Infinite Games - Stephane Gosselin
Founder @ frontier.tech
𝕏: @thegostep

Here we get to the real content of the day.

This will be a quick presentation, 50 minutes.

So I wrote a blog post last week about Infinite Games and it sort of reflects a lot of my thinking over the last year.

I think we had a beautiful event where I know quite a few of you here today were attending last year in Amsterdam and it was sort of the initial place where we got to meet everyone who's building in the space.

I think at the time, there was very few other sort of companies aside from FlashBots that were really building infrastructure, and it was mostly a place where we saw a lot of a lot of searchers come together and discuss some of the The ideas and and and trading strategies or insights about the space It feels absolutely mind-blowing how far the space has come from a year ago to today I sort of mentioned it earlier in the day, but you know we had a huge amount, we had thousands of applications for the event today and we had to handpick from these and we made sure that basically every single person in this room is touching MEV in their day-to-day job in some way.

So I think it's really exciting to have seen such an industry flourish.

It's something that I was really excited to be able to help put together.

a little bit about how my thinking has evolved since then.

I had put together last year that aimed to be a bit of a prediction of how the space would evolve.

I think it introduced and hit on a few new ideas.

1 of them was the builder role, which didn't exist at the time when this was produced.

And that's been sort of something that's blown up in the past few months in terms of attention.

and the exciting prediction that I had was that we'd have a diversity of different approaches to implementing these things.

bit of a show of hands in terms of the crowd that we have today.

Which part of this stack would you say that you are sort of working on?

So maybe we'll just start at the very beginning of it with the intent layer.

Show of hands, who in the room is working on the intent layer and the expression of user preferences?

Who is at the wallet application layer and actually building interfaces for users to help them interact with the chain?

Wow, we actually only have a couple application layer people in the room right now.

Who is working at the searcher level, actually doing searching, market making, any of these games?

All right, we've got quite a few, actually.

Where were you guys when I was asking for people

We have quite a few block builders in the room right now.

And then finally, Who's at the validator layer actually helping provide custody and services?

Who's at the protocol design for base layer?

So we still have a good range of the supply chain that's represented today.

The framing that I had put on this on my on my previous talk was the idea of this dichotomy between utopia and dystopia and the direction in which monetary flows would go in this supply chain.

I think this ended up being the wrong framing.

I think in as much as we did spark a lot of innovation over the course of the last year and new teams that entered the space, I think a lot of the conversation didn't go in the direction that I wish it did.

It was a lot more focused on how do we sort of distribute the value that's created?

Who gets to capture the amount of value our user is getting?

You know a good deal or a bad deal Our validator is getting a good deal or a bad deal rather than I think the more important question Which is Are the systems that we're building actually useful for the world?

Are they things that are improving upon the status quo?

Are they things that are going to actually be able to drive meaningful value to the crypto ecosystem as a whole and deliver on this promise that we have as a space to do things better, to build better mechanisms.

And I don't think that that's been really the emphasis that the MEV space and the MEV infrastructure has had to date.

And so part of the event today is to get everyone in the room together, have discussions about what they're building, and realize that really, instead of looking more like a supply chain, what we're building is a lot more complex.

There's a lot more back relationships between all the different items.

What this really is is it shows that you can have a diversity of different experiments and different mechanisms that compose with each other and compete with each other in the creation of better systems.

We've seen a lot of these just emerge in the last year, and so I have no doubt that a year from now this diagram will be outdated and a lot more complex.

But I do think that this represents much better what we're actually trying to achieve here, which is not just a single way of getting transactions included on-chain, but rather a way to fulfill different needs that different counterparties have and specialize in the production of those, in the servicing of those needs.

I think framing the transaction supply network as being an optimization not just of who gets the value.

But rather framing it as, how does the entire ecosystem as a whole become useful in a way that maximizes the value across all of it should really be the optimization function that we aim for.

And so what does this actually end up looking like?

I think it ends up looking like we have a bunch of different teams that are building different games that are composing with each other in a way that aims to provide more value.

The games are competing with each other for players and for users to submit either transaction flow, liquidity, information, preferences into their system.

Whichever is the game that's able to better service those preferences is likely the game that's going to win.

It's also likely the game that's able to generate the most value for the ecosystem as a whole.

And so I really want us to think about, OK, how do we optimize the entire system and the value the entire system creates as an ecosystem rather than linearly who gets

In thinking about, okay, what do these games look like and how do they compose with each other?

I think 1 simple abstraction that we can bring to mind is the fact that all of these structured games are composed of 2 very simple things.

They're standard interfaces for messages that are being communicated between different parties.

