Supra’s Vision For Automatic DeFi aka AutoFi Featuring Joshua Tobkin

The key fundamental thing is that we move away from just classical oracles doing just data price feeds, you know, to systems that can actually debate and reason. And then in order to keep it safe, you need to randomize those— the inclusion of these different models. They have to debate with each other and cryptographically sign off on their their output, which allows us to bring in qualitative data, not just quantitative data. Qualitative, like what is sentiment? All kinds of things that you can't get otherwise into a blockchain.

Nothing said on Empire is a recommendation to buy or sell any investments or products. This podcast is for informational purposes only, and the views expressed by anyone on the show are solely their opinions, not financial advice, or necessarily the the views of Blockworks. Our hosts, guests, and the Blockworks team may hold positions in the companies, funds, or projects discussed. All right everyone, welcome back to Empire. Very excited about this episode. We have Josh Tobin, co-founder and CEO of Supra. Uh, Supra has been a longtime supporter of Blockworks across permissionless and podcasts, so really excited, uh, that they made this episode possible. And yeah, Josh, just welcome to Empire for the first time, man.

Yeah, no, it's great. I've been listening to your podcast for years. I learned so much from it and really love your takes. So it's such a privilege to be here.

I have horrendous takes, but we bring on people with good takes, so I'm glad to have you on. So the reason I want to bring you on is you guys just released this white paper, Super 2.0, Automatic DeFi and the Future of Decentralized Finance. You rolled out this term that I had never heard of called AutoFi. And the way that my, you know, smooth brain thought about this was, look, right now in crypto, there's a lot of actions that users have to take. But what if you could run DeFi essentially on autopilot, managing risks, moving money in real time, generating real revenues? And obviously for that, you need the infrastructure. And I think that's what you guys are trying to build. So I want to get into that maybe in the second half of this. But for the first half, can you just give us the setup for this conversation? Like, and maybe the first place to start is just what is Supra, and what are some of the design decisions that you guys have made to get here?

Yeah, yeah. So Supra, our point of view and our kind of general thesis is that if a layer 1 can handle more services than just one thing, then they should. If there's enough compute and bandwidth available, uh, to— for these nodes to be doing more than one service simultaneously, they should. So we— our thesis is basically the opposite of the modular thesis. Ours is the vertically integrated thesis, full vertical integration, single-layer one that also has built-in Oracle protocol that's very high-performing, very secure, cross-chain communication, and on-chain randomness, as well as automation. The automation is a big one because it can enable many things. But really, the magic happens when you combine smart contract platform, high-performing consensus with parallel execution for smart contract execution. With access to Oracle data. So you have all these different triggers in which you can do things and to trigger automations. Having them all together on single, single stack is really where the magic happens.

Okay. How did we get to this as the philosophy?

Right?

Because on one hand, maybe you have like these general purpose chains. On the other hand, you have like AppChains, obviously vertically integrated. Maybe you could also say like purpose-built. You guys are maybe somewhere in the middle. I don't know if that's the right if it dumbs it down too much to use that kind of spectrum. But how did you arrive here?

You know, the first origin of the idea is back in 2017, like November 27th, San Francisco. You know, it just clicked to me that we could organize the network into a very unique architecture we call tribes and clans. And but in order to keep the shared security, we needed randomness. To randomize the allocation of nodes to different committees. Right. And to get randomness onto a deterministic, perfectly ordered blockchain was a difficult problem. And in my naive kind of point of view, like, 8 years ago, I was like, okay, we need oracles for this. I was actually originally, like, back in the day, very supportive of Chainlink and Link Marine because I was like, this makes so much sense that smart contracts can do so much more with oracles. And also basically just the way that a very naive approach was like, okay, with oracles we can take stock prices, crypto prices, weather data, combine them together to create randomness. That was the actual origin of this. Of course, what we do today is much more quick, lower latency, faster, cryptographically secure, unbiased, unpredictable for the randomness component. But that opened up the door of like, yeah, a blockchain should have built-in oracles. And by then, or back then rather, this was not a common idea. I mean, people were still trying to wrap their head around oracles. And that naturally led towards, well, what else is an oracle problem? Well, it turns out cross-chain communication is part of the oracle problem. Automation is part of the oracle problem. So, yeah, as well as, of course, getting data on chain from the outside world. So that's the origin, and that's how we kind of got inspired to pursue this. And then ultimately, just fundamentally, I will say that the advent of Solana really kind of shaped my worldview. It was very much shaped about— actually, I went into blockchain trying to scale Ethereum originally. I was a nobody, but That was my intention, right? And then Solana comes several years later after we created our own crypto lab. And Anatoly really shaped my point of view in terms of like, well, these nodes can be doing a lot more. And then that kind of just matured the line of thinking that we already had on the oracle side of things.

So right now, if an oracle data in Supra is able to land in every block, as I understand it— and again, I'm going to ask some really dumb questions because I'm not too deep into the weeds of Supra, but As I understand it, Oracle data actually can land in every block. Others, I think, don't do this because there's this, right, it can kind of congest the blockchain, which can hurt DeFi. So what you guys have done is make this design decision to actually land the Oracle data inside of the block. Can you just, A, tell me if I'm even right on this point, and B, expand on why this is so important?

Yeah, so our blockchain is super fast. When things are on the fast path, we're producing 3 blocks per second. So we're talking about 300 milliseconds. Seconds full finality, by the way. So very fast. However, our Oracle protocol right now in the current version is 600 to 900 milliseconds to finalize that data. And we're not talking about producing the data. I'm talking about the data is ready to be used, has been verified. So we can't necessarily update every single block because our block is so fast, our blockchain so fast. However, approximately every 2 blocks or every 3 blocks, if there's an update, there's an update. Now, what we do since we're designing the system this way is we realize, especially by Looking at the history of DeFi, there's been major problems with blockchains not being able to update Oracle data when there's a lot of congestion, you know, because they have to try to enter the blockchain just like every other transaction. What we are doing is prioritizing, giving priority access and block space to Oracle data that's generated from our system. So whenever it's ready to deploy or when it's finalized, we, we, we let it be included. It's, it's outside of the scope of block congestion. We prioritize it so that in case of high congestion. That's usually, by the way, where a lot of liquidations are occurring and such. We want to make sure the Oracle data lands, so we prioritize it. Yeah. So the purpose behind it is primarily to make sure that Oracle data lands so that DeFi can continue to stay solvent, not creating this like, oh, the Oracle didn't update for 20 minutes and then caused cascading failures for a period of time.

