We talk intents with Hart:* Harts experience trading treasuries &amp; the unique market structure there* Intents ~= Market Making* Potential use cases beyond trading* Bridging and Harts project Across* Crosschain Intent systems &amp; their Capital...

This is where we interview founders, traders, and investors to take a deep dive into the crypto ecosystem and its market structure.

This show is brought to you by Parsec, which is the premier DeFi and NFT analytics hub.

We figured the best person we could bring on to talk about maybe some new tech in crypto is Hart.

Talk about all things intense, some of the stuff he's been working on, market making and more broad DeFi stuff.

Yeah, so I think like some people, especially like more seasoned DeFi people will know you, but I think you get to get a quick rundown of your background.

You know, you have a pretty extensive background that's pretty impressive and would love to just hear like how you ended up, you know, in the DeFi world.

Okay, well, I will share what I think is relevant to our discussion.

So yeah, I am, like I have the evolution of being a tech guy that became a finance guy that became a fintech guy that became, I guess, a crypto dude.

So that is there's some usefulness there.

But it is like that he computer science worked at Goldman Sachs as a bond trader for 8 years, which is relevant to I think the conversation we're going to have about the role of market makers, particularly in like an intense framework and in like a cross-chain environment, but learned a lot about market making and the efficiencies and inefficiencies there.

And then I started a fintech business that was acquired by an asset manager in 2017.

And I'd been playing along with crypto the whole time and then went full-time into crypto in

sort of before DeFi was actually a phrase and was interested in just how we could build financial contracts on smart contract platforms like Ethereum.

And then just so everybody has the context, I started a project called UMA, which originally was sort of focused on some synthetic assets use cases.

Synthetic assets was really a showcase for an Oracle system we built called an optimistic Oracle.

Because there was no DeFi at that point, so we wanted an example of showcase how we could use this Oracle.

We've since really focused, Uma, on just the Optimistic Oracle, which is a type of Oracle that can validate any type of verifiable data.

And then we built a cross, which is a cross-chain bridge that uses the optimistic oracle as its security model.

And it turns out a cross actually works really well.

And a cross has got us closer to an intense like architecture for cross-chain bridging as well.

Yeah, so the reason that I reached out to you is because I think I was like, they're posting about intense and because they got kind of like quite hyped over the summer and say it's that out that down a little bit.

But there was a lot of talk about how this is like, you know, going to be everything x, y, and z.

But You gave a really good example of an actual intense flow that you had experienced as a trader.

Maybe you could just walk through that example you gave in your thread and why it makes you bullish on intense and kind of the DeFi context.

So at Goldman, I had a really old school job.

And people, still today, hedge funds call you up or central banks call you up on the phone and are like bid 100 10 year treasuries.

And it's just a weird old school job with a bunch of technology that's getting layered on top.

But the idea of why this was relevant to the intense framework is like, you have a market maker and the client doesn't care how you do the thing.

They just want to sell you bonds at the best price or buy bonds at the best price.

And this is a little bit different from like equity markets where most equity markets don't have a market maker acting as principal.

That's actually taking the other side of the risk.

They will actually go and route your order for you or like execute it.

So we'll take your order and then they'll go and they'll buy some of it over there and some of it over there.

And this looks more like a transaction in the traditional sense, where you're being more explicit about how you do the thing.

You're like, I want to buy this much on this algorithm, whatever, from this source.

Whereas the market maker that principles risk, like in fixed income, is much more intent-like, where the client just wants the thing and the market maker takes the risk at how they get them the thing.

Well, why do you think, why are those like markets so different?

I mean, is it just because bonds are less liquid?

I mean, obviously treasuries are very liquid.

Like why do those markets converge to different kind of architectures?

So first of all, treasury markets are by many, I don't know exactly how you measure this, but they're considered 1 of the most deeply, deepest, most liquid markets in the world.

So it's not like you need a traditional equity exchange to have deep liquid markets, you actually can get these in a principle based market like the treasuries market.

I think some of the reasons why this existed is like there are many, many types of treasury bonds.

Like they issue a new 1 every month or every 3 months.

And so you kind of have the most recent tenure treasury is the on the run 1 that people transact liquidly.

And then you have this tale of 9 and three-quarter-year-old bonds and nine-and-a-half-year-old bonds, and those trade a bit differently.

So there's just historical reasons why I think this developed that way.