And they're an optimization function that these parties are competing against each other on to find an optimal solution.

So whether we're talking about an order flow auction, whether we're talking about a protocol layer system, like MevBoost, or an intent system, they're all instantiation of a similar game that are being played between these parties.

1 of the things that I've come to realize is while there's some sort of similarities between a lot of these games, there's also a lot of benefit from specialization.

So it's likely that the games that end up being the most useful are the ones that are the more locally tailored to the users that it's trying to service.

The goal is not necessarily to just generate and deploy a single game that captures all of the benefits or all of the preferences of the end users, but rather develop games that are specifically tailored for the narrow use cases that a subset of users want.

1 of them is you better service the user and you create more value for them.

The second 1 is that it's actually much easier to achieve decentralization.

I think 1 of the sort of underbelly and dark truths about the design of these sort of games is a lot of them devolve into political games.

So I'm gonna talk a little bit about these difficult topics and difficult lessons learned, I think, from being in a position of having designed some of these games in the past.

I think basically all consensus games are a political game by definition.

So consensus games are, in the most abstract, a collaboration of people that come to agreement on something.

And the tools that they have for doing this are basically sharing information, whether that's through language or through Ethereum transactions or through protocol consensus rules.

These are informations and bits that are being shared from 1 party to another and that information sharing also expresses contention.

These contentions end up being resolved through mostly political games.

Now, I have some abstraction of what political games actually means here, which is a game that benefits from information access.

And information access I sort of define in 2 ways.

1 of them is through, I should say, information advantages.

Information advantages get expressed in 2 ways.

So if you are part of a closed source of order flow and you're able to benefit from that, to be able to have some preferred access to doing an arbitrage opportunity, or you have some preferred access to a validator set that others aren't able to access.

And you're therefore able to sort of win the political contentious game of who gets to define the next state update.

Access is 1 of them and it's a trust game.

It's an elitist game that is all about sort of connections and relationship.

So not only the access to information matters, but the speed at which you get it.

Whether you are just 5 milliseconds before someone else on the access to an arbitrage opportunity or in a more slow political game, sort of can see way ahead of time, you know, what are the next protocol changes that are going to happen on Ethereum, and be able to start investing in your business to be able to be ready for those protocol changes.

The latency at which information gets propagated does define the winners and losers in political games and that huge that is sort of a cause of Centralization at the end of the day and so I do want to emphasize that these games happen at many different scales right both time scales and and sort of information size scales.

So every single block, every single transaction in a block is a game that's being played, whether it's an auction or something else.

But then also, every single change that gets introduced to Ethereum as a protocol or a new company that sort of spins up and creates a game are also changes to the game.

And the information access doesn't stop at a transaction level.

The information access and the advantages that certain parties have exist at these much more meta layers.

I think 1 of the lessons that I've also learned is the idea that the more global a game is, the more political it'll be and the more centralized it'll end up.

The more information and the more contention that you have in a game, the more you're trying to represent the state of the world within a single abstraction, with a single price discovery layer, the more competition that you'll have for being able to win that game and the more people are going to be willing to invest in developing those advantages.

So this is sort of a blessing and a curse, right?

Ethereum as a protocol has amassed a huge amount of value and is sort of the home of DeFi, but also by proxy is sort of the biggest place where all the players are sort of competing on latency and being able to control parts of the stack.

If we're able to fragment these games into smaller games, we're able to sort of isolate the amount of centralization pressure and political pressure that there is on those systems, and ultimately create redundancy in a place where there otherwise would not naturally be redundancy.

So instead of thinking big, maybe the solution is actually to think smaller and try to think about, okay, what are the more niche games that we can design that are not great for everyone, but are protected from being influenced by the outside world.

I think as a part of this is also the question of monolithic and non monolithic game design.

This is not necessarily something or like language or terminology that's being used a lot these days, But if you have a monolithic system to me that means a system that Benefits from network effects typically through the form of price discovery so if you have a system that aggregates a bunch of different orders and benefits from the fact that you have to go and trade on this venue, or you have to send your transactions to this venue to be able to benefit from the best price discovery, you're encouraging and sort of developing a more monolithic system.

That makes it harder for external parties to spin up and sort of transfer over their usage to a different system.

The bootstrapping cost in developing a new game increases the more that these systems are designed in a monolithic way relative to non-monolithic system design, in which it really doesn't matter where you send the flow.

So maybe 1 example here is a non-monolithic system design would be something like MevBoost.

So MevBoost or PBS as a system, the price discovery doesn't happen at the block builder level.