Happened the other day, actually.

Yeah, I heard about that. Yeah.

Yeah. What else do you feel strongly opinionated about in terms of design decisions for building base layer infrastructure in blockchains?

Yeah. So I feel very strongly that the modular approach is a problem. I think that it introduces security vulnerabilities, it introduces additional costs and complexity for developers. It's also slower. And okay, let me explain why. Because now you have a layer 1 or layer 2 that has to communicate with a data availability layer. That data availability layer has to finalize its proof that the data is available before you can move on to the next step. Same thing, modular approach, external oracle, external network call. They have to finish the protocol, send it back to you. Same thing with on-chain randomness, cross-chain communication, AK bridge, as well as automation. These are all typically external networks, so you don't have the unified shared security. You rely on other networks for their security. And if your application taps into any of those services, your application security has now been reduced to the weakest component, right? But beyond security is, is cost, and a big one is latency. So, um, we think that, uh, you know, the whole world, especially last couple years, we're moving towards this modular world that's like a thousand flowers bloom, and I love that too, you know. But Supra's philosophy is full vertical integration. Now, there is a caveat. While we have been described as like a mega monolith, we do have modular components, meaning the consensus algorithm can update without affecting the Oracle protocol. The automation V1 to V2 can update without affecting Oracle or Layer 1 execution, et cetera. So it's a monolith, but it's built in a modular fashion. Actually, any global scale technology has to be built this way. So why not blockchains? Yeah.

Okay. So as I'm understanding it, so you've got this mega monolith, or maybe we call it like a monolithic unified system where you've got these maybe modular components inside of this monolithic system. Okay. So, and okay, I'm just wrapping my head around this. And then you've got this vertical— you also describe it as like this vertical stack. Maybe my next follow-up question to that would then be, which part of the stack do you want to build and own? And which do you want to outsource? Right? Because one, one of the early strategies of like, you know, L1 builders was, hey, we'll just build the very bottom of the stack and then we want all these great developers and founders to come build. And now I feel like it's starting to actually change where if you talk to a lot of the best L1 founders, the new folks, I feel like they say, actually, we want to build more and more of this in-house. I'm curious how opinionated you are on that note.

Yeah, actually, we're very opinionated. So first of all, Supra is led by Dr. Anata Kate. Dr. Kate is the inventor of KZG commitments, which underpins EigenDA, other ZKP protocols and beyond, as well as even Ethereum's kind of data availability roadmap relies on this cryptographic primitive that our chief research officer is the inventor of. He's the K in KZG. We have him and we have other academics and former professors as PhDs. They're on staff. Now, when you have that, right, full-time, that type of caliber on staff, you got to understand that even myself as primarily focused on product, right, I can't say let's take shortcuts. It doesn't work. So everything we do has to go through a very rigorous process. In fact, we publish to multiple academic venues. We've been accepted to top— some of the top academic venues in computer science. In fact, the top academic in computer science. Recently we got 3 of our white papers into CCS, and this is basically like the Harvard of academic computer science papers. So we, you know, as much as sometimes it's kind of a little bit frustrating for me, it's like, let's go, let's go, let's go. We are disciplined to take a rigorous process, peer review, et cetera. So why am I saying this? I'm saying this because The contrarian view is that we actually think we can develop protocols for each of these components that are actually outperforming the other projects. And it's because we assembled like the Avengers. We have an incredible team. Now, that said, of course, if someone comes up with something better, since Supra is modular, we can always swap something out or make an adjustment. So that's always on the table. Uh, so the, the contrarian kind of— and I hope I'm not being seen as arrogant here, but I truly mean this— that a single expert team can outperform 1,000 mediocre teams, right? And yes, occasionally someone does something better and we can adopt it if it's allowed to, uh, or we can draw inspiration from it. So, um, more or less the kind of point of view at Supra is that we study a lot. Anything that comes out that's making an impact, we want to understand it and we want to analyze it, and we always like to pay homage where credit's due. So that's something very important to us, and this is why we always cite. We have a rigorous citation process in all our papers, because if someone came out with an idea before us and it's a good idea and we want to build upon it, they should get the credit. Yeah, yeah, it's very important.

So yeah, there was a You're making me think about it. So I tweeted this the other day, but I'll repeat it here. I was listening to the Acquired podcast and there's this Steve Ballmer quote, um, and I'll just read it here. He said, platforms need apps. You want to own the top first-party app that runs on your platform. Otherwise your platform can't get good. Office was the best first-party app on Windows. Outlook was the best first-party app on Exchange. You want to make sure that you actually own the first-party app in addition to just your platform. You'll get stuck in the mud if you're just a platform. And so my take on that is, look, if you apply that Steve Ballmer take, who's obviously a phenomenal operator, if you apply that to crypto, L1 should not only build the chain, they should also build a few amazing apps as well, which is very different than the current strategy today, which is basically outsourcing 100% of your app building. How do you think about that as applied to Supra?

