采访｜专注 ZK 投资的 Geometry 如何看待硬核技术公司的商业模式？
Scenius Studio 播客采访了 Geometry 的联合创始人 Tom Walton-Pocock。Tom Walton-Pocock 曾是 Aztec Network 联合创始人和 CEO。
Scenius Studio 的本期播客采访了 Geometry 的联合创始人 Tom Walton-Pocock。Geometry 是一家致力于支持密码学和零知识证明研究的公司，通过利用硬核的数学和密码学技术的协议来推动 Web3 的发展。他们的投资组合包括了与零知识技术相关的项目 Socket、Scroll、Cubist、Cartridge、RISC Zero、Herodotus、Ingonyama、Capsule、Slush，以及其他的 Web3 项目Cub3、Soap、MatchboxDAO、Magicave。
Tom Walton-Pocock 在英国剑桥大学攻读数学专业，获得本科和研究生学位。之后进入金融领域，并创建了 Layer2 项目 Aztec Network。基于他作为 Aztec Network 联合创始人和 CEO 的经验，Tom 与 Kobi Gurkan 等人于 2021 年共同创立了 Geometry。
关于创立 Geometry 的动机，他认为硬核技术公司在寻求融资时，由于许多投资者对硬核技术的理解不足，无法与公司建立共鸣。为了弥补这种技术理解的差距，研究驱动的 VC 可以帮助投资者更好地了解公司的技术实力，做出更明智的投资决策。
其次，Tom 看到了加密货币和区块链领域的巨大潜力，以及这一领域的技术前沿性和复杂性，需要具备深厚的技术背景和知识才能进行投资和研究。大多数投资者缺乏这方面的技术知识，而他和 Kobi Gurkan 等人则具备专业知识和经验，能够更好地理解和评估这一领域的投资机会和技术创新。
「零知识证明和密码学是未来互联网的核心技术，可以涵盖未来市场的很大一部分，因此成立一个以此为主要投资方向的基金具有很大的前景和机会。」
在 Tom 看来，Geometry 的目标是确保不具有商业经验的技术背景创始人能够成功地启动商业协议。对于那些拥有深厚技术背景的创始人来说，他们可能会热衷于技术本身，而不擅长解决现实的商业问题和找到真正的用户。因此，Tom 及其团队为这些创始人提供研究合作、技术支持的同时协助他们开发商业目标，寻找第一个用户，使产品实现商业化运营。
「Geometry 团队的目标是帮助技术背景的创始人将技术转化为具有商业价值的产品。」
在 Web3 生态系统中，实现技术协议的商业化是一项困难的任务。Web3 的使命是减少对中心化机构的依赖，这对商业模式的发展不利，因为商业模式通常需要建立在信任的基础上。然而，许多协议和密码学开发正在削弱对信任的需求，这使得技术协议更加难以形成商业模式。Tom 认为有一小部分与协议相关的代币可以成为宏观资产，类似于黄金这样的储值资产，例如 BTC、ETH 等。这些代币可以被视为一种货币形式，其价格可能非常高，但这需要全球范围内对其储值属性的认可。然而，对于有些协议，例如与以太坊相关的 Layer2 协议，虽然它们可以通过提高交易速度和增加以太坊的计算能力来缓解瓶颈问题，但这些网络的代币是否能够产生未来的现金流仍存在争议。因此，协议的商业模式需要根据其特定情况进行评估和制定。
Tom 还向大家介绍了 Geometry 投资的两个早期项目—— Scroll 和 RISC Zero。这两个项目都致力于解决不同类型的开发人员进入以太坊的门槛问题，以扩大以太坊的用户群。Scroll 专注于解决熟悉 JavaScript 的开发人员在使用 EVM 编程语言 Solidity 时遇到的问题，而 RISC Zero 则针对那些熟悉更为正式的编程语言如 C++ 的开发人员，它可以将使用这些编程语言编写的代码编译为以太坊智能合约所需的代码。因此，熟悉 Rust、Go 或 C++ 等编程语言的开发者就可以直接编写智能合约。
当被主持人 Ben Jacobs 问及大多数创始人和加密货币投资者忽略的事物时，Tom 分两部分阐述了他的观点。他认为在加密世界中，Layer2 协议层面上的价值提取情况非常不明确，赚钱的目标与保护用户的冲动并不一致。Layer2 的安全性将来自于更规范的资产，如 ETH，这种资产比本地代币更好地提供安全性。
此外，他指出 AI 对于加密货币的投资价值有着非常重要的影响。AI 已经使计算系统高度可变，产生高度风险，加密货币作为一种更不可变、更坚固的基础设施资产，将变得非常有价值。
「如果想要在投资 AI 的领域占有优势，就需要对加密货币以同样的重视来审视。」
在加密货币之外，Tom 认为，虚拟现实（VR）是未来工作的一个重要工具，由于 COVID-19 的出现，以高度去中心化形式建立的公司使虚拟环境的房地产变得非常有价值，VR 也可能对房地产和其他资产有重大影响，目前许多人低估了它。
以上是本次采访中的重点总结，Tom 还提及一些 ZK 和 EVM 相关的基础话题，感兴趣的朋友可以收听播客或者查看以下用 AI 生成的中英文文本。
播客 🔗
中英文全文本
Ben：Hello, and welcome to Scenius Studio, hosted by me, Ben Jacobs of Scenius Capital. With this podcast, we aim to give listeners inside access to the best and brightest investors in the crypto asset management industry. In each episode, I will chat with a leading crypto and blockchain venture fund or hedge fund manager as we explore the complexities of operating an investment fund at the bleeding edge of innovation. In this episode, I sit down with Tom Walton-Pocock, managing founder of Geometry. Building on his experience as a co-founder and CEO of the Aztec Network, Tom launched Geometry on the premise that deeply technical founders need investors who can relate to their engineering-first approach to business building. Geometry is now a leading cryptography and mathematics research firm with a track record of successful early-stage blockchain venture deals. Let's get into it.