It happens at the validator level, which means that although it's non-trivial for a new block builder to get set up and get going and aggregate the flow, it's still trivial for them to actually start participating in the MevBoost system and submit openly to new validators, just like it's sort of open for any new relay to come in and plug into the system or a new new validator to come and plug into the Mavbooth architecture, there isn't sort of this shared global state of price discovery that you need to tap into to be able to participate in the game.

And so part of what I want to encourage is also the development of more of these non-monolithic system design.

And when we're evaluating the games that we decide to play.

It's a lens through which we should consider what are positive some games and what are negative some games.

I'm going to end up with talking a little bit about Frontier and the reason why we put this event together, the reason why I decided to start a new company in this space and continue working here.

I think the reason why I joined this space, the reason why I got excited about crypto as a whole when I started back in 2017, was the design of games.

It was creating games that create more value for the end users that we're developing for.

And the goal for us is to help the vast majority of the people here who are either designing or playing these games think about the design that they're making.

And so on the Frontier Research side, we publish articles and mental models, I'm sure quite a few of you in this room have read, that I think are useful for being custodians and designers of these games.

And we also take advisory engagements with a lot of the teams here and work with them on how to think about, are these games good?

How do we build a business around this kind of game design?

And I'm super excited to be able to continue doing that.

time, we're also developing an adventures portfolio, which is a set of efforts and incubated projects that aim to support various different teams in the transaction supply network.

And so the initial project that we were bootstrapping is called the Faith Builder, or the F1B Builder.

And it's a generalized sort of solver that aims to plug into all these different games and help them get bootstrap and provide better experiences.

So right now we're participating in a block building game.

And it's been sort of fascinating to be able to meet people and get to understand the side of the player in the games that are being designed here.

And so a lot of really interesting lessons that we'll be keen to share with anyone who's interested in learning about it.

I'm really excited for everyone here to get to chat with each other and foster in the next year of MEV.


Speaker 0: All right. Hey, everyone. Here we get to the real content of the day. This will be a quick presentation, 50 minutes. I wrote a blog post.

Speaker 1: Oh, actually, I have a mic.

Speaker 0: I don't know why I'm holding this 1. Alright.

Speaker 1: Can everyone hear me okay?

Speaker 0: Yeah? Perfect. So I wrote a blog post last week about Infinite Games and it sort of reflects a lot of my thinking over the last year. I think we had a beautiful event where I know quite a few of you here today were attending last year in Amsterdam and it was sort of the initial place where we got to meet everyone who's building in the space. I think at the time, there was very few other sort of companies aside from FlashBots that were really building infrastructure, and it was mostly a place where we saw a lot of a lot of searchers come together and discuss some of the The ideas and and and trading strategies or insights about the space It feels absolutely mind-blowing how far the space has come from a year ago to today I sort of mentioned it earlier in the day, but you know we had a huge amount, we had thousands of applications for the event today and we had to handpick from these and we made sure that basically every single person in this room is touching MEV in their day-to-day job in some way.

So I think it's really exciting to have seen such an industry flourish. It's something that I was really excited to be able to help put together. And

Speaker 1: I want to talk

Speaker 0: a little bit about how my thinking has evolved since then. So a bit of a year of learnings.

Speaker 1: So this is a graphic that

Speaker 0: I had put together last year that aimed to be a bit of a prediction of how the space would evolve. I think it introduced and hit on a few new ideas. 1 of them was the builder role, which didn't exist at the time when this was produced. Another 1 is the idea of intents. And that's been sort of something that's blown up in the past few months in terms of attention.

Speaker 1: Because the exciting thing

Speaker 0: and the exciting prediction that I had was that we'd have a diversity of different approaches to implementing these things.

Speaker 1: I want to have a

Speaker 0: bit of a show of hands in terms of the crowd that we have today. Which part of this stack would you say that you are sort of working on? So maybe we'll just start at the very beginning of it with the intent layer. Show of hands, who in the room is working on the intent layer and the expression of user preferences? All right.

So we have a handful, 5, 6 people. Who is at the wallet application layer and actually building interfaces for users to help them interact with the chain? Don't be hype. Put your hands up. OK.

Wow, we actually only have a couple application layer people in the room right now. Who is working at the searcher level, actually doing searching, market making, any of these games? All right, we've got quite a few, actually.

Speaker 1: Where were you guys when I was asking for people

Speaker 0: to come on the searcher panel? No 1 messaged me. What happened? All right. Who's doing block building?

Put your hand up. We have quite a few block builders in the room right now. Nice. And then finally, Who's at the validator layer actually helping provide custody and services? Perfect.

Who's at the protocol design for base layer? Beautiful. All right, cool. So we still have a good range of the supply chain that's represented today. The framing that I had put on this on my on my previous talk was the idea of this dichotomy between utopia and dystopia and the direction in which monetary flows would go in this supply chain.

