GRTiQ Podcast: 37 Brandon Ramirez (Part 2)

The following podcast is for informational purposes only. The contents of this podcast do not constitute tax, legal, or investment advice. Take responsibility for your own decisions, consult with the proper professionals, and do your own research.

Brandon Ramirez (00:13):

The end vision is that The Graph is a global decentralized market of Indexers that operates efficiently, that consumers can reliably query that data from their device, and that this is a stable platform that can be relied upon for every future layer and stack that accretes on top of it.

Nick (01:09):

Welcome to the GRTiQ podcast. Today I’m picking up the conversation with Brandon Ramirez, co-founder and research and product lead at Edge & Node. And as you may already know, Brandon is one of the original founders behind The Graph. If you’re new to this podcast, then you should know that Brandon was incredibly gracious with his time, and our interview lasted for nearly two and a half hours. So I split the interview in half and created a two-part series.

Nick (01:36):

During part one, which already aired, Episode 36. Brandon talked about his corporate experience working at Microsoft, his move into entrepreneurialism, and then he provided a great backstory about the origins and early days of The Graph. During part two, which you’re about to hear, we pick up the conversation, and Brandon shares many new interesting ideas, including the story behind Edge & Node, a core dev team working at The Graph, along with the design of the protocol incentive structure and how certain stakeholder roles emerged. We finish the conversation with Brandon’s vision for the future of The Graph.

Nick (02:16):

That’s something I’ve experienced in having interviews with so many different people on this podcast, which is there is this dispersion of people throughout the world working on this huge opportunity on the hope that it will change the future, it will enrich people’s lives, but it’s all based on a shared value system. And it’s been remarkable to me that so many people who, some of which have never sat in the same room, have the same values. If you could characterize that, what are those values?

Brandon Ramirez (02:47):

Let me start by characterizing the mentality, and maybe I can back my way into some values. But I think the biggest change in mentality is going from kind of a zero-sum to a positive sum mindset, and all the sorts of decisions that you make differently when you’re in that mindset. So there were periods before the network launched where we were actively thinking about how do we give away control of the network? And how do we bring the right stakeholders in to become core developers or active participants and make Edge & Node less of a key player in the system? And that wasn’t out of some, certainly it aligned with our values, but it wasn’t pure altruism. We knew that the way that these protocols could be designed would actually make the network more resilient if we had a large and diverse ecosystem of participants and builders working on the thing.

Brandon Ramirez (03:46):

And that’s very, very contrary to the way you would make decisions in the past where it would be all around: how do I build as big of a moat as possible around the product that I built? How do I lock in users as much as possible? How do I compete with adjacent markets? So you get a foothold in one market as a SaaS product, and then you need to figure out, okay, what’s the adjacent market that I’m then going to compete against in a zero-sum with the competitors on that market? Whereas you look at something like The Graph, there’s not even really a strong reason to quote-unquote “compete” with it, because you can just join and you can build it and you can make it better. And people have built companies that are fully devoted to working on The Graph and participating in that ecosystem.

Brandon Ramirez (04:32):

And so I think the mentality is just completely different from what we’ve seen in the past, and that was very alluring to us. And then I think the shared values first and foremost really are decentralization, but everything that kind of comes with that. So all the value of web3 stuff, empowering users, giving users ownership. I think people are very committed to the changing nature of work and just how humans coordinate at scale.

Brandon Ramirez (05:00):

So we talked about scale earlier in the interview, one thing that I’m always struck by working in The Graph’s ecosystem is that I often feel like I’m working at a larger scale than my work at Microsoft. When you look at the amount of people I interact with on a day-to-day basis or week to week basis and the amount of different stakeholders and developers and builders and designers, it is a massive ecosystem of which Edge & Node is just a small part. And that level of coordination is enabled by this new primitive that we have in the blockchain in token design, crypto economics, and so forth. And so I think to the degree to which this is an experiment in human coordination at scale, I think everyone is interested in proving that this is a viable way of creating value. And so I think that’s a strong undercurrent, is just this shared belief in the way that work is going to change in the future as a result of web3.

Nick (06:02):

Another thing that’s come up on this podcast is, is this technology reflecting back its ethics, its possibilities on us? Or is the technology and expression of our optimism and our ethics?

Brandon Ramirez (06:15):

Yeah, I mean it’s a really fascinating observation. I think you could definitely say that crypto is this manifestation of these really long-term trends that preceded blockchain, preceded Bitcoin white paper by decades if not hundreds of years.