「Ben Jacobs is a partner at Scenius Capital Management. All views expressed by Ben and the guests of this podcast are solely their opinions and do not reflect the opinions of Scenius Capital Management. Guests and the host may maintain positions in the assets and funds discussed in this podcast. You should not treat any opinion expressed by anyone on this podcast as a specific inducement to make a particular investment or follow a particular strategy, but only as an expression of their personal opinion. This podcast is for informational purposes only. 」
Ben：Hello and welcome to Scenius Studio. I am your host, Ben Jacobs of Scenius Capital. Today's guest is calling in from the UK and we are delighted to have Tom Walpoe of Geometry. How's it going, Tom?
Tom：Thanks for having me, Ben. Very well. How are you?
Ben：I'm doing well. Where exactly are you calling in from?
Tom：So I'm calling in from a small city about an hour south of London, but London is definitely where Geometry has, I guess, sort of most people, but we're pretty sprinkled all over the world from SF, New York, right across to Singapore. So yeah, most of my work's done from here.
Ben：Great. Well, we have a lot we want to jump into today, talk about your background, geometry, and then your perspective on a number of deeply technical topics that we haven't really touched on yet at Scenius Studio. But before we get into that, I know that your background is actually in music. And so I wanted to hear a little bit about that, and then maybe how you've drawn parallels from your music background and skills to your current work in crypto.
Tom：So I really do want to overstate the music thing. This was a best of glancing blow. It was a very brief stint. And really, my background as an undergraduate was in maths, and then I ended up doing some singing, and then that became a bit more serious, and I took it seriously for about six, nine months. And the market spoke. I was one of those lucky people, I think it was an awful midway where if you are talented, but not hyper talented, you might still go after the music. And I was fortunate to be below that rung. And the market spoke pretty quickly. And so I was never put in any kind of question, am I going to be threatening anyone's jobs at Covent Garden or the Metropolitan Opera or whatever. So I feel like I had it easy. Maybe there's an element of overlap between the experience of being a singer and entrepreneur, I guess, in terms of mentality and sort of battle resilience and that sort of thing. But I think that they're probably pretty, pretty minimal.
Ben：Got it. Well, I'd love to touch on your background and experiences, both from where you went to college, grew up, but then how all of that led to your current work in the formation of geometry.