Whereas equity markets where it had these like, you know, the stock exchange pit, and then that got electronified and that had an electronic matching engine on it And they kind of went in different paths.

So, and I also get like, I guess the treasuries and bonds are generally more like institutionally facing versus like, you know, equity markets have retail in the mix sometimes.

So maybe that's, that's part of the story.

But anyway, like back to the, to like example.

So like, what was the flow for like, It becomes this like prime broker flow, right?

Where you're laying it off to Goldman or whatever to actually make this trade that you want to make happen.

What happens behind the scenes and why is it like, how does the asynchronousness of that transaction like, you know, come to bear?

Well, so, okay, let's like, what would a typical, let's say BlackRock has their, you know, bond fund, and they want to go and buy a hundred million of some bond.

What could this look like in crypto is I think a useful thing to go through.

So there was a bit of technology, it's called TradeWeb, where BlackRock would type in their order and they could pick 5 dealers.

And we can talk about how weird this all is, by the way, like why this all existed.

But BlackRock could pick 5 dealers that they would go and show their order to.

And their order is like, where can I buy a hundred million of the 10 year treasury?

And those 5 dealers all type in a price, hit enter.

The dealers have to hold their price for like 2 seconds.

And if BlackRock executes on 1 of those prices within 2 seconds, that trade is now done.

And then the dealer that does the trade, they now have sold 100 million 10-year treasuries to BlackRock.

And it's up to them to figure out what they do next.

So do they go and buy them back from the interbroker dealer market?

Do they sit there short and try to cover whatever they want to do?

It's their risk and they warehouse it now.

So, turns out this is actually a pretty efficient way to do business.

It's like if you put 4 or more people in comp in competition, you will get a sharp price.

And I generally speaking think Some of the research suggests that this is like a fairly efficient way of transacting.

And the advantages over an exchange here is like in the exchange, if it was like equity market exchange or whatever, like there might not be a hundred million treasuries at that price.

Here you have a specialized role, which is the market maker that really is trying to be this intermediary, like a useful middleman that is providing you the best liquidity at that price.

And there shouldn't be excess profits because they are competing for the right to do this.

First off, does that flow make sense to you guys?

Yeah, but I mean, would this be qualified as like Kind of just an RFQ, like

I mean, there's a little bit of nuance around it.

It's an RFQ where there is a commitment to like holding your price.

You're not, at this time, there's a wire time where you hold your price for 2 seconds.

So first of all, BlackRock can only put 5 dealers in comp on this piece of software called TradeWeb.

Well it's because the dealers all got together and they're like, well, we don't want them to talk to more than 5 people at once.

It's just a weird made up rule that the dealers sort of conspired when they were first into, you know, like introducing this piece of software.

I mean, there's information leakage here.

So now the dealers might not show their best price because they're worried that the 4 other dealers that aren't them that saw it now know what the trade was and might try to punish them.

And as a market maker, this really did happen, right?

If somebody showed too good a price, I would try to hurt them in interbroker dealer market, stuff like this.

So there's like, but if you go back and you think about crypto, you're like, well, could we design a better version of this where there is no barrier to entry, where like anyone can compete to fill this order.

And could we also design a market where maybe in the future, you don't actually see what price they got filled at immediately.

There's like some, maybe not all the information is communicated all at once.

We do some cool cryptography, 0 knowledge thing or something else like that.

And then seems like we're going down the path of like, okay, wait, we have, you know, a decentralized system that could implement this protocol, this like back and forth RFQ protocol in a very efficient way.

And turns out, people just started using the word intense to describe this.

Yeah, I mean, it seems to me like the key to sync this is like the asynchrony of it, right?

Versus just like a synchronous transaction and then intense or async because you're kind of like saying up front, I want X to happen, ideally, can you do it for me?

And there's some communication over the wire for that.

Just going back to what you're talking about with the 5 participants, I don't know if there's much research about this, but is there like a kind of a Dunbar's number equivalent of like X, you know, you need X dealers for this to be efficient and it can't be more than X.

Otherwise someone's gonna defect and start like hitting the other direction.

Now I need to, I have a very, 1 of my very good friends is a professor of economics and he's the guy that I go to for some of this stuff, but I don't have a source in front of me.

Maybe I can get the source and we can put it in.

They've studied this and there's assumptions being made.

But What I'll tell you from my own experience is it is pretty remarkable how competitive like 3 or 4 people will make things.

You really do have to collude to protect your excess profits.