Brandon Ramirez (06:31):

So one of those big long-term trends is a shift from collectivism to greater individuality and greater individual freedoms. You see this going all the way back to some of the earliest European history where you had the individual was part of some large family structure, or you saw this in other parts of the world as well with clans. And it was actually quite a large innovation socially when you started seeing things in places like England with respect to contract law and the atomic family, this idea that instead of being part of these large collective structures, we were now part of these smaller agile units. And then you kind of play that forward a couple hundred to 300 more years, and now you’ve seen this incredible empowerment of the individual, to some pretty insane degrees in some respects.

Brandon Ramirez (07:26):

I remember as an example, I remember even 10 or 15 years ago, if I would fly into LAX, for example, to visit friends, I would call friends and say, “Hey, can you pick me up from the airport?” And it was this sort of collective challenge of, okay, how do we get Brandon from point A to point B? And then now fast-forward 2021, I’d almost feel rude asking a friend to pick me up from the airport. It’s like, okay, I’ve got this app that at the tsp of a button, I can have this car here on demand, and it can take me from place to place. And in some ways you might lament the loss of that interdependence, but what we gain in that process is sort of unprecedented choice, autonomy, and the ability to be deliberate with how we express that interdependence.

Brandon Ramirez (08:16):

So it’s not that we’re no longer dependent on one another, it’s just that we opt in to the places where we want to collaborate with one another. I think one of the things that’s so fascinating about crypto is that for a movement that’s ostensibly predicated on individuality, on sort of not trusting one another, at least the protocol level of being all about technology first as a solution to solution to problems, it’s a very human movement. A lot of people in crypto spend a good part of their year just pre-COVID, and now as life is getting back to normal, collaborating with other people in person. You see these incredibly rich and vibrant communities that have self-organized around these protocols. And so even as we’ve enabled an unprecedented level of choice in how we spend our time, on where we work, we’re also seeing a lot more intentionality behind what communities do we choose to spend time in.

Brandon Ramirez (09:18):

And that’s not just a crypto thing. I think when you look at the changing nature of work, I mentioned Uber a second ago, like gig economy and gig work is another great example of how we’ve started freeing up individuals from the default path of only working for a large organization. And now we’re saying, “Hey, you can go that path, but you could also work on an open platform where you can come in and work as you see fit on the hours that make sense for you,” which now opens up the labor force to maybe students, or part-time students, or parents or that otherwise are primary caretakers for the children that maybe have some spare hours here and there. So it’s just this new level of flexibility that I think we’ve been moving towards for quite a long time. And I think the thing that’s lagging, if anything, is the institutional context that supports that shift in the nature of work.

Brandon Ramirez (10:15):

And so we saw that a little bit with gig workers, with Airbnb hosts, Uber, Lyft drivers, et cetera. Early on, some of these companies wanted to issue equities to these gig workers kind of recognizing this changing nature of work that sort of values the primacy of the individual. And because of the way securities laws were written at the time, Airbnb wasn’t allowed to give equities to their hosts because they weren’t full-time employees. And that would’ve been life-changing amounts of money for these early supporters of these new platforms, for these early workers and these new platforms.

Brandon Ramirez (10:52):

Another way that we’re seeing that tension is with just the way that we’ve always thought about healthcare benefits and this idea that the default path that we’re supposed to go down in our parents’ generation was, “Hey, you find some company that you want to spend 30 years at, and that becomes your social safety net. It becomes your source of healthcare.” And in California, you had a piece of legislation sort of trying to re-entrench that mindset. It was assembly bill 85 that basically was trying to mandate that gig workers be treated as full-time employees, essentially taking away some of their choice to spend their time or split their time freely across these different platforms in the way that they see fit, and reinforcing this idea that there’s this really strong link between the company and the individual.

Brandon Ramirez (11:46):

But all that being said, I’m still a huge believer in teams. I’m still a huge believer in a group of people committing to a problem over a long period of time and developing strong working relationships. I just think that the fact that there are alternative methods of large scale human coordination, that there’s alternative options for individuals to have unprecedented choice in how they work, I think is just a really important check on the power and on the size and the bloatedness that can kind of come from the traditional way of building companies and scaling organizations.

Nick (12:27):

Brandon, I really appreciate all the time and thought you’ve taken into helping me and the listeners better understand the backstory with The Graph and all the design and thought that went into that. And then in recent memory is Edge & Node, which as I understand as a core dev team, a lot of the founding members of The Graph migrated over to. I’d love to know what your long term vision for Edge & Node is. And in fact, that name’s quite interesting too. What can you tell us about the origin of the name?