Tom：Yeah, so I grew up on a wee island called Jersey, the name that will be familiar to you, but this is old Jersey, after which New Jersey was named. It's a tiny little lump of granite floating in the English Channel. It's a lovely place. And it's kind of between the UK and France. Maybe what typified the existence there was you always slightly less well connected to the world. The internet started coming online, I guess we were probably always very slightly behind. And so maybe some information flows were sort of slightly less good there at the time. Went to university in Cambridge in the UK. I studied maths at undergraduate and then postgraduate. And there's this kind of, I still have my postgrad photo on my wall. It's kind of amazing how many people in there are now actually leading pretty major projects, including Ethereum and WorldCoin and various other really important pieces of research driven infrastructure. Maybe it's not a surprise, there's lots of mathematicians in a room basically. It's a very brief stint in music, bit of time in finance, and then co-founded a company called Aztec, which is a layer two network. So this is kind of one of these companies that is scaling Ethereum. I guess we're going to talk a little bit about scaling during the course of this podcast. And it's done a lot of the work to create the sort of the cryptographic tricks that enable you to make these systems faster without imperiling their decentralization. So this new kind of hard consensus substrate, which is going to be the basis, I think, for all future computation. You need to make sure that it remains hard and that it remains fully decentralized. And it was an encryption trick that was created. I say it was a trick, it was a really impressive piece of work, not done by me, but by my co-founder, Zach Williamson, and our chief scientist, Ariel Gabasson, that I think is now forming the basis of a lot of the scaling that Ethereum is now able to enjoy. And we can chat about that maybe later on. And as far as geometry is concerned, geometry arose really sort of not long after I left Aztec. I co-founded this with Kobi Gurkan, and we wanted geometry to be a new flavor of investor. So it would be a research-driven company first. So we do a lot of work in applied cryptography, these things, zero-knowledge proofs, which we can also touch on later on. And we also use that as a mechanism by which to invest. So it sort of acts as a way to generate inbound traffic that allows geometry to invest and justify the higher cost of operating geometry as a fund.
Ben：What was the gap in the market that you identified that you thought, given your experience at Aztec and where you thought there was opportunity, how you thought it made sense to form this research-driven investment fund?
Tom：Yeah. So I guess this could break it into two questions. One is sort of why that particular interest, and the second is sort of why a research fund at all? I think sort of, maybe answer the why a research fund. I noticed a lot of the time that when deeply technical companies were looking to raise money, they were struggling to form a rapport with the investor, who I guess suffered from a sort of a lack of deeply technical understanding of the company. It was therefore maybe a sort of lack of empathy between the investor and the company. And the investor then typically went where the most technical people were. And I think investors, they're constantly trying to work out how can we justify our position in an increasingly heated investment environment? How can we sort of justify why we should be on the cap table rather than someone else? And a lot of the time, investment funds maybe make promises, "Oh, we'll make introductions to other people in our population." I think that can work for very powerful investors, but otherwise I think you just need a USB. And then the sort of question of why particularly did we pick zero knowledge proofs and cryptography? Really this sounds like a very narrow interest, but actually it's likely to devour all computation on this future internet which we are building. The zero knowledge proof basically sits at the corner of everything that you are going to want to do on this new decentralized verified internet that we are now piling into. And so it's a very universal thing to be interested in because ultimately every time you want to interact with this internet, there will be this thing as zero knowledge proof that is created, executed, verified. And so it really allows you to sort of to see a very wide section of this future market as we felt that from a strategic perspective, it's a great place to start. And then I think also it was the commonality between me and Kobe. It was the thing that sort of, I guess, united our interests and seemed like a very natural place to start. And we still wanted to keep really closely for real, for not just investment reasons, but for reasons of intellectual interest, this is one of the most fertile areas of intellectual output. Obviously we're very parochial and the way we think about this is we breathe and eat and this stuff and we spend a lot of time in it, but I really think this has been one of the most sort of fertile periods in the advance of computation sort of through history. There have been several of these, but I think this really has that level of importance. And so we wanted to stay close to the research.
Ben：I wanted to double click on your comment about founders with deeply technical backgrounds. What do you find most investment firms aren't able to understand about these founders and how does geometry work to solve that and ensure that these deeply technical founders who may not have traditional business experience, that they're able to launch a successful business protocol, et cetera?