Yeah, no, we actually are totally aligned with that. We had always originally planned to, uh, enshrine, like, build into the system, uh, something called a dynamic function market maker. It's a new type of AMM, and this is pretty much meant to be socialized and be a social good in, in our pursuit of value capture for the network. So that is something that is, uh, on the way. And, uh, more recently, I guess I can just go ahead and say it, Supra is going to be incubating a couple of protocols including next generation Perp DEXs that leverage our stack. For example, Hyperliquid really did an incredible job with their product, right? And also their kind of dynamics in terms of community, the revenue share, et cetera. I think they've done a really great job. Now, that said, of course, that's not going to scale beyond their whatever. I think they launched with 4 nodes. Maybe they're at 14 to 21 nodes today. They're not going to scale to hundreds of nodes, right? So there's always going to be that tension there. Nonetheless, the vertical integration, that is a another example of vertical integration. There's a lot of similarities between their philosophy and ours, meaning they actually have their nodes running Oracle for their price feeds. I think their automation is centralized, but they also have built their DEX and they custom built this for what they're doing. And I think somewhere in the middle is the answer, meaning you want to have decentralization and at the same time the project that knows the infrastructure the best, why not launch products as well? And in our case, and especially in blockchain, I'm really trying to navigate this towards we launch products if they're successful, that capture the value for the ecosystem and other stakeholders instead of just the labs. I really want to enshrine these, frankly, financial primitives into the system. Anything that's a primitive, aka it's always going to be on every chain, might as well be owned by everyone. That's kind of my point of view.

So you want to enshrine— you almost want to enshrine not just the, like, an ERC-720, you know, like, you're— I think you're talking about enshrining applications in a sense.

Yeah, yeah, yeah. I mean, yeah, no, absolutely. And so there's all— there's certain things that all chains need. They need the ability to exchange value. They need ability to do that efficiently. Now I know that this is a little bit of a touchy kind of topic. This— I want to make clear, this does not preclude other protocols from launching on us. In fact, they can tap into the product suite.

If you guys build the best DEX, can anyone go build a DEX?

Yes, they can. Anyone can do anything.

They really get traction. Like, would you support them and stuff like that?

Yeah, actually, more activity is better for the ecosystem and we would not shut anyone down at all. In fact, we would just say, hey, if your product is better, let that win, actually. The thing is, if we developed anything with foundation support, et cetera, then it really should be owned by the network. Right. So there's some nuance there because some things can be, quote unquote, socialized. I am an American and I want to be the bottom line there. But somewhere between democracy and communism, there is probably a sweet spot there. I'm not saying that's Europe either, but in blockchains, the value really ought to accumulate to the stakeholders. And there's more than just— the stakeholders include node operators, they include dApps, they include token holders, community, et cetera. So we are designing a system that is meant to make sure that the value is captured, by the way, not by a small set of sophisticated MEV connoisseurs, which are very smart. It's like the value is getting sucked into 1,000 people out of tens of millions, whereas what we're trying to do is make sure that value is is kept into the ecosystem, keeps the ecosystem alive. And this is a kicker. If we do this successfully, we hope we can wean off of block rewards within 2 to 3 years, so there's less inflation, and that's good for everyone.

Yeah. What are your thoughts on— I mean, I know you strongly disagree with the kind of modular thesis. What are your thoughts on other L1s, like, um, who seem to be doing quite well right now, like the Solanas or Base or Hyper— like, I mean, you could say Hyperliquids. You could say they're an L1 because they say that they're an L1, right? But what are your thoughts on those guys?

Yeah, no, let me start with Solana. So Supra, actually, we've paid a lot of attention to Solana, but we always knew that proof of history was like an inefficiency in consensus. It will never perform better than a classical consensus algorithm. More recently with their Alpenglow, they finally got the academics involved here that shaped this up, and it's very good direction. In fact, Solana is moving closer to what Supra is doing. They're using BLS signatures. They're doing all-to-all communication. They're moving into a classical BFT algorithm. Right. That was a bit of a scare. I was like, crap. I always knew that Solana's Achilles heel was proof of history. So I was like, we always knew that we're going to outperform them. Right. But now they're making these changes and By the way, they are making changes that are in line with our existing research. So what I'm saying is, like, actually just a couple of days ago, we released a paper called Hydrangea. Hydrangea is a new type of consensus algorithm that is in the same kind of network assumptions that Solana is moving towards. We also pointed out some things where we have more resilience in them, etc. But Solana is the pursuit of high performance is something that I think we all have to pursue because that's what users want and demand and expect. Now, when it comes to Hyperliquid, they're also an instantiation of vertical integration, right? Layer 1 with Oracle, with bridge, with DApp, right? So they basically cement our thesis. We were kind of alone for a long time there in the modular world. Everyone was in the modular world. It's like, this is going to be incredible. We were like, well, there are some major drawbacks, which is latency, security, costs, right, which do affect the end consumer more or less. I would say that it actually has been a little bit of a lonely path in that we were standing out as saying, nope, we think this is the wrong direction. We think vertical integration is very natural for technology. It always happens in every field. And Yeah, we think that as long as computers and nodes have enough bandwidth and compute, they should be doing more than one thing. And yeah, that's kind of our point of view. Do we draw inspiration from everyone? The answer is yes, we do. We are kind of like very much a curious research firm and an organization, and we love good ideas. Let's put it that way.

Yeah. Speaking of good ideas, can we talk, can we talk Autophy?

Yeah, for sure.

I would love to just hear your vision of what the world looks like in like 5 to 10 years as it relates and where blockchains fit into that. And then we can get into this white paper.

Sure. 5 to 10 years. Of course, these are very hard to predict, right? Nonetheless, I think it's very natural for folks to— the things that don't change, like the whole Jeff Bezos kind of quote about you build a business in which you know that these attributes are always going to remain, aka people will always want cheaper prices at faster delivery, right? That applies directly to blockchain. People want better performance with lower latency and at a cheaper price. So that's always going to be there. Now, the things that are missing from blockchain today include, as you mentioned at the beginning of this podcast, automation, meaning right now people are manually having to try and set this transaction, etc. Or you do use an automation service, but sometimes it doesn't trigger. And it's not deterministic. It's not— it also has latency because they have to decide to do it or not and then submit it to the network. There's no on-chain automation, really. There's definitely not a layer 1 with oracle, with automation. That's actually where the magic happens. You need all three. So the other component is also in relation is 5, 10 years from now. I mean, AI is going to get— AI ain't going nowhere, let's put it that way. You know, that, that is going to be It's a major platform. It's a new computation paradigm, so to speak. I think everything gets reinvented with it to a large degree, and we need to figure out where does blockchain and AI intersect. The problem with AI is that it's not deterministic. It can hallucinate, right? And that's the opposite of blockchain. Blockchain is perfectly ordered, transparent, and clear. But how can they work together? And so we think in the next 5 to 10 years, you're going to have high throughput, you know, parallelized execution with access to AI. So you do things on chain in a very safe manner. And also we think that automation's going to play a big role here. Part of AI, major thing about AI that's going to help us is automation of workflows. But these workflows exist in layer ones as well, whether that's a sophisticated strategy, you know, it could be a trading strategy or hedging strategy, whatever it might be. There needs to be automation. So AI and automation go hand in hand. AI, blockchain, and automation go hand in hand. How do we do this though? And our conjecture is that you need to vertically integrate to do this. And, but remember, vertical integration does not mean centralization. That's the key tension here. We're designing— yeah, we're designing the systems to be decentralized.