Brandon Ramirez (12:57):

So Edge & Node is, it’s a reference to graph theory, so it’s kind of a nerdy name. You have edges and nodes in a graph. But the inspiration for it actually came from these kind of classic craftsman style fashion brands where you have Rag & Bone, or there’s one I think called Grown & Sewn. It evokes this more classic type of brand in contrast to some of the more like futuristic and technological brands that you classically hear in tech. And we wanted to adopt this identity of protocol craftsmen while at the same time taking a step back and leaving space for more core developers to come into The Graph ecosystem. So that’s why we felt it was important for our company name to not be synonymous with The Graph and simultaneously take on this identity that would set us up to be the craftsman of more protocols in the future.

Brandon Ramirez (13:48):

So I think for the next few years, easily the majority of our time will be spent contributing to The Graph ecosystem and making the protocol better. But part of the vision of web3, we talked about the changing nature of work, is that technology accretes in layers, and technology can eventually be essentially feature complete these platforms that we can then build future platforms on. We’re not going back, and certainly we iterate, but we’re not going back and fundamentally reinventing the railroad to facilitate the stuff that was built on top of it. I mean, at a certain point, The Graph is going to provide all the functionality that it needs to provide, and you’re going to see a lot more innovation at the cutting edge on these new layers in the stack. And so already we’re trying to do what we can to support these projects and as developers of a platform that many of them are building on.

Brandon Ramirez (14:40):

But one of the things that web3 does is it separates the creation of software from its operation. And so that’s something that is sort of unprecedented. In the past, to have a sustainable business model as the creator of software, you also had to stick around and operate that software ad infinitum. So it created this weird tension where people that would maybe self-identify as creators would start these projects, and then the business would rapidly shift to one of creation and innovation to one of maintenance extraction operation. And part of that comes from the fact that, the sort of perverse incentives we discussed before. Part of it also comes from the fact that software wasn’t permanent before. We didn’t have software or technology layers that were permanent. We had this sort of obsolescence that would happen if you didn’t stick around and operate the thing that you created.

Brandon Ramirez (15:39):

And that’s pretty unique to software with respect to information products. If you think of software as an information product and you compare it to other information products like books, movies, audio, songs, like yes, the media that these information products are written onto can decay, right? Like a book can dissolve or get torn, or a file on a computer can decay, but that the information artifacts themselves never go bad. It’s not like the movie studio that produced Jurassic Park has to go back and keep reproducing Jurassic Park or else it won’t work. Well, I mean, Spielberg does like to do that, but it’s not because he has to. And so what that does is it allows companies to emerge that focus primarily on the act of creation and to always continue focusing on where the act of creation is going to have the highest leverage.

Brandon Ramirez (16:34):

And that’s, in the long term. I think what our hopes are for Edge & Node, is that we can continue being creators, being builders at the layer of the stack where the most value to decentralization and to end users can be delivered. And then as part of the genesis of these new protocols and layers, handing over control of the operation to the community in the same way that now today the Indexers run The Graph. We are not the operators of the network. And so I think we’re several years away for our focus shifting in any meaningful way, but that’s kind of the long-term vision behind this new identity and this new company.

Nick (19:37):

And so the idea of The Graph is hatched. You write the white paper, and my sense is that you guys get your feet underneath yourselves as you’re kind of carrying this new idea, this project into the marketplace. But then you have to build an ecosystem with a bunch of actors that are incentivized. And I know that you were kind of the brains behind that. So how did you understand or architect early on the ecosystem, the Curatorss, the Indexers, and what was that process like for you?

Brandon Ramirez (20:06):

Sure. Yeah. So in retrospect, you can always tell a story that seems more methodical than it probably felt at the time, which was a little bit more exploratory, and kind of taking one path and then taking a few steps back and a few steps forward in a different direction. But at least in the broad brush strokes, the place that we started was just outlining what were the essential behaviors that needed to happen in order for The Graph to deliver a service of querying this blockchain data that was comparable to a centralized SaaS product. So that’s always been our, I wouldn’t say it’s our benchmark because we want to be much better than that, but it’s always been a constraint from day one. Knowing that we would be compared to centralized SaaS products meant that we had to go for scale, go for some level of feature completeness from day one.

Brandon Ramirez (20:59):

And that was kind of the first invariant, is like this is where we need to start. And once we had all those behaviors outlined, and who are the users, or who are the roles that would be engaged in those behaviors, we kind of took an approach that I would describe as even more micro than microeconomics, in the sense that microeconomics traditionally deals with these idealized incentives and rationality of the economic agent. Perfect information, perfect rationality, unbounded computational ability. And this is I think where my experience in UX design actually kind of came forward a lot. One of the models that we used to use in UX was this model called the BJ Fogg Behavioral Model. I believe he was a researcher at Stanford. And it was basically this, it’s really more of a heuristic and it goes like B equals MAT, which translates to behavior is equal to the product of motivation, ability, and trigger.