Tom：Yeah. So I think it basically boils down to what motivates the founders. So, you know, it's a very hackneyed thing to kind of, to comb back over history and pick out the Steve Jobs and the Bezos and all these wonderful founders. And the thing that very clearly unites them is an utter obsession over the user and the ease of use and addressing the mainstream user as much as they can. It could be said that the computation and all this computer industry grew up on a diet of hobbyists, basically financing these companies to be able to care about some mainstream user that could exist. But thereafter it was all about sort of the mainstream user and caring a lot about them. The interesting thing I think that possibly plagues deep tech founders is they come in with a thirst, a passion for the intellectual stuff. So they've come straight up possibly out of a university department. They may have worked at some deep tech startup before, but it usually will have been at the cutting edge of the research, not necessarily having to do that sort of existential problem of how do we actually find a real user? And you could say that crypto in its entirety has been plagued by the fact that it is amazing economics meets amazing mathematics, and it's really struggled to find a, I would argue, a sort of a long-term user so far. It's been built up on a diet of hobbyists. These hobbyists have actually largely been financially driven so far, and it's now really seeking its long-term end user. So the trouble is someone comes in and they say, "We've got this great thing, MPC, multi-party computation. We've got this great thing, homomorphic encryption, or this great thing, zero-knowledge proofs," or whatever it might be. And so they end up, instead of it being a light bulb moment, "Ah, this user needs this. Where's the technology?" It tends to be, "I've got this amazing thing. My old accelerators used to call it the orb, this magical piece of technology that animates them and animates all their friends." And then they spend years, and as arguably I think Aztec did for a while, and we definitely wrestled with this, you have this magic thing, but what does the magic thing do for the user? And so the one thing we always make sure that we do is try to find somewhere in the DNA of at least one of the founders, there is that obsessive impulse over a future end user, even if the end user is at the moment possibly ill-defined, possibly too general. We always try to make sure that they have the capacity to care about an end user, so that when that phase change happens in the company, research, research, research, research, product, the product is going to take over, and the product is not going to be hamstrung by the desires of the rest of the founding team to essentially run a faculty and not a business. So what does running a research-driven VC fund mean in practice for those founders that you just referenced? How are you supporting them both in terms of developing the orb, but then also bringing that orb and developing a commercial purpose around it? Yes, so we've had several ways. First of all, very often we end up doing research collaborations before we invest in a company. Often, actually, that's at least on occasion, that's how the relationship is built. We often do, we support the technical part of the founding team with some of the early breakthroughs they need to make. I think sort of a notable one lately has been Anika Monblatt's work on Sangria, which is basically a way to make these proofs even faster. We can go into more detail, but it's basically a sort of, in some ways, it's a recapitulation of things we've been doing before, but it's a kind of different layer of the cryptographic stack. And so we have that sort of very pure research-driven interaction. And I'd say probably what I bring to the team more is someone who has sort of carried the existential dread for three, four years, sort of worked out how to make the horse lead the cart. So I spend a lot more time sort of trying to work on the joinery of the business to the encryption system or the business, so this new widget, and try to draw out a definition of that first user, which very often, because the R&D looks like such hell, it's such an uphill struggle, and there's so much to do, and there's so much performance gains to make, very often that can get lost in the discussion. So I think that's where the team is quite well balanced to help on both sides.
Ben：What are, thinking through the successful projects over the last three, four years, a lot of them have been token incentive-driven and just playing financial games. What are the business models that you have conviction in that can be applied to these more technical protocols that are being developed that you guys are backing at Geometry? What are, I guess, a couple different examples of which this deeply technical work can be monetized and developed into a real business?