Okay, let me read— let's get into the Autify a little deeper because instead of using this word verticalization, I want to get like maybe into a little more of the specifics. And there's this one line from the white paper that I'm going to read Okay. By embedding— ba ba ba— skip, skip, skip. Autophi transforms BlockFi. Okay. Autophi is automatic DeFi for people who missed that at the beginning. Autophi transforms blockchains from passive settlement ledgers into proactive, real-time economic systems capable of autonomously coordinating liquidity provisioning, adaptive vault strategies, arbitrage, and liquidation. Tell me what this means, Josh, and importantly, how are you guys doing this when other— like, I think this is a commonly held idea that AI is going to change things and the world will get more automated. A, what does this actually mean? And B, how are you guys able to do this?

Yeah, so today, blockchains today are these passive ledgers. Things happen, it's recorded. After it's recorded, someone triggers something for something else to happen, usually. When you have automation built into it. So the way our automation service works is every automated task is registered across our layer 1 itself. Every node has this information, so they don't have to communicate and run another consensus algorithm to decide, yep, we should do this. They all know if these triggers in this block are correct and these conditions are met, we automatically in the same block execute this, right? And we run automation every block, which means up to 3 times per second. So it's very fast, virtually instantaneous. Now, that alone will help you automate on-chain transactions based off of things like Oracle data or state changes or transactions that are trades. You can say, hey, look, if this happens and these other conditions are met, do this without human intervention while you sleep. So it's like autopilot. Now, this idea of autonomous this is a little bit futuristic, but we're so close to this that it's worth talking about, is that not only can it run on its own with its own logic, the automation logic, but it— that logic could be tweaked with the introduction of a secure AI oracle. You know, you need— when I say oracle, I'm saying this is a something outside of the layer 1 itself that is bringing information into the layer 1. So we do have something called Threshold AI Oracle which basically what they do is there are committees of AI models that debate each other in natural language. It's actually in English. And they go back and forth. They have to agree. And if they agree, they cryptographically sign it. You have to hit a threshold. What this allows us to do is get decision-making done in real time by these autonomous agents, but not just a single actor. It has to be a committee of them who one might be programmed to be pessimistic. Very pessimistic, like, no, no, no. Another one might be you look for opportunities and then they still have to kind of reach agreement before it's brought on chain. And this can power adjustments to automation. So it can kind of be this reflexive system based off real world events.

So you've got these agents. So the threshold AI or threshold AI oracles, TAOs, these are really agents that you've pre-programmed to essentially vote on what should happen. Like, I use vote in quotes, but like direct the protocol and things to happen. What are— so how many agents do you have? Is it like 7 and you need 4 out of 7, or is it like 700? Or yeah, what does that look like?

So it's so early, this idea. We went from idea to paper to prototype in less than 3 months. And so right now in our prototype, which we shared online, It's only like 4 or 5 agents, but this is supposed to grow and we're going to invite— it's going to be permissionless, meaning anyone can stake, anyone can participate. There's going to be models that are fine-tuned for various use cases. Right. And there's a lot more to do here. But the key fundamental thing is that we move away from just classical oracles doing just data, you know, price feeds, you know, to systems that can actually debate and reason. And then in order to keep it safe, you need to randomize the inclusion of these different models. They have to debate with each other and cryptographically sign off on their output, which allows us to bring in qualitative data, not just quantitative data. Qualitative, like what is sentiment? All kinds of things that you can't get otherwise into a blockchain. Yeah.

I'm trying to wrap my head around what this actually like a use case of this. And I think I've kind of got something in my head. But so if you combine AI and these oracles and a blockchain and cross-chain communication, what is the thing that that allows you to do? It's like, I think you mentioned on the, on the Bell Curve episode with Mike, it's like when the price of Bitcoin is X and the price of ETH is Y and CPI comes in lower than Z, then buy this amount of, you know, ABC token or something like that, right? It allows you to do these more complex automations. It's almost like more complex if-then statements. Is that— am I thinking about that in the right way?

Absolutely. That's exactly right. You can even throw into the mix, if the sentiment of, you know, X token is X, Y, and Z as well, do this. So you can add these conditions, right? And how do you calculate sentiment? Well, there's not really an oracle price feed for that, per se. You may have to analyze the news, the world events, you know, and, you know, so basically just expands the aperture of things that can move on chain to trigger automations. Basically what this means is individuals can have their custom strategy and their own worldview as to what moves the markets, and they can do what they want with that. It's basically just giving you more tools to decide how you want your assets to be managed.

Have you thought about— Mike and I have been just thinking nonstop about prediction markets and what they will Our take is kind of like Polymarket and Kalshi might be V1 and V2 is going to be— might look a little more intricate. What are your thoughts on prediction like this? Maybe this is because in my head I've been thinking about prediction markets nonstop, but how do you think about deploying this new AutoFi tech into prediction markets and if there's any sort of crossover or I don't know if you've— maybe you haven't thought about that much, but—

Oh, absolutely. And by the way, if you guys want to work on this, let me know. But basically, basically the Threshold AI Oracle is can resolve the prediction market without human intervention. So Polymarkets relies on UMA Oracle, U-M-A. And there was a recent attack where there was a governance attack. Someone had acquired enough UMA tokens and voted incorrectly to arbitrage their position. Right. And I think they made money off of that. So that's like human in the loop. Right. But threshold oracle, if it's robust enough, can just resolve the answer to all these things at the end of the prediction market deadline. The oracle can just run its protocol, cryptographically sign it, reach the threshold of the system, and then just, it's done. No humans required. So this is something that could speed up the settlement process, but also remove human bias, at least. Of course, AI is going to have its own challenges. Nonetheless, this is why we introduce committees that debate each other.