Brandon Ramirez (22:05):

And so one of the things that crypto economics and blockchains gave us early on was this ability to get really, really fine-grained with incentives. But incentives just kind of cover the motivation. When you’re actually designing the full system, it’s not enough to just define the on-chain incentives that are expressed in the smart contracts. You also need to think through, okay, what is the ability of users to actually respond to those incentives? What are going to be the triggers that actually elicit them to take action in response to that motivation, in response to those abilities? And that’s where things like UX design, that’s where things like just basically the products and tooling that we built around the network came into play. Things like the indexer agent, things like The Graph Explorer, the subgraph studio, the different various CLIs that enable engaging with the network, as well as tools that weren’t built by us.

Brandon Ramirez (23:02):

There’s many other community explorers and dashboards that kind of actually act as this really important translation layer of turning those incentives into a real motivation and understanding of how to respond to them. So that was one broad philosophical kind of underpinning with how we approached this. But then kind of looping back to the protocol economics, one thing that was clear to us that became evident really early on, which is the complexity of The Graph when you compare it to some other protocols, I mean kind of complexity in the sense of many interacting components that are difficult to model in a really closed form way. Blockchains are incredibly technologically sophisticated and deep in terms of how they act, and there’s a huge line of consensus research. But if you look at the proof that was in the Bitcoin white paper around why Nakamoto Consensus was secure, it’s actually pretty reasonable to understand in sort of a closed form way.

Brandon Ramirez (24:03):

Whereas it became clear to us early on that what we were designing at The Graph was much more of a system of interacting agents, personas, and incentives. And so the first step was defining what those personas were. At the very least, we knew we had a consumer and we had a service provider. So the service provider here being the indexer. And that service provider needed to make that service available to consumers in a efficient way, like an economically efficient way where supply is perfectly met by demand at an optimal price. And the best sort of precedent that we have for that is markets. A well-designed market that’s open, has low barriers to entry, low transaction costs, where there can be efficient price discovery, where there’s no asymmetric or very little asymmetric information. Those are the ways, that was our initial instinct for how can this service be made available to consumers at the highest quality for the lowest price.

Brandon Ramirez (25:03):

A corollary to that was like, well, okay, so we have these markets, well, how do we bootstrap them in the first place? So a well functioning market already assumes some level of market depth or liquidity, that there’s already actors on both sides of that market that want to engage in some exchange. And we didn’t have that at the time. And so that’s kind of where the curation mechanism came into play. It was, okay, what would be the information signal between either consumers or developers or some other ecosystem participant to say, “Hey, this market for this particular subgraph for this data set is useful to bootstrap”? And that’s one thing that’s important to note with the market design is that despite there being a global set of service providers and consumers, the actual market really is around the specific data sets, right?

Brandon Ramirez (25:53):

Because if you’re an indexer on one data set, you can’t serve consumers that are interested in data on another data set. So you really need to think about demand and supply being bootstrapped at the data set level. And that’s kind of where curation came into play, we’re Curatorss. And then a whole set of incentives kind of fell out of that. Why would Indexers trust the signal from Curatorss? How does the price, we ended up choosing bonding curves for curationism, I’m sure you know. How does that price actually have any meaning to Indexers? Is there efficient signaling that’s happening between Curatorss and the rest of the network?

Brandon Ramirez (26:31):

Other concerns that came up were, how do you guarantee the… So another part of market design is standardizing the products. So when you go to an exchange, like a commodities exchange, you don’t know where that barrel of oil or bushel of wheat is coming from. These exchanges have already taken care of standardizing in such a way that when you get some bushel of wheat, you know that it’s essentially interchangeable with all these other bushels of wheat. And the corollary for our protocol design is kind of determinism and having a well specified interface between consumers and Indexers.

Brandon Ramirez (27:09):

And so that implies not just defining what type of data you or what data you get back when you send a specific query, it also implies figuring out a way to standardize costs, or at least express the standardized way of expressing costs or prices. And it also implies, because this market is no barriers to entry, literally anyone can come in and participate in this market, it also implies that we need to have a strong underpinning of verifiability. If a consumer can’t verify or trust that the data that’s coming back is correct, then no consumer is going to participate in this market.

Brandon Ramirez (27:47):

And so that’s where arbitration came in, the role of the arbitrator. That’s where some of our lines of long-term research, which I’m happy to get into later, are around verifiable indexing, verifiable queries, have come in. And that’s where economic security also came in. This requirement that Indexers have to stake graph tokens to participate in the market so that if they serve some incorrect data, that there’s a penalty for that. So the consumer gets this guarantee that if an indexer lies to them, they’re subject to be slashed by some amount.