Tom：Yeah, okay, so the funny thing about a lot of these protocols and the development of cryptography is it takes a lot of the, what do you call it, so the residues of trust, which actually you often look for in a business to be able to justify your ability to charge for it, right? You need the whole thing about stickiness is I can charge you more if it's much harder for you to leave me or if some actor could do something bad to your information. So if I'm Oracle, I could forget all the information I've stored in my database or I could mutate it, right? So you're prepared to pay Oracle for a whole load of services, but one of the core services is please don't screw up. And that please don't screw up-ism is kind of extracted from the risks of doing stuff on Web3. Web3, actually, its whole mission has been about try to reduce the dependencies on centralized people, which is also structurally not always great for the business model. So again, that's what we look for. Where are the trust residues likely to be still left after this amazing work has been done? So maybe some good examples. So first of all, I'm probably a believer that there is some small clutch of protocol attached tokens that can have value without cash flow. And I think the reason is they are striving for candidacy as a macro asset. They want to be a store of value asset that can sit alongside gold or currencies or whatever. And so sort of, let's say Bitcoin, Ether, and maybe there's a couple of other sort of canonical layer one assets that can make this claim. So they can basically become money-like. They can essentially sit at a crazily elevated cash multiple. And that's fine because the world has agreed that they're a store of value. It still leaves an article of faith at the base of this thing, but that's fine. So then there's what we call the layer two network. So this has been, maybe we should also talk about sort of structurally how Ethereum is scaling differently to other networks. We can do that. But there's this thing called the layer two network, which is essentially it borrows off the security of Ethereum. So it kind of, in the long run, it is as secure as Ethereum. There's some stuff that it's allowed to do to kind of make transaction volumes faster and increase the amount of computation that Ethereum actually can host. And these layer two networks have also had tokens attached. Now, I think the jury is really out on whether they can, those tokens, have a future claim to cash flow. There are kind of two things to examine here. One is, is there a mechanism by which they can extract cash? And that brings in a whole load of questions like, well, from a regulatory standpoint, is this a security or does it get classed as a royalties? Somehow it kind of sits outside that framework. Does it need to generate cash? My view is probably it does. I'm not sure that the layer two has quite the canonical role in the world that it does at a layer one. You know, Ether has a very sort of central position to the entire financial machinery of Ethereum. Can the same be said for a layer two token? Not sure. But there are a lot of people sitting around the layer two network who will help it to operate where there are going to be these trust relationships and there are going to be basically input costs that cannot go away to how they operate. So a good example is, I mean, we've invested in a company called Ingeniawa. There are several of these companies that are doing acceleration of the type of computation that is going to drive all of this future internet. Basically putting together these things called zero-knowledge proofs. Like, basically it means any program that's going to go into the system, it has to go through this process of being turned into a succinct proof. And the reason it has to go through that process is that Ethereum can't check billions of transactions per second. But the succinct proofs are a way of relieving Ethereum of having to just check everything. It can just check something that allows it to know that it's all true without checking every single detail, right? So that means every piece of software, any code, anything will end up running through this zero-knowledge proof creation system. And so there are particular hardware providers that, very like the CPU, we had the GPU, which focused just on graphics. This thing called the ZPU, or ZPU, which will just focus on the construction of this very important paradigm because it's going to be everywhere. It doesn't matter what code you're running, what computer program you're interacting with, you will need a conversion to a set KPI. So that's maybe a good example of where you have to follow the line of cash flow. You have to work out, can the layer two make money? Don't know. Jury's out. The other thing about the layer two is it probably has to freeze up over time, which is kind of the opposite of what a company does. A company obviously needs to keep on innovating to drop its margin. These blockchain infrastructures have this interesting thing where the more money they host and the more people are relying on it, the less they can rebuild themselves in midair, right? Ethereum has just done this for arguably the last time. Further tweaks to Ethereum probably are now tweaks. And the layer twos might end up following that same pattern. And so that might further question, is there any impulse to extract cash? Is there any mechanism or desire to extract more than the cost of operation from the user and therefore be a good return at the protocol level? But there's loads of businesses you can build around that, like hardware, possibly even software services that help engineers to maintain their software. Bear in mind, if you're running smart contracts in the future, the threats around blockchain programs are potentially an order of magnitude higher than sort of classical programs or classical websites or whatever. And so the role of software that is constantly monitoring the threats for blockchain programs can probably extract a lot of value from engineering teams. There's lots of opportunity. It probably creates a lot of opportunity, but it also deletes all business models. Of course, we've seen this with tech lots of times, so it shouldn't necessarily surprise us.
Ben：I think this is a good time to dive into the Merriam-Webster section of today's episode. Let's start with ZK Proof, just because that's a core tenant of Geometry's fun thesis and a lot of where the innovation is going towards right now. And it seems to be a key primitive that has unlocked the next wave for blockchains. What is ZK Proof and what is the history of it? Because I find that fascinating. And why is it important?
