time, your code has access to these fields, 1 of which is misspelled.

And so, yeah, you get a confidential back end that your execution gets access to.

And then as well, this off-chain ETH back end.

So you can call functionality that's used in traditional block building, such as like simulate a bundle or build a block.

You get access to all of those inside of your smart contract, which is really awesome.

So you can deploy a program that will take in a transaction, simulate it, do something with it based on the results of the simulation, maybe like route it to another smart contract, and you can kind of like stack these things.

So you could imagine like multiple layers of OFA contracts and then maybe multiple layers of like block building contracts, and then multiple layers of relay contracts.

And the point is not to be prescriptive about what the right route through the supply chain is.

It's, again, allow anyone to propose new ways that these like mechanisms can interact and operate and then just like let the market win for which one's the best.

Yeah so what are the differences like in terms of UX on like using Suave?

So this is like the you know transaction supply chain before you go from like user looking at the UX to a wallet.

That RPC will somehow end up at a block builder.

Sometimes it's like, you could view Infura right here, who sends it to a mempool.

And that's like the public broadcast to all the block builders.

And then yeah, block builder sends to relay and the relay sends the proposer.

Yeah, so Suave does not make any types of like opinionated claims about this portion, like that is for anyone to sort of go out and design like an intent framework that can operate here.

And then yeah, so what are the differences from there?

We have a new transaction type called an off-chain transaction type.

It basically is like a regular Ethereum transaction, except the call data is for a function call on suave, and then there's an additional field called confidential input, which is like your encrypted bundle.

But other than that, exactly the same, like 1 or 2 field differences.

As well, you also specify which contracts are allowed to execute over your transaction.

Yeah, so the RPC is also slightly modified.

SendRawTransaction just takes this new off-chain transaction type.

How the RPC and the block builder communicate, that is also different now.

They communicate through the confidential data store.

So your OFA can output a combination of the user's bundle with whatever transaction the OFA might have inserted.

So in MevShare, that's like a refund transaction.

And the way they communicate that is just by passing it again through SWAB internally.

And then, yeah, as well, block building, the way it communicates with the relay is through this confidential data store as well.

And typically, these 2 have been like HTTP.

And then, yeah, this portion will probably stay the same for a while, the relay to proposer.

So that will definitely always be the last mile.

We hope that there will be like some cool peer-to-peer network standards That'll come out so that we could just like broadcast these commitments to proposers and then they could accept But that's probably a long way to go Yeah, so there's intent infrastructure on just like many different domains being worked on.

Yeah, PBS like you can view as sort of like the last mile intent delivery.

If the guarantees we have with PBS aren't satisfied, if a builder's block is able to be like unbundled, is an attack we saw like in Ethereum where a malicious proposer basically said they were going to propose a block that came out of the MebBoost auction, but then they didn't.

And then once they received the block, they took out all the transactions, just sandwiched everyone, made like 20 mil, and then just like left.

And they of course got a slash on the Ethereum network because that's an equivocation.

But I think they made like, yeah, a couple mil and the slashing penalty is like 20, 000 or something.

So it was nowhere comparable or incentive compatible.

Yeah, 4337 is also approaching probably the earlier part, like maybe intent propagation.

All OFAs right now take transactions, just like regular transactions.

I think if someone just made a partial transaction, like OFA, similar as MevShare, like literally just fork MevShare, I think that would be a super cool thing that people might use.

So yeah, you can build an OFA in Solidity deployed on Swab.

Dex aggregators, you could also build them there.

Dex aggregators are really awesome today.

And I think if you want the best deal, you probably go to 1 inch, or like CalSwap as well.

So yeah, you can do all of that same logic, except just on swab, where you know the computation is guaranteed to be what you allowed.

See all SevenX Ventures transcripts on Youtube

Intents, Suave, and Off-Chain Auction Infrastructure - Dan Marzec @Flashbots | MEV DAY