I think this ended up being the wrong framing. I think in as much as we did spark a lot of innovation over the course of the last year and new teams that entered the space, I think a lot of the conversation didn't go in the direction that I wish it did. It was a lot more focused on how do we sort of distribute the value that's created? Who gets to capture the amount of value our user is getting? You know a good deal or a bad deal Our validator is getting a good deal or a bad deal rather than I think the more important question Which is Are the systems that we're building actually useful for the world?

Are they things that are improving upon the status quo? Are they things that are going to actually be able to drive meaningful value to the crypto ecosystem as a whole and deliver on this promise that we have as a space to do things better, to build better mechanisms. And I don't think that that's been really the emphasis that the MEV space and the MEV infrastructure has had to date. And so part of the event today is to get everyone in the room together, have discussions about what they're building, and realize that really, instead of looking more like a supply chain, what we're building is a lot more complex. It's a lot more convoluted.

There's a lot more back relationships between all the different items. And this is a great thing. What this really is is it shows that you can have a diversity of different experiments and different mechanisms that compose with each other and compete with each other in the creation of better systems. We've seen a lot of these just emerge in the last year, and so I have no doubt that a year from now this diagram will be outdated and a lot more complex. But I do think that this represents much better what we're actually trying to achieve here, which is not just a single way of getting transactions included on-chain, but rather a way to fulfill different needs that different counterparties have and specialize in the production of those, in the servicing of those needs.

I think framing the transaction supply network as being an optimization not just of who gets the value. Is it users? Is it validators? Is it middlemen in the middle? But rather framing it as, how does the entire ecosystem as a whole become useful in a way that maximizes the value across all of it should really be the optimization function that we aim for.

And so what does this actually end up looking like? I think it ends up looking like we have a bunch of different teams that are building different games that are composing with each other in a way that aims to provide more value. And they're competing with each other. The games are competing with each other for players and for users to submit either transaction flow, liquidity, information, preferences into their system. Whichever is the game that's able to better service those preferences is likely the game that's going to win.

It's also likely the game that's able to generate the most value for the ecosystem as a whole. And so I really want us to think about, OK, how do we optimize the entire system and the value the entire system creates as an ecosystem rather than linearly who gets

Speaker 1: to capture the biggest slice of the pie.

Speaker 0: In thinking about, okay, what do these games look like and how do they compose with each other? I think 1 simple abstraction that we can bring to mind is the fact that all of these structured games are composed of 2 very simple things. They're interfaces. They're standard interfaces for messages that are being communicated between different parties. And they're an optimization function that these parties are competing against each other on to find an optimal solution.

So whether we're talking about an order flow auction, whether we're talking about a protocol layer system, like MevBoost, or an intent system, they're all instantiation of a similar game that are being played between these parties. 1 of the things that I've come to realize is while there's some sort of similarities between a lot of these games, there's also a lot of benefit from specialization. So it's likely that the games that end up being the most useful are the ones that are the more locally tailored to the users that it's trying to service. The goal is not necessarily to just generate and deploy a single game that captures all of the benefits or all of the preferences of the end users, but rather develop games that are specifically tailored for the narrow use cases that a subset of users want. The upside of this is, I think, twofold.

1 of them is you better service the user and you create more value for them. The second 1 is that it's actually much easier to achieve decentralization. I think 1 of the sort of underbelly and dark truths about the design of these sort of games is a lot of them devolve into political games. So I'm gonna talk a little bit about these difficult topics and difficult lessons learned, I think, from being in a position of having designed some of these games in the past. I think basically all consensus games are a political game by definition.

So consensus games are, in the most abstract, a collaboration of people that come to agreement on something. And the tools that they have for doing this are basically sharing information, whether that's through language or through Ethereum transactions or through protocol consensus rules. These are informations and bits that are being shared from 1 party to another and that information sharing also expresses contention. These contentions end up being resolved through mostly political games. Now, I have some abstraction of what political games actually means here, which is a game that benefits from information access.

And information access I sort of define in 2 ways. 1 of them is through, I should say, information advantages. Information advantages get expressed in 2 ways. 1 of them is through access. So if you are part of a closed source of order flow and you're able to benefit from that, to be able to have some preferred access to doing an arbitrage opportunity, or you have some preferred access to a validator set that others aren't able to access.

And you're therefore able to sort of win the political contentious game of who gets to define the next state update. Access is 1 of them and it's a trust game. It's an elitist game that is all about sort of connections and relationship. The second 1 is speed. So not only the access to information matters, but the speed at which you get it.