Tom：Sure. So, ZK Proof, first of all, the first moment you encounter this thing, it seems really unlikely that it exists. It's a very odd thing. So, what are we trying to do? We're trying to allow Ethereum, let's say, or some layer one blockchain to process tens of millions, hundreds of millions, billions of transactions all at the same time. You can't do it, right? Because it only can run as fast as the slowest computer checking the system. So, what the ZK Proof enables for blockchains, and by the way, ZK Proof is a bit of a misnomer, should really be called succinct proof. And the historical reasons why we use this bad word, zero knowledge is a bit misleading. It's basically a way of taking a computation of which, let's say, 100 million transactions, that would be a big computation. And instead of asking a node on the Ethereum network, like basically everyone participating in the Ethereum network to check this thing, you can do a much bigger amount of computation. And in moral exchange for this big computation that does this big argument, produce a tiny proof that shows that those millions of transactions are all correct, which is a crazy thing to say. So, in other words, I can convince you that the result of a change in the information on Ethereum from one block to the next was authentically updated by, let's say, a million or 10 million or however many transactions, all compiled together, and you need to do no more work than it would cost you to check just a few transactions. I can really get almost limitless scaling. Now, this sort of inherits from a body of work that started in the 1980s. I actually explained this on another podcast, I'll try to find another way to explain it, but essentially, maths had a bit of an existential crisis. Mathematics, I should say, or math, I should say, on this podcast, it had an existential crisis. So, proofs in maths were getting really long. People were having to sort of basically, there was one called the four-color theorem, which basically says that you can color any map in four colors. And it was impossible for mathematicians to check it was true because it had been reduced to thousands of configurations, 500-page proofs, so no one could check it. So, this is really like a blockchain. Blockchains have too much stuff to check. And so, the maths community couldn't check these proofs, and this was a problem. And so, this sort of kicked off a body of work at MIT where they were trying to work out, is it actually possible to allow someone to verify that a computation has correctly happened and there are no mistakes, but not actually have to look at the original computation. And this is kind of where it all arose from. Blockchain really gave the sort of the economic stimulus to that going from a sort of, theoretically, this is possible, who's ever going to use it, to an actual application. It actually arose first in the context of privacy. So, one thing maybe I've not explained on podcasts in the past, I think this is really interesting, privacy and scaling are really, really closely related phenomena. And here's why. If I want to do a private transaction, let's say on Ethereum, I do a ZK proof that means I want to show you the transaction was true, but I don't want to share the information with you. And the scaling setting says I want to prove the transaction or batch of transactions is true, but I don't want to have to send you all the information because you can't possibly check it all. Now those are different motivations, but it's the same paradigm. I'm not going to send you all the information, but I still want you to be able to verify it. So, the same thing, the same kind of mathematical requirement arises out of both privacy and scaling. We actually saw it in privacy first, 2013, Zcash, 2013, 2014, something like that. But it's actually now come to the rescue of things like Ethereum as a scaling mechanism. So, yeah, these ZK proofs, they're everywhere, they're really important. And they are probably, I think without them, I don't know how we would have scaled Ethereum and it had no chance of ever catching up with the surge capacity of Visa, which can do up to 60,000, 70,000 transactions a second. We would have just not had no hope. The close correlation of privacy and scaling is an interesting one. And so I'd like to make that tangible via a number of the ZK EVM, Ethereum Virtual Machine, scaling solutions that are coming live now. We saw Polygon launch at ZK EVM, Scroll, and basically everything that's not built on optimistic rollup technology.
Ben：Can you explain what, and maybe we use Scroll as an example, what does Scroll do? Why is it important? And what is the EVM component of it?
Tom：Yeah, so the way I tend to analogize this is like the kind of microchip semiconductor industry, going back a wee bit. So what have we started with? We started with this layer one. What are the technologies it brought to us? It brought to us basically a new type of computational substance, something that's really hard and really immutable we've not had before, and is going to become incredibly important in the coming years. And I'll explain why later on in the show. It carries with it, it needs to carry with it, an agreement on how you write computer programs. So that's what a microchip does.