Brandon Ramirez (28:16):

Now there’s a whole but a bunch of devils in the details on what makes that mechanism work. For example, we have the Cobb Douglas mechanism, which part of what allows the token to play that role in economic security. And so that’s where we start getting into more mathematical definitions of the system. And that’s more maybe traditionally in the realm of economics or game theory. But it started with these kind of qualitative requirements around, well, what does a functioning market need to look like? What are the information flows?

Nick (28:45):

When you’re architecting this early on and you’re thinking about the ecosystem and the different actors, are you doing something which seems as trivial as building a spreadsheet and modeling out the different incentives? Or is this something far more complex, game theory, where you’re just trying to reason away to a market where everyone can participate?

Brandon Ramirez (29:05):

Yeah, so I would say it’s both. So early on, yeah, it did look a lot like a spreadsheet. Or you could imagine just an exhaustive table or bulleted list of what are the behaviors, what are the roles, what are the incentives, filling in every single gap. And then the challenges I think with the protocol design is each time you introduce one new incentive, there’s sort of secondary behaviors and secondary incentives that need to exist to make that first thing work.

Brandon Ramirez (29:33):

So a great example is the primary incentive for serving queries is query fees between the indexer and the consumer. But then it’s like, okay, well how do you make that market safe to interact with? And it’s, well, you need verifiability, so then you need the arbitrator and the fishermen. It’s like, okay, well then what are the incentives for disputes? And then you know, there’s layers and layers. And I think part of what you try and do, because there’s so many different permutations of how these things could work, is you put the biggest stakes in the ground first, and then you see what falls out of that. And then you kind of go down that in order.

Brandon Ramirez (30:07):

And then at the black box level is where I think a lot more of the game theory comes in, when you’re getting to specific components in the system. We have very well-defined mathematical models of the bonding curves and how those work. And there’s this whole paper that the 0x research team did on the Cobb Douglas function that kind of goes through the equilibrium analysis there, again, for that component in the system.

Brandon Ramirez (30:31):

Where things get more complicated and more difficult to model is when you put all these components, all these different block boxes, staking, delegation, curation, query market, all the sub-components of that, the automated negotiation, the price mechanism there, put all of that into a complete system and trying to predict what’s going to happen. And then also just things that are outside of the model. So you can have these models of how a mechanism is supposed to work, but those models inherently bake in assumptions. And so there’s sort of things that emerge when you see these things all working in one system in an environment with real incentives.

Brandon Ramirez (31:11):

And so there’s a few threads that we have going on to address that complexity, and these are ones that are still ongoing as we speak. One of them is this investment in simulation. So the system that we’re building is complex to the degree that it defies useful, or it limits the usefulness of closed form representations of the system. And what I mean when I say closed form, is you could actually come up with a single equation, or a set of equations to say, this is how the system is going to behave. And a great example of this is, to give an appreciation for this type of complexity, is the three body problem if you’ve come across this. So if you have two objects in space, one orbiting the other, Newtonian physics or Newtonian mechanics gives you these really nice working equations. And I could give you an equation that says, okay, this is going to be the path of this object at time T=50, T=100. This one equation can tell you the entire future of this object’s movement.

Brandon Ramirez (32:20):

As soon as you introduce a third problem, it becomes intractable to solve analytically. At least you won’t have a closed form solution for it. You have to start solving it numerically using numerical methods. And as that increases, your only option really is simulation. And that’s true of any system of medium scale where you don’t have enough… So at a small scale in the example that I just mentioned, Newtonian mechanics, because the amount of interacting particles is so small, you can just perfectly model how those things will interact. At the opposite end of the spectrum, at the large scale, you kind of have statistical uniformity. So a great example here would be like Boltzmann’s equations that define how gas is going to fill a container, where you have many, literally millions, if not probably billions and trillions of atoms comprising this gas. But because they effectively interact randomly, you can sort of model them with these aggregate level laws.

Brandon Ramirez (33:20):

And where things become really difficult to model and understand is when you get to medium scale complexity, which is where a lot of the social sciences lies, where you have heterogeneous agents with heterogeneous incentives and skills and types and preferences all interacting, that becomes really, really difficult to model using equational reasoning. But a simulation can at least start to suss out some of the dynamics of that system. And so we have this long-running research driving block science where they’re looking to model protocol economics using cadCAD. Currently, they’ve been working on modeling the delegation subsystem of the protocol. So this work takes time.

Brandon Ramirez (34:02):

And then the idea is that once you have this fully robust simulation environment, you can start running agent-based simulations on top of that, where you basically construct a set of agents with some defined heuristic for interacting with the protocol. And you see, okay, what happens when we play forward the simulation with this set of decision making heuristics? And those could be anywhere from really simple naive heuristics, like just delegate to an indexer that had the highest yield at the previous time step, or it could be some heuristic that involves learning reinforcement learning or the solution to some optimization problem.