Yeah, sorry to interrupt that. I was just thinking there's an interview that Onchain Times did with Domer, who's the number one trader on Polymarket. and there's this quote he said, which I didn't know this existed. He said there's a cottage industry of users on Polymarket who actually specialize in trying to convince UMA in the dispute process. And there's all this like schmoozing and back— I forget the actual details because the article is like over a year old, but all the schmoozing and backroom deals involved. And it, uh, he's like, he's like, look, it borders on like fraud. Um, but he's like, you have these token holders with a vested stake in, uh, he was saying if Polymarket had a token, it would actually make a lot more sense because you'd have token holders with a vested stake in Polymarket's success, where they would be the people deciding the resolution in unclear markets. And as it stands right now, UMA votes on the disputes, but their interests are a little misaligned with Polymarket. So anyways, not to go down rabbit hole.

No, no, I, I, I can see that. It's a little bit not surprising to hear that there's that smooching and kind of, hey, what if you change your vote? And, you know, um, yeah, but, uh, and, and I want to say that threshold AI oracle, especially in unclear situations, may need eventually human intervention. Uh, but for vast majority of use cases, I think probably 95% of the use cases, uh, maybe 99%, you're, you're smooth sailing, fast response. This is definitive. This is it. Yeah, it's only when, like, they cannot come to an agreement that humans might have to be involved there. But yeah, back to kind of what you're saying is like for Autify, pretty much like automating your strategy is really what it comes down to. I'm trying to do like, for example, Bitcoin doesn't have smart contracts. We'll bring wrapped Bitcoin to Supra and tap into these access to smart contracts, automation, Oracle to do more interesting things and to trade your asset. By the way, some of these strategies with Threshold AI Oracle, you'll be able to do on Supra that you can't even do on a traditional DEX. Sorry, CEX. You can't even do this in Trezor, you know, which is really interesting because we think that this is going to be a compelling reason why folks will over time— we're actually rolling out our Supernova crossing communication bridge. We think in like maybe less than 2 weeks. This is layer 1 to layer 1 communication. We think that when that's implemented, folks over time will start to have a strong reason to bring their assets to Supra to execute these things you just can't do elsewhere.

Interesting. Like, can you say more examples of what you can't do elsewhere?

Yeah, yeah, yeah. So big ones are sentiment. Oh, of course you could do this in a centralized manner if you're a sophisticated developer, can set up the data pipelines and all those things related to that, right? But for the average retail user, these are just— it's a no-go. But imagine a strategy builder that's AI-assisted. You're just talking to it in natural language, even just voice, where you're saying, hey, this is what I think, this is what I'm trying to do. Can you help me set it up? And then it can help you set up your strategy, but you don't want to quite deploy just yet. You're going to want to backtest it on historical data. We have Oracle data, you know, sub-second kind of finality, many ticks that you can start to test it against as well before you deploy to make sure your strategy does perform or outperform the market. Other, other things you can bring on-chain. Yeah, so sentiment's a big one. World events may include like start or ending of wars, interest rate cuts. Another one, technical indicators. Many things that you might otherwise get, as long as we can bring it into the blockchain safely, which is an oracle problem, then you can mix and match all these variables. And the combinatorics are very large, so it's kind of unpredictable how people use this.

At the center of a lot of this is the Super Treasury, which, as I understand it, is this basically auto vault. It's this vault that's like a programmable vault that sits at the center and dynamically allocates capital. To different Autofi strategies. Can you just expand on what the purpose of the Super Treasury is?

Yeah. So this decentralized treasury is actually meant to be network owned. It's actually meant to be a public good, right? And we should definitely talk about how it plans to capture value. But the idea is that we have these strategies that are built into the system through what we call auto primitives such as auto liquidations and auto arbitrage, meaning community goes first. But any opportunity that's left, the system-level automation can kick in and we can rebalance DeFi. We can make sure that any liquidation that was not triggered, we trigger it. Keeping DeFi solvent, by the way, this is good for everyone. But these also tend to come with significant revenue, right? So of course, community can go first, but they're also paying for their gas, for execution fees, etc., and automation fees. But any opportunity that's missed, this treasury at the system level will capture those opportunities, become a revenue source for the treasury. Now, the treasury, moreover, will take that revenue. And first of all, that revenue is designed to be split across dApps where the opportunity came from, become a new revenue stream for the dApp. They might not be capturing— they're not necessarily capturing this value today. It's being leaked out to these very sophisticated MEV searchers, extracted from the network. Right. Whereas if you are a dApp and you create this opportunity, whether it's liquidation or arbitrage, you really deserve part of that revenue cut. That's our point of view. We capture it, we redistribute it in real time to where it came from, but also— and we can talk about this a bit— but we are trying to remove toxic MEV as a revenue stream and incentive for node operators. This is very sophisticated in how we approach this, but we replace it with this revenue capture, revenue share in real time. So node operators have, you know, yeah, we take— we try to mitigate toxic MEV, but we replace it with a new revenue stream. The treasury, let's just say that maybe it's 50% to the treasury, which represents the network and community. What we do here is the idea here is that we build, accumulate the treasury. This is eventually going to be completely governance-based where we could do buybacks, we could do reinvestment into other initiatives. It could be a protocol development, it could be investing into promising dApps. But also where it gets really interesting is with automation, the treasury can slowly— with approved dApps that are deployed in our community, the treasury could actually deploy as a liquidity provider, starting with a little bit, deploying a little bit, monitoring it. How's the activity? What's the ROI? Increasing the amount that's deployed slowly over time. Right. To provide liquidity at, you know, you know, not by not, you know, and by the way, these, these types of activities are still fee generating. The motivation is still simultaneously value revenue generation for the treasury, but also redistribution. This can help dApps that are, you know, have been qualified, meaning maybe they've passed the audits and, you know, maybe there's a docs team and, you know, other things. If they pass this, this kind of rating system or community-driven kind of review system, then the treasury could automatically start to deploy some capital, providing liquidity to DEXs or lending protocols, and then scale that as performance improves. And doing this not in a mercenary manner, it's actually completely missionary. We don't have to charge as much either. And the idea is that the whole point is to— it's really important for sustainability that we do the treasury for the network's sake, for the stakeholders' sakes, collects this value. But it is meant to be a public look good and redistribute it. So that's the intention behind it.