Obviously it's got these kind of instructions to which every program you ever run, Microsoft Word, Microsoft Excel, whatever it is, everything is compiling down to that same set of a finite number of instructions. Often it's a matter of a few dozen or a few hundred, and every bit like an alphabet, every single computer program can be built out of those instructions. The EBM was the first such instruction set to appear on a blockchain as a way to codify all programs. Now a lot of people are not sure whether it's necessarily the best design, but there was a whole lot of constraints they were operating with at the time, but it was the first one. And very like, maybe a good example is JavaScript people, a lot of serious engineers who don't particularly like JavaScript, but it kind of, it was the sort of the go-to market for browser programming. In a similar way, the EVM has that role in blockchains. It's still become really, really important, and you see now the EVM everywhere on other layer ones, etc. So a natural thing for layer two networks, whose job was scaling Ethereum, was to say, you know, a lot of secure programs have already been written in this thing. Things like Aave and MakerDAO, they're all known to be secure when they're written in this Ethereum script. This is an Ethereum sort of assembly set. So we need to find a way to scale those exact programs, which are already known to be secure, already have, they're very heavily battle-tested, they carry billions of dollars, and so therefore we need to fit this thing into a zero-knowledge proof, so that we can now do lots and lots and lots of these same transactions in the same computing paradigm, but at the layer two. And that's exactly, as you say, it's what Scroll has done, it's what ZK-Sync has done, is that they're taking this instruction set and they are putting it into zero-knowledge proofs. Now it turns out that's a really hard thing to do, and unfortunately there are one or two historical choices that were made in how the Ethereum virtual machine was designed. I won't get into sort of details, but it made them really bad for fitting inside zero-knowledge proofs, but because when the Ethereum virtual machine was designed, no one realized zero-knowledge proofs were the great chariot coming to the rescue, they did some stuff around the design of the EVM that actually meant that when you had to put the EVM into zero-knowledge proofs to scale it, it actually took longer. It required more brainpower to fit the Ethereum virtual machine into zero-knowledge proofs, because no one saw zero-knowledge proofs coming as a scaling solution.
Ben：What's the implication of zero-knowledge proofs and the launch of Scroll for the end user, for example? What benefits does all of this technology bring to the average person walking on the street who's interested in doing something on-chain? The short term and the long term.
Tom：The short term is, it isn't going to make your transactions an order of magnitude cheaper, because you are now basically having your transactions bundled up with lots of other people's transactions, and Ethereum just has to check one consolidated proof to know that all that batch of transactions was secure. Now, there's some reasons why actually you need to make some extra assumptions to get even better scaling, but that's the short answer, is you just get cheaper transactions for the same stuff. Well, I think by far and away the more important thing is this now opens up the basket of ways that engineers can innovate, so they can now build more complicated programs, because those more complicated programs were unaffordably expensive to run on Ethereum mainnet, and now they're about to get an order of magnitude cheaper. So the more interesting thing from the investment standpoint is, what are now all of the programs that can now be built, that couldn't be built for the user because they were too complex? You have a very simple example, is something like an order book, an exchange, right? So that the amount of searching that needed to be done across all of the bids and asks in a classical exchange environment with self-custody, that was way too complex to put onto a blockchain because it's just prohibitively expensive. And so that is maybe a good example of something that now can be built on EVM or some other computing paradigm because of zero-knowledge proofs that was prohibitively expensive. By the way, very interestingly, because it was prohibitively expensive, that's why we got Uniswap, right? We got something that was computationally simpler, and it was really like this sort of creature born deep under the sea. There are certain types of creatures that can survive at the very lowest levels of ocean depths where the amount of pressure in the ocean is so high. And so you get very particular types of creatures, and they're not maybe the most sophisticated-looking types of creatures, but they're very sort of hardy creatures. But as that kind of pressure releases, as things get cheaper on blockchains, you can get more sophisticated sort of economic organisms, if you like, more sophisticated pieces of software.
Ben：Interesting comparison to Uniswap and how it was bred of a need for simplicity at a time when the tech wasn't quite there. As the leader of geometry, your job is to look into the future and stay ahead of trends and make sure that your team is conducting research on where the most bleeding edge tech is going. With your crystal ball, where are you steering the ship? And to ensure that geometry is always looking to what blockchain and crypto technology needs, looking past ZK, which has now really become a core component of blockchain infrastructure, what is the next big leap forward that we should be anticipating?