Brandon Ramirez (34:38):

So you can actually code some of the heterogeneity in the way that people make decisions, right? Because you and I know from our own experience that we’re not just these automaton rational optimizers. We’re not all going around handwriting equations and then solving some mathematical optimization problem. Oftentimes we’re making decisions quickly using some set of heuristics. And so that’s one stream of research that I think is really important there.

Brandon Ramirez (35:06):

Another one that sort of dovetails on this is what Semiotic, one of the grant recipients that the foundation is working on, which is primarily focused on reinforcement learning. We know with a similar eye towards sussing out what behaviors and dynamics emerge when you have reinforcement learning agents optimizing themselves for the rules of the protocol. So they’re applying some of the same techniques as like GPT 3 or AlphaGo. AlphaGo was this project by Google DeepMind that beat the world’s best Go players. Basically coming up with agents that are optimized for interacting with the economic incentives of the protocol. And then again, we can observe sort of experimentally, what are the behaviors that emerge when these agents optimize in this way? And their line of research currently is focused specifically on the negotiation protocol in the query market. So the place where we have price discovery, because there’s some interesting constraints that we have there with respect to if you were to compare as to the way that Ethereum gas prices work for smart contracts, we have our own unique set of constraints.

Brandon Ramirez (36:14):

And then in addition to that, we’re still leveraging the classic economic toolkit. So the foundation this past few months, had a grant to the Prism Group team. They’ve done some foundational research in the block science space. I think they had a paper a few years ago, it was blockchain forks as a coordination game. And they’ve been working on an equilibrium analysis of the protocol. And so this is trying to come up with an analytic representation of agents’ behaviors. But in order to do that, you need to make a lot of simplifying assumptions around agents’ behaviors, around the type of equilibrium that will form. And so it’s a sort of useful juxtaposition to the more simulation experimental set of approaches.

Brandon Ramirez (36:57):

And then finally, I would say that we had an incentivized test net before the network launched, and obviously now we have a running network with real live incentives and billions of dollars flowing through it. And so there’s a lot of just empirical observation that’s coming from just seeing what works. And honestly, this is where I think that the UX approach really compliments the traditional economic approach, is that in traditional economics, again, you’re making these idealized assumptions of agents and why they’re doing things. And for us, we can actually do user interviews. We can sit down with users and say, “Hey, how are you making this decision? What are the heuristics you’re applying?” Why did you make this decision? Why are you responding to this in this way? And it’s really fascinating when you do thatm because you realize how far away real people are from this sort of idealized economic agent that we get from traditional economics.

Nick (37:49):

So given all the work that went into thinking about and designing the incentives, what’s surprised you the most as you’ve seen this thing out in the wild and performing on its own?

Brandon Ramirez (37:59):

I think a surprise was just how in engaged everyone is in the protocols across the board. It’s really incredible to see. Anything from the indexer office hours, but also indexer working groups, where they’re working through the math of the Cobb Douglas equation together and working through the math of the optimal way to time their allocation management. And then a particularly surprising one for me was all these sort of telegram groups that emerged with Curatorss where they’re exchanging, they have basically this knowledge economy where they’re sharing, “Hey, what do you know about this subgraph? What do you know about this graph?”

Brandon Ramirez (38:39):

We did a user study actually with one of the PMs on the Edge & Node team, Adam, who has some side projects in the NFT space, where he deployed his subgraphs that he had been running on the hosted service to the decentralized network. And within minutes, a Curators from one of these Curators groups DMed him and said like, “Hey, what are your intentions with the subgraph? How many queries do you expect? When are you going to production?” And that’s something that we wouldn’t have had the chance to observe have had we not run through that process ourself with kind of a real test case. And so that’s been a surprise to the upside, was just seeing how engaged people are participating in the way that we expected them to be.

Nick (39:57):

Brandon, you’ve been so generous taking us through the early days, helping us understand the ideation behind The Graph, and so much of your personal story. And then the time and effort and thought that went into different incentives for all the stakeholders within The Graph. Where are we, in your mind, as you think about where you’ve been, where we are presently, where are we in the evolution of The Graph? And what are you most looking forward to next?

Brandon Ramirez (40:21):

So what are the things we would like to improve and iterate upon in the future? A big theme here is, well, even taking a step back, it’s helpful to think about where we want to get to. So right now, today, still most users access the decentralized network through a gateway. And that gateway solves a number of problems. One is it reduces the amount of complexity between Indexers and consumers, because there’s basically fewer point to point relationships between gateways and Indexers than there would be if we had every indexer connected to every consumer in this many to many fashion. But our end goal is to eventually have consumers just querying Indexers in the network directly. So from everything that a gateway does today would just happen on a consumer’s device.