Do you think there's an idea that actually the best business for a lot of L1s and L2s to be running is more of a fund type model with their treasuries? Because there was this moment in time where you were able to get a lot of money into a treasury, and I've seen other founders of L1s and L2s actually think about this as well, which I mean, you guys could take this to the far end of the spectrum. You say instead of the treasury being a public good and the business being the infrastructure, you could flip it where the infrastructure is the public good and the treasury becomes the business model. Have you thought about that at all?

A little bit, I would say. I mean, the thing is, the treasury would still have to be for the benefit of the stakeholders of the network instead of just team investors, right? So it has to be for everyone. But certainly it makes sense to allocate in diverse manners. So if it makes sense, you know, the Treasury should allocate to other VCs, frankly, as well. But we also can allocate to— at least that's my opinion currently. But at least, you know, the system is designed to capture some of the revenue that's lost to MEV otherwise. And then to redistribute it to dApps and ecosystems. These things can be governance-based. So lending protocols to DEXs for liquidity. By the way, these are still fee-generating activities, right? Investing into other dApps ecosystems. It could be we want to invest in education and have like a hacker house in the physical world, whatever it might be. These are things that are possible. But I do think there's strong balance between— ultimately, the treasury becomes a completely public good for the public good, and that's its mission, right? But should it allocate to other fund managers, almost like a fund of funds? The answer is yes, in my opinion. I think they absolutely should. And we intend to do this, by the way. Yeah. But step one is we do have tokens allocated for treasury. But step two is to actually implement this flywheel we have automation going to mainnet a couple of weeks from now, implement the auto arbitrage, auto liquidation, generate the revenue capture redistribution mechanism, which we already started working on. As we build the coffers, so to speak, those funds should not sit there idle. We want to put to work, help the ecosystems. Of course, in the short term, that could just mean purely just capture the value, lock it up as basically buybacks. But also instead of burning the token, right, we are thinking to introduce something called freezing, where those tokens are frozen for, let's say, 3 years, removed from circulating supply for the 3-year period, having the same effect as burning. But 3 years later, it becomes accessible to the treasury and the network that can be deployed according to what is required at that time. But a big one is we do want to get to the stage where we generate generate as a system enough revenue through auto arbitrage, auto liquidation, automation fees, layer 1 oracle transactions, et cetera, such that we can wean off of— make block rewards minimal so there's less inflation. These blockchains are supposed to eventually wean off of block rewards, right? Especially because Super is a fixed supply asset. So we need to do that. So we need to focus on revenue now. So our moonshot kind of aspiration is to be off of like 90% of block rewards within 3 years, driven purely from the other value capture that we're able to— that's frankly leaked out to MEV. I'll be talking a lot about this and getting into the details of how we do it in my keynote at Permissionless, by the way, in a week. Yeah.

Oh, nice. Love it. Love it. Yeah, people should get a ticket to Permissionless. One week out. Let's go. Okay, let's talk about revenue because you've mentioned the word revenue a lot. While you were talking, I actually just pulled up your guys' white paper and did a Command+F for the word revenue, and it's mentioned 43 times, which I think is the most I've ever seen in the white paper for an L1. Tell me why— um, I was going to say, tell me why this is important. Obviously everyone knows why revenue is important because it, you know, generates the money, and you generate money and stuff like that. But tell me your view on, like, how there's obviously, like, the revenue meta on Twitter and, like, fundamentals. And it's funny that it's even, like, a meta to make money or buy but, uh, tell me how you think the industry is changing and maybe like this shift into fundamentals and shift into revenue being important. And I'm curious if you would agree or disagree with that take.