And collusion is actually quite difficult to do in these contexts.

People have accused Goldman Sachs of colluding and well, it never happened.

Goldman was not a Libor participant and I didn't stay away from that.

You have to have a pretty fucked up market structure to enable easy collusion.

So yeah, man, I think you have 4 people competing and you get really sharp pricing.

But it's worth going back to your earlier point, right?

This all makes sense because there's this asynchronous thing and there's like gaps in time.

And why I think intents are of the moment is purely because we've moved into this multi-chain world.

And now there is a synchronicity between different chains in a way where there wasn't in the single chain world.

And so I actually kind of think intense in a single chain world don't really make sense.

Like you in a single chain world, you want everything to happen atomically.

It's just got really nice features and properties to it.

So you, you can't do this back and forth kind of negotiation market maker thing if you are doing atomic transactions where things are done and there as a group or not at all.

And I kind of don't, I think it's lost a little bit in the dialogue that like intense isn't a new concept.

But it didn't make sense a year ago or 2 years ago, because we were in a single chain environment.

And then you do want to do things atomically.

I think like intense on a single chain environment, you know, the TLDR and like a simple level, I just think, you know, I'm trying to sell a million dollars worth of CRV, I send it to CalSwap, they're processing all my transactions, they're trying to give me the best quote.

That's kind of how I think about it on a simple level.

You're kind of giving that, you know, permission on a signer level to multiple things, could be other than the swap.

When you say multi-chain, What are we unlocking here besides, I don't know how to say it, the cross-chain, like, I mean, we're talking swaps, can we like prove collateral outside of a single chain?

In a single chain world, the average normie user prefers, you do on-chain things that happen instantly, you can pose them, it's done all together, not at all.

Cowswap, I'd argue, is like less normy because wait, I'm going to get better execution by accepting a delay.

So I'm gonna like go through a process, send an order.

And then in this in cowswaps case, wait for things to get batched together and then wait for execution.

And I get better, better execution for it.

Like there's sort of like, and this is where stuff gets a little bit nuanced, but I send my orders on chain and I get my fill on chain, but there's kind of an off-chain solver that figures stuff out here.

In a multi-chain world, when I want to do things across chains, I don't have the option of doing things like Uniswap v2 does, where I atomically execute.

Fundamentally, these 2 chains are asynchronous.

Let's ignore shared sequencers and that line of research for a second.

And so I think the fact that this is becoming kind of like going to be the new normie case where I am communicating between chains, it means that my new default is asynchronous and there is sort of an off-chain component.

There's some component that is messaging between these 2 different systems.

How can we guarantee kind of, if we're talking about just the normal swap and the intense case, without the off-chain computation to basically separate either malicious actors or best execution, who is actually solving that in this case?

How do we decide the best, whatever, processor of my intent, right?

Well, yeah, and then I think it gets interesting.

So this is where I think the finance angle and like a market maker, I think it can be clarifying, but also different from the way a lot of crypto people think.

So generally speaking, the crypto mindset is there is an answer.

There's like a deterministic way to figure out the best answer.

And that's also often why you hear the word solver.

It's just somebody solved the best answer and like proved it deterministically.

The angle, and I think that's true in a sense, Like in there are use cases where you're just like, here's the set of rules we need to follow, and somebody solved them.

And we now have the answer, and we can prove it.

The other sense is that there, I can take a more of finance mindset and this is where I think trading and like risk can be useful where there is no 1 answer.

It is literally just best price and in this case, fulfilling my intent is building a system that is designed to give the user the best price.

And That generally looks like an auction.

Doesn't have to, but generally looks like an auction where we are going to auction off the right to fulfill that intent and the winner of the auction paid the most to the user.

I mean, I feel like we see a little bit of this.

It was like, I think the roles are getting quite big, obviously, but they're still developing and there's not a ton of like frosting yet.

But I do think we see it like basically, if you imagine finance as a chain and like e-development as a chain, Like we see a lot of examples of like this type of behavior happening on there, like just yesterday when like, you know, someone sold a bunch of SYN and then, you know, absorbed it and then goes and sends it to Binance.

We do see some of that already in the inside the MEV market and kind of inside of FlashBot bids, right?

So this is where, I mean, This intense model checks a few boxes that are kind of relevant at the same time, which is why I think it got a little bit of a hype boost behind it.

We can also do like protect MEV things or kind of extract MEV things because we run an auction for the right to execute this intent which allows us to capture the excess profits that would otherwise be going to a validator and kind of reroute them back to the original user.