Brandon Ramirez (41:12):

So there’s a few things that need to happen for us to get to that reality. We need more state channel scalability. So we have a few efforts there. We’ve been working with the Connects team for quite some time. We also have a cryptography working group in the community that spans several teams and grant recipients that’s focused on allowing state channels to scale more effectively. And I think we’re going to be talking more about this in future posts. So I’ll kind of pause on that thread for now.

Brandon Ramirez (41:41):

Another part is making the query market work a lot more efficiently. So we have this V1 of the Agora DSL, which is this way that Indexers express price between themselves and consumers. One thing that’s really unique to The Graph as opposed to something like Ethereum… Well, a couple things. One is that in Ethereum, when you pay for a transaction, it’s this one to many process. You’re imposing a cost on the entire network. And because of the way the EVM was designed, it’s relatively simple, is the costs can be pretty well modeled and are well understood across a number of different client implementations and different configurations.

Brandon Ramirez (42:24):

With The Graph, there’s a lot more heterogeneity in the configuration of an indexer. You could be running a different database, you could be using a different ETH client, you could be running your graph node in a horizontally scaled cluster where you have different read nodes and write nodes that are scaled separately. That’s how the Edge & Node’s hosted service works today. And so this idea of coming up with some global uniform price that is reflective of every indexer configuration is just sort of, it’s impossible.

Brandon Ramirez (42:55):

And so we addressed that initially by giving Indexers a way to express their own prices in the market interface, that’s this Agora DSL. Most Indexers haven’t been able to navigate setting that in a way that’s price efficient. And so a lot of semiotics research and our goals for the next iterations on the negotiation protocol and this price DSL are A, giving Indexers the tools to model for their own configuration, their own setup, what the cost of a query is going to be, and to express that over the wire to consumers in a way that’s going to lead to a market-efficient outcome. So that’s another thread that’s kind of just necessary to facilitate this efficient many to many interactions between consumers and Indexers.

Brandon Ramirez (43:40):

On the verifiability side, we want, the arbitrator kind of acts as a bottleneck on verifiability. It’s a set of actors that are set through governance that have to repeat and redo the work of proof of indexing or a query at the station to arbitrate the outcome of a dispute. There’s a long lag. It’s a lot of work. It’s a slow process. You don’t know in real time if a query response you’re getting is correct. It’s only weeks later when this thing is adjudicated. And so we have some lines of research in the progressive verifiability space. One is that we would like to make indexing verifiable. We’re already fairly well set up for that, given that from early days, we made the choice to put subgraph’s execution model inside of a Waza RunTime.

Brandon Ramirez (44:28):

And now there’s some very sophisticated proof systems to prove the correctness of Waza computation. So I think Truebit was the first player in the space using the refereed game approach. But you’re seeing other projects like Optimism and Arbitrum actually put those approaches into production in a meaningful way. And so I think there’s a really credible roadmap here on the verifiable indexing side. It’s a bit more complex in those systems,

Brandon Ramirez (44:58):

Because you’re not just verifying the computation that’s happening inside the Waza RunTime, but you also have to verify the data that’s witnessed in from outside the Waza RunTime, right? Because subgraphs can call out to the blockchain, they can make a smart contract call on Ethereum, or they can call out to the database. And so the engineering complexity is substantially higher, but the theory behind it is the same. So that’s kind of one line of research on verifiable indexing.

Brandon Ramirez (45:27):

On verifiable querying, actually, I mentioned this cryptography working group that’s taking place in the ecosystem. Currently, they’re focused on state channel scalability, but there are some long-running research threads within that group as well that are focused on verifiable querying. And I’m sure that we’ll have entire podcast episodes and blogs about this again in the future. But there’s some nice properties about our problem that are sort of unique, or differentiated from some of the other spaces where you’ve seen zero knowledge proof supplied in the crypto space so far.

Brandon Ramirez (46:00):

And one of those is just that a lot of queries can be modeled in this highly parallelizable way, where instead of thinking about it as query as a computation, one long-running computation on the entire database, you could think of it as many smaller computations on individual rows. For example, whether or not a entity in a table matches a filter predicate. That could be a single computation, that could be parallelized with other computations that are checking whether an entity should be included in a result set based on that same filter predicate in different rows of the databases. So that lends itself to a unique set of proof systems that are maybe of a different flavor and character than what most folks are familiar with from like maybe ZK-rollups or StarkWare or some of that other stuff. So that’s kind of another line of research, and in that area we can go pretty deep.