Whether you are just 5 milliseconds before someone else on the access to an arbitrage opportunity or in a more slow political game, sort of can see way ahead of time, you know, what are the next protocol changes that are going to happen on Ethereum, and be able to start investing in your business to be able to be ready for those protocol changes. The latency at which information gets propagated does define the winners and losers in political games and that huge that is sort of a cause of Centralization at the end of the day and so I do want to emphasize that these games happen at many different scales right both time scales and and sort of information size scales. So every single block, every single transaction in a block is a game that's being played, whether it's an auction or something else. But then also, every single change that gets introduced to Ethereum as a protocol or a new company that sort of spins up and creates a game are also changes to the game. And the information access doesn't stop at a transaction level.

The information access and the advantages that certain parties have exist at these much more meta layers. I think 1 of the lessons that I've also learned is the idea that the more global a game is, the more political it'll be and the more centralized it'll end up. The more information and the more contention that you have in a game, the more you're trying to represent the state of the world within a single abstraction, with a single price discovery layer, the more competition that you'll have for being able to win that game and the more people are going to be willing to invest in developing those advantages. So this is sort of a blessing and a curse, right? Ethereum as a protocol has amassed a huge amount of value and is sort of the home of DeFi, but also by proxy is sort of the biggest place where all the players are sort of competing on latency and being able to control parts of the stack.

If we're able to fragment these games into smaller games, we're able to sort of isolate the amount of centralization pressure and political pressure that there is on those systems, and ultimately create redundancy in a place where there otherwise would not naturally be redundancy. So instead of thinking big, maybe the solution is actually to think smaller and try to think about, okay, what are the more niche games that we can design that are not great for everyone, but are protected from being influenced by the outside world. I think as a part of this is also the question of monolithic and non monolithic game design. This is not necessarily something or like language or terminology that's being used a lot these days, But if you have a monolithic system to me that means a system that Benefits from network effects typically through the form of price discovery so if you have a system that aggregates a bunch of different orders and benefits from the fact that you have to go and trade on this venue, or you have to send your transactions to this venue to be able to benefit from the best price discovery, you're encouraging and sort of developing a more monolithic system.

That makes it harder for external parties to spin up and sort of transfer over their usage to a different system. The bootstrapping cost in developing a new game increases the more that these systems are designed in a monolithic way relative to non-monolithic system design, in which it really doesn't matter where you send the flow. So maybe 1 example here is a non-monolithic system design would be something like MevBoost. So MevBoost or PBS as a system, the price discovery doesn't happen at the block builder level. It happens at the validator level, which means that although it's non-trivial for a new block builder to get set up and get going and aggregate the flow, it's still trivial for them to actually start participating in the MevBoost system and submit openly to new validators, just like it's sort of open for any new relay to come in and plug into the system or a new new validator to come and plug into the Mavbooth architecture, there isn't sort of this shared global state of price discovery that you need to tap into to be able to participate in the game.

And so part of what I want to encourage is also the development of more of these non-monolithic system design. And when we're evaluating the games that we decide to play. It's a lens through which we should consider what are positive some games and what are negative some games. I'm going to end up with talking a little bit about Frontier and the reason why we put this event together, the reason why I decided to start a new company in this space and continue working here. I think the reason why I joined this space, the reason why I got excited about crypto as a whole when I started back in 2017, was the design of games.

It was creating games that create more value for the end users that we're developing for. And the goal for us is to help the vast majority of the people here who are either designing or playing these games think about the design that they're making. And so on the Frontier Research side, we publish articles and mental models, I'm sure quite a few of you in this room have read, that I think are useful for being custodians and designers of these games. And we also take advisory engagements with a lot of the teams here and work with them on how to think about, are these games good? Are these games useful?

Are these games going to get adoption? How do we build a business around this kind of game design? And I'm super excited to be able to continue doing that.

Speaker 1: At the same

Speaker 0: time, we're also developing an adventures portfolio, which is a set of efforts and incubated projects that aim to support various different teams in the transaction supply network. And so the initial project that we were bootstrapping is called the Faith Builder, or the F1B Builder. And it's a generalized sort of solver that aims to plug into all these different games and help them get bootstrap and provide better experiences. So right now we're participating in a block building game. And it's been sort of fascinating to be able to meet people and get to understand the side of the player in the games that are being designed here.

And so a lot of really interesting lessons that we'll be keen to share with anyone who's interested in learning about it. So That's it for me. That's all I have to share for today. I'm really excited for everyone here to get to chat with each other and foster in the next year of MEV.

See all Frontier Research transcripts on Youtube

Infinite Games - Stephane Gosselin