Oh, I totally agree. I mean, I think it's— how do you have a sustainable system, you know, that's not dependent on good feelings, you know, and memes, right? Like, oh yeah, we're in it for the— we're going to change the future someday. But like, you know, we need to get back to reality. Like, you know, traditional finance and Startups, et cetera, SaaS, they all need to prove this stuff. Now let's talk about the various fees out there. So a base fee, the transaction fee, what is that really there for in a decentralized network? It's actually, it's to prevent spamming, DoS, denial of service. And it's also fees. Those base fees need to be there to prevent the system from buckling, right? So it has to get exponentially more expensive as you approach your bandwidth capacity as well as execution capacity. Now, Supra is designed to handle so much throughput that we probably won't hit that for a long time. That's where fees are minuscule. You can't build a business model when your fees are minuscule, right? So once again, base fees are there to— blockchains, frankly, should be public goods, and fees are there to— these base fees are there just to make sure that you can't expand the network and cause a liveness failure, as well as if the system is approaching its execution limits. It needs to spike up in fees so that folks back off. Right. So that's base fees. Base fees are not a good business model as a result because they're designed to be as cheap as possible. Sure. If the network is constantly at maximum capacity, then it works. But like, you know, our system can handle hundreds of thousands of transactions perhaps. Right. And yet the whole industry combined is less than like 15,000 transactions per second. So we're not going to hit those limits anytime soon. Base fees is not a good business model. Now, there are a couple other avenues here. So in the world of blockchain, it'd be so nice if we could do first come, first serve and prove it. But the reason you can't is because these nodes are distributed around the world. Data is submitted from all people around the world. They land at different nodes at a different time, right? So the block producer actually has a lot of control over the ordering of the transactions. Okay. And therefore the introduction of like priority fees or tips has been introduced. Like, oh, there's a high value transaction. I want to be at the top of that or above this other one. I will pay you a priority fee to include me. Right. So this is kind of a typical thing that's happening on Ethereum, Solana, where you have— you pay for higher your fee, the higher probability you're included in the block. Of course, first come, first serve would be the best, but it just doesn't work in a decentralized network because of the distribution of the nodes. So the priority fees, it's basically the thing is this, right? Who gets those fees? It's that node operator at that particular time. It's not the whole ecosystem, it's not all the other stakeholders, it's not necessarily to the dApp or to the token holders. Now, folks introduce burning as a way to kind of, uh, provide benefit to everyone. However, when you look at the numbers, when you compare it to the actual value that's been extracted through MEV, maximum extractable value, my understanding is it's like less than 5%. The burning is less than 5% of the value that's been extracted. Also, um, Solana at the moment in this example, I believe, has something on the order of 5% inflation recently, and it's also like less than 5% of inflation. So you're still inflating 95%, something like this ballpark, right? And meaning they're not really having an effect. So if you don't stake on Solana, you're getting inflated away, right? And the intention was to have fees that are burned that benefit everyone. But if you're only burning a small percentage of the base fee and these priority fees or whatever it might be, and they're only 5% of the inflation, then it doesn't really do anything. Moreover, on this kind of lane of thinking, what's happening is, back to priority fees, is how do you incentivize node operators to include your transactions or your bundle? Because they might not even know what's in your bundle. They're just looking at, oh, you're willing to pay me this much money? Okay, I'll include you, right? Because they're revenue maximizing entities. Which creates this kind of toxic MEV cycle, which is you have the searchers, they're looking for the maximum opportunities of front-run sandwich attacks to create the largest opportunity so that they keep like 70% of their revenue, but they can say, node operator, here's 30% of it. This is maximally extracted. You get paid more. Include me. What this translates to, though, for the end consumer is worse execution. right? Yeah, yeah. And that's not great for the end consumer, cuz in, in the end, we're here to serve the end consumer. Uh, and if they keep on not getting the execution they expect, that becomes a problem. Now imagine you had a system where you could capture that, that, that behavior, the extracted value, and redistribute it in real time. Now that is where I will be talking about and presenting it with slides and everything and going deep into it in Permissionless. So do show up, um, in a week, I think, right? It's, it's really around the corner. Uh, this is gonna be a very compelling conversation because how do you— what you have to do, and the reason why it gets a little bit technical, is because you have to have mechanisms where you mitigate toxic MEV. You also have to delineate between what is good MEV versus bad MEV. Toxic MEV is front-running, sandwich attacks. Good MEV is arbitrage and timely liquidations, right, that keeps spreads tight, that keep DeFi protocols solvent. Now we want to prevent the toxic MEV through our various mechanisms we have. Now I'll go into it in the presentation more, but more or less we do randomized execution. We have localized fee markets to make it expensive to spam, and then also encryption, decryption, just-in-time decryption. So you don't even know what the transaction is in the first place, so you can't front run it. So suppose we have that, but we also have this good MEV, right, which is timely arbitrage and timely liquidations. Meaning it's better for the end consumers. This is nuanced, but I think some of you folks, your audience, would understand it. Arbitrage is not bad, it's just neutral. And having tight, good arbitrage allows spreads to be tighter to the actual market price, which results in better execution and more liquidity generally for the end consumer. They get what they expected. Right now, it's not necessarily happening on Ethereum's line. Sometimes you There was like a crazy thing where someone put $250K in and got $5,000 out of it for a swap, right? Like 3 or 4 months ago on Ethereum. That was wild. But anyways, the other is liquidations. Why is that so important? If you have a DeFi protocol that's underprovisioned, it actually is underwater. You need to solve that to keep the solvency of the system. So you want to have fast resolution of liquidations. This is why Aave, Chainlink, and Flashbots are collaborating. My understanding is that Aave has oracle-driven liquidations. Chainlink, when they, when they're producing their price feeds, though, that may trigger a liquidation. By the way, liquidations are large revenue-generating opportunities. Like, uh, it could be up to 5% or more. But let's suppose 5% on $1 million is $50,000 to trigger that transaction. Okay, that value is usually being siphoned out by external actors. I think Aave's intention here is to work with Chainlink, know when that liquidation or that, that price feed is going to liquidate that position, work with Flashbots to bundle it and route the value back to the Aave and Chainlink and Flashbots ecosystems capturing it. So my point here is saying that is dApps are not necessarily capturing any of this MEV, nor is the layer 1 ecosystem Now, what we're doing is we're designing a system where you can capture this. And this is achieved by, as I mentioned, we have automation first and retail goes first. They pay for their automation fees. But at the end of every single block, any opportunity that's missed, Super will run our system level automation to capture the arbitrage, to capture the liquidations that were missed, capturing that value and redistributing it. Treasury, the dApp where it came from, node operators. And I guess it's important to mention that we also have a— the auction allows us to create— I'm sorry, the automation allows us to create an auction model where folks we believe will bid on priority execution of automation that gets basically executed at the top of every block or after the Oracle update, rather. Now, what does that And I'll try to keep it simple. Well, the best analogy I have, or the comparison, is what drives 80% of Google's revenue, 95% of Meta's, Facebook's revenue, is ad networks. They run auctions basically. And why does that work? Because there's a limited amount of capacity. Uh, there's a limited amount of ad units out there, and folks will bid for priority display of their ad unit to drive traffic to their website. Similarly, in blockchains, there's limited amount of execution capacity. There's limited amount of automation priority. If there is a large position, let's say a $10 million or $1 million liquidation potentially about to happen that is worth $50,000, like a 5% liquidation fee perhaps, right? People would actually bid to be the first one to execute that, starting maybe at $1,000. Eventually, theoretically, someone might bid up to $49,000 to be the one to execute that. Knowing that if they are the one that executes it, they make $50,000, right? So that automation priority auction, that revenue also can get captured by the treasury to the dApp where the opportunity came from, redistributed to node operators, replacing toxic MEV. So this is what we're going for. And this basically socializes MEV to the good actors of the system, the stakeholders, the believers, not siphoned off to very sophisticated genius-level MEV searchers, right? I mean, what they do is really fascinating. It's pretty smart. But it's not really fair because it's like 1% is of these really insiders that are just so brilliant that they're able to take this value from the other 99%. But what if we can capture it, mitigate that? It's a more fair system. Yeah, that's basically our philosophy.