Brandon Ramirez (46:54):

And then some other big high level threads are sort of just fulfilling the original vision of The Graph as this global interconnected graph of public verifiable data. So we started, if you look at where we started and where we’re at today, The Graph is primarily used on Ethereum, EBM-based chains. We support tons of different L2s or other L1s even that are EVM and Ethereum based. We would like to expand that. We believe in a multi blockchain world. And so I think you’re going to see a lot more support for other chains in the future.

Brandon Ramirez (47:30):

And then thinking about where a subgraph is today, even though we’ve broken developers out of these data monopolies and centralized data platforms, subgraphs themselves are still relatively siloed with respect to one another, right? So let’s say you have the synthetic subgraph and the chain link subgraph both deployed to the decentralized network, and it just so happens that the Synthetic subgraph wants to use chain link price feeds in its logic, because that’s important to the operation of the Synthetix protocol. Well, today, Synthetix’s subgraph developer couldn’t just go and say, “Hey, pull in this data from the chain link subgraph.” They have to recreate the work of accessing that lower level data from these other sets of smart contracts in the chain. So we don’t have true reusability and interoperability at the subgraph layer. And I think you’re going to see a lot more efforts towards that in in the future when it comes to things like composing subgraphs at query time, or having data pipelines where one subgraph connect is the input to another subgraph. I think that’s going to bring us much closer to the original vision.

Brandon Ramirez (48:39):

And then those stages in the pipeline, those subgraphs, they could also be smaller and more purpose built. So you might have a sub-graph that just focuses on some derived analytics view on top of the data that’s already been produced or indexed by some upstream subgraph. I actually think this is going to enable some of the first examples we see in the wild of bring your own algorithm. This is one of the many promises of web3 to address the problems of things like Twitter, Facebook, where we’re all sort of forced into this one recommendation algorithm to rule them all.

Brandon Ramirez (49:17):

Imagine with subgraphs, you have one giant dataset represented as a single subgraph, and then you could have many smaller subgraphs that comply with some interface of the recommendation or the sorting. And each of those reflects a different point of view on how data should be sorted. And then you could have end users selecting on those recommendation algorithms client side, right in the app. So this composability, this interoperability, there’s many other benefits to it, but that’s just one example of how this is going to enrich what’s possible in the web3 app space.

Brandon Ramirez (49:50):

And then just across the board performance, I think you’re already seeing some large investments in performance with some of the grant announcements with teams like StreamingFast and Figment that came from the foundation. But performance just makes everything better. It’s not just about the performance of the developer, it’s like all the protocol economics work better when subgraph indexing performance is better. Indexers need to do less work upfront to see whether or not query demand is going to materialize for some subgraph, reduces their costs so they can pass those savings on to consumers.

Brandon Ramirez (50:25):

So performance at the subgraph level, and then also at the protocol economics layer. A lot of the heavy costs right now to sub graft developers just come from L1 gas costs. So there’s a lot of work and talk within The Graph ecosystem right now around what is the optimal L2 scaling strategy look like? What is the right deployment target, what’s the right architecture? So what contracts do you leave on the L1, for example, for DeFi interoperability versus what stuff do you put on the L2 for maximum performance gains?

Brandon Ramirez (51:00):

And so those are a lot of the enabling research threads to get to where we want to end up, which the end vision is that The Graph is a global decentralized market of Indexers that operates efficiently, that consumers can reliably query that data from their device, and that this is a stable platform that can be relied upon for every future layer in stack that accretes on top of it. And I’m just as excited about what this is going to unlock for users and developers as I am for just what this unlocks as a case study and how these things could be built in the future. And I hope this is the way that future protocols and products are built as well. And I hope that we can provide a useful example on some of these fronts.

Nick (51:43):

Brandon, thank you again for your time. You’ve been so gracious. And I know for me personally, and I think I can speak on behalf of all the listeners, this is incredible information and access to you and the way you’re thinking through this. So I’m grateful, and on behalf of all my listeners, I express the same appreciation. If listeners want to stay in touch with you, learn more about the work and everything that you’re working on, what’s the best way to stay in touch?

Brandon Ramirez (52:06):

Yeah. First off, thank you so much for having me. This was an absolute pleasure. If folks want to stay in touch with me, they can find me on Twitter. My Twitter handle is @rezbrandon, that’s R-E-Z-B-R-A-N-D-O-N. And if you guys are interested in collaborating, please check out The Graph Foundation’s many grant opportunities or go to edgeandnode.com/jobs. We have a number of open job postings, so if you are just someone talented wants to get more involved in this space, please check them out.