Yeah, so I think for those reasons it was like it's kind of again a simple idea, but it makes sense at this moment in time.

The stuff that again is just different and weird is there is this kind of off-chain component to it.

So it feels a little more centralized, I guess, in a sense.

There's kind of like a spectrum component here where the systems all function in a decentralized way.

But if you're having a relatively limited set of specialized actors be this market maker function, it feels like we've kind of in a sense gone backwards and now have a kind of a vector of centralization here.

Right, I mean, that's what looks like a really nice property of AMMs, right?

In that they have like market maker liveness, like better market maker liveness.

Because like you saw when FTX went down, Serum just went dark, right?

Because they were just the entire market making set.

And things like that are obviously more likely to happen with a smaller group of participants.

But then you kind of have the other side of it.

It's like with some of the MEV games, you have so many participants or so many people that are willing to try and pick up pennies that you kind of get worse outcomes.

You always get sandwiched and things like that.

So AMMs, Uniswap V1, V2 style AMMs, X times Y equals K, Capital inefficient, but always on maximally decentralized.

And if you go back to the problem of if you wanted to market make like a low liquidity, infrequently traded thing before the X and Y equals K AMM, you like had to hire, you could run algorithms or whatever, but you needed people watching this market-making service.

And then there was this sort of invention, it's like, oh, we use this always-on thing.

It has some trade-offs, it's capital inefficient, but it will always provide a price.

And that was like really cool and really useful and felt really interesting.

And yeah, well, like I kind of agree that now if we are doing if we are having specialized roles that enable cross-chain transactions or fulfilling intents or whatever else.

We are kind of taking a system that worked with no...

It was beautiful in the sense that it just worked chain, all the computation was done on chain.

And now we have these other actors that we're dependent on and that feels a little unsatisfying, even though I would argue that if you have 3 barriers to entry or like no barriers to entry to becoming that role and you have a properly designed economic incentive, it's good because you will have people show up to do that.

The AMMs being online all the time, good for some bad for others, right?

In terms of intense, do you think I think a lot of what we've been talking about has been majority swaps?

That's where I sometimes feel a little like mid-weighty where I reduce everything to kind of like finance terms.

It's like, okay, this intense stuff all makes sense to me when we're talking about orders and like trading things, including like, okay, NFT is going to add NFT, like fulfilling something there.

But I don't have a good answer as to like what the other innovative or useful use cases here are outside of like your financial, I wanted to buy or sell things.

I mean, I naturally agree with you, but I'll try and take it to the other side and see if you like my show.

So imagine all these things hedge funds want to do, right?

It's still kind of a trade, but there's certain types of trades that you couldn't really get autonomically, right?

Say, I want total return swap on some stock or something, that's some derivative or something.

The prime broker in that case is this intense agent.

Do you think that's the end game, a potential end game where it's an intense entity and becomes this basically some sort of prime broker risk manager system that like, that sells these kind of more higher level derivatives trades?

Sort of, but again, it still feels very financially, right?

Like, And I feel like there could be more, but I just haven't seen it yet.

So, okay, base level, I want to buy or sell something.

I wanna buy the most frog coin that I can with 10 each, right?

And then you can kind of like add features to it, like the price has to be at least this good.

And then I could maybe have like a delivery component and I want you to deliver it to this other chain over here.

Maybe I could have some intents out there that are kind of perpetual, that they can hang out.

weighted between 60% ETH and 40% FROG coin.

Keep rebalancing me every time we're out of whack by more than

But again, I feel I just keep stuck in like, financial.

I also like 1 of my concerns about these like layering of intense is that it I get concerned about MEV in these contexts, because the fewer people that can fulfill this order, the more likely you are to get destroyed on the trade and to get a bad price.

Once you start sacking things, I get concerned that like more MEV enters the system in that way and that you get like, you know, single bidder type scenarios that could be potentially bad.

Obviously if you layer on your risk, you know, your like actual intents correctly, that's not super possible, but it is something that I've thought about a bit.

Yeah, I think there's other people that are far more informed than me on this, I think, but to the extent that there are like generalize searchers, or people writing searcher algorithms that are like super sophisticated, where they don't even know, they're just figuring out the best way to do things in a general sense.

See all Unknown transcripts on Html5mediaembed

Intents with Hart Lambur - The Parsec Podcast (1)