Nice. Love that. That's awesome, Josh. I'm excited about the presentation at Permissionless. It's going to be good. Maybe as we shift gears a little bit, you guys, this is a more less focused on Supra and more just, I mean, you guys launched the token 6 months ago. What do you think about— there seems sometimes it seems like in crypto there are the haves and have-nots, or it seems like some people are in with the cool kids or deemed cool by the in-crowd and others are not. How do you think about this dynamic in crypto? I was actually just talking about the fact with the founder of another L1 the other day. They've been around for like 7 years and just for one reason or another haven't really gained traction. Their token's been out for for years and years and years. And he's like, I just don't get it. We're just not in the cool kid crowd. How do you think about this weird dynamic that exists in crypto?

Yeah, so there is such thing as a cool crowd. There always will be. It was in high school, it's in life in general. But does it always make sense? And the thing is this, right? If you cannot not necessarily get into the cool crowd, right? And Supra struggled with this a lot, you know, because, uh, as Americans, me and my co-founders, we left America to go to Asia. And, you know, when you're in Asia as a foreigner, you're not always accepted either, right? And, um, uh, the, you know, part of that though is that we tapped out of the— we didn't really tap into the cool crowd by not being in the States directly, going to all the USA events, meeting like, you know, you guys years ago. We only met last year, first time, right? And the thing is, there's basically one, you're in the cool crowd and you get a lot of boost in the beginning, but you still have to prove yourself, or you're not necessarily totally into the cool crowd yet. And especially because Supra, we are so contrarian. We're not friends with layer 2s, so to speak. I mean, we're not super friendly. Of course, I love the developments. We all do. But our thesis is counter to that. And that was like the main kind of focus of the last 2 years. We're also competing with oracles. We're also competing with bridges. Right. We're competing with pretty much— we're also competing with layer 1s. So we didn't really get that synergy of, oh yeah, we're working with LayerZero. By the way, we would love to work with LayerZero and Chainlink and others as a layer 1 itself. They will have services that we don't offer. So this is totally allowed. I just want to make that clear. This is not like we don't allow them in. No, no, they're welcome to deploy. We're welcome to work with them. I'll personally guide them. That's, that's fine. We would love that. But still, you know, there's kind of always a little bit of friction there because we're competing with everyone, like literally Layer 1s, Layer 2s, oracles, cross-chain communication protocols, automation, etc. Right. And, and as a result, I think there's less cross-promotion. Attention. You know, the only way to get out of that kind of bucket, to actually get the attention when you're in kind of that position, is to do something no one else is able to do that actually has product-market fit. This is why you just have to become a popular product on your own, right? You know, so this is why we are— we, we, we actually really believe there are one smart contracts with Oracle data, Threshold AI, Oracle Automation. We think this is going to have product market fit. We're moving automation to mainnet very soon, like in a week or two, and we're building all these automation open source automated primitives that anyone can mix and match and do things. We're maturing that stack. We think that this is going to be a very powerful tool for retail to craft their intention on chain and to execute these automation running their strategies, people, I think it can be very fun in that they'll start to monitor their performance relative to others and adapt it in real time. And it's, it's, it's real stakes. It's real money, right? So to speak. Um, so you do wanna have security, you wanna have determinism, but also, um, it, I, I think it can bring out that competitive nature. Now we all have also seen cases, um, I don't want to punch down here or anything, but, uh, we have basically our major competitor, going to market was Movement Labs because we're the third layer one to be Move language based. By the way, we are rolling out EVM support. It's already on testnet, so it's coming. We're going to be a multi-VM execution environment. But back then, I also never understood. I was like, wow, these guys are really great at promotion and connection and getting support because they're young. They're articulate, seem ambitious, right? And I was like, wow, I don't understand why we don't get that attention either. Of course, they spend most of the budget marketing than technology. But my point though is that we've also been accused of almost being too straight edge, maybe too research focused, too no backdoor, backroom deals, too not doing OTC deals, et cetera. And we're just like, well, we avoided the FTX collapse, Luna Terra collapse, and other things by keeping it simple, not playing. You know what I mean? So this is our habit, is just to keep it simple. Now, that said, I think for us, it's up to us to just prove it by building things and showing capabilities you can't do elsewhere to attract actual activity, real activity, not just speculation. And I think we're going to get there very soon in the next several months as it becomes more apparent, as more dApps deploy that are using automation to do things, to automate things like auto liquidations. They can deploy that after themselves that they may not have elsewhere. So, yeah, the truth is KOLs, media outlets, they do to a large degree, uh, VCs, but all really media, you guys kind of do have a tremendous effect and influence on the ecosystem to some degree, kind of like gatekeepers. Um, but also like, but that's natural. I think it's very natural way for things to, to evolve. So I think there's nothing wrong with it. And, and I think it's my job and Super's job is to keep on deploying, be steadfast, build something people actually want to use. For example, Hyperliquid, build a great product. Sure, it's not as decentralized, and actually our algorithms are better. Both the oracle and consensus are higher performing what they have, but they have like 10 nodes. We have 10 times larger network, right? 20 times, sometimes maybe 20 times larger. Nonetheless, they built a product that got so much traction that they earned attention. And that's pretty much my attitude, is we're not going to beg and say, oh, why, why not us? It's more like, no, well, it's our job is not to just have another layer 1. Let's do something different. Let's bring the industry to the future, like build something that people will need very soon. They may not even see it today. And that's kind of our point of view. So I think, I think we'll become the cool kids by next year.

Nice.

But, uh, yeah, it'll take some Nice.

Well, Josh, congrats on everything, man. Excited for, excited for Autophy and everything you're building. And yeah, again, thanks for being such a big supporter of Blockworks and excited for the presentation at Permissionless next week.

Yeah, thanks for having us. Appreciate it.