Tom： So it's something that people are starting to work on now. And I don't want to sound like one of those people that sort of suddenly has seen AI bolt from the stables and suddenly that's all I can think about. AI does usher in a couple of interesting opportunities, both from the research side and also from the investment standpoint. So first of all, I think AI has just made crypto investable. I think it's also made AI investable, but it's certainly made, in my view, crypto investable because crypto for a long time, I heard one investor describe the view of crypto as a very gothic view. You really have to believe a lot of stuff is going to go wrong for crypto to be sort of universally relevant rather than just sort of staying in lane and just handling these kind of macro assets. And I think AI might be the calling. That might be the answer, is you now have these programs, they get to decide not just how they execute, but when or whether they execute. This means that the storage on which we all rely for an account of our history, things like files, pieces of work you've done, all of that stuff is sort of increasingly potentially mutable. And so the role of crypto as a computational backend for the world, for the new internet, is increasingly important. That level of redundancy and verification that it offers is incredibly important. So that's kind of one thing to say. From a research standpoint, this is quite interesting because, and it means that by the way, the course of research that will interest geometry will remain thematically quite similar for at least the next 12 months, which is that zero-knowledge proofs now need to get a lot, lot, lot, lot faster. So if you're now going to have a world where the internet needs to be increasingly verified, that means that when people run artificial intelligence, or let's say sort of at a lower level, sort of machine learning models, and increasingly if the world is moving from this sort of compiled computation environment where you have all programs are hard-coded by an engineer, and they all go down into this instruction set, and instead increasingly the utilities that we're using online are more kind of machine learning generated, they're sort of more intuitive in that sense, intuitive compute versus maybe sort of classical compute. It's going to be very important to know that they were executed correctly when they are operating on your state. If you think blockchain state right now, yeah, it's your assets. In the long run, this could be everything that you are storing, everything you're doing on the internet, that in the long run, backed up essentially, sort of executed on this sort of decentralized network. So this means secondary proofs need to get a lot faster. These machine learning models are enormous, and so whereas before what we needed to be able to do was do a unisort, well it's pretty simple, it's x times y, it needs a really simple equation, and now you've got to verify whether an enormous model, billions or trillions of parameters, numbers flowing all over, it was actually very simple to get a lot better. So that's where these two worlds, machine learning and crypto, made very well life and produces a requirement for massive improvements in the improving system we've all been working with way in the past.
Ben：I like to think of this end user and who the two different primary groups are. There's the average consumer, and then there's corporates, institutions, businesses, et cetera.Why is ZK, verification, security, privacy, why is that so important to that second cohort, the businesses, the companies, so that they can start getting more involved on chain?
Tom：Right, well first of all I think the first beneficiary is are not on chain. So these zero knowledge proofs are massively important off chain. On chain basically just means I'm recording the order in which transactions happened, but it might be that it's really important for you to check that a computation happened but you're not on chain. Here's a great example. So you've always, as a basic input to this, you've always got to look for where is the cost of not knowing that a machine learning model was executed massively costly. Where is either a huge financial cost or a compliance cost or something, right? That's where you've got to be staring to work out where to apply this technology. So a really good example is medtech. So increasingly we're going to see, for example, oncology. So people sort of, you know, doctors, until now it's been doctors stare at a scan, try to work out is there a tumor, is there a medical problem with this person. And increasingly the move will be towards AI or sort of machine learning, right? So you actually run some model and it's able to detect all these extra features that the human eye is not able to detect from the input data. So great example of this technology would be like, can you imagine the medical, the sort of the indemnity breach that you would have if you couldn't prove a paper trace and maybe someone had just failed to run the correct model, the authorized model that is the one that is known to be the best at detecting cancers, for example, would be monumental. So great example of proving that you executed a machine learning model would be oncology. And so accompanying your scan, wherever it goes, would be this proof that says the best in class known algorithm for detecting cancer. Here's a succinct proof that you can check to make sure that the model actually ran. That's a really good example of where sort of the cost of producing the proof is not that high compared with the massive cost of not having a paper trail that proves that you did the thing that was medically the correct thing to do using early versions of AI or ML.
Ben：So I think it's kind of your job helping steer these very technical teams think about that end user. Likely the oncologist has deep knowledge of oncology, but may know nothing about crypto or zero knowledge proofs. And so that user experience, like the complexity of the crypto and the mathematics needs to be abstracted to the background. Given the very technical nature of these topics, how can these teams think about the UX and is that where geometry and other investors could step in and really try and drive an emphasis on that topic?
Tom：Indeed. I mean, the first thing to say here is that as the cryptography team, you are probably not going to be a very good person to sell directly to medical specialists. So you're going to need really to think about platformizing yourself. So what you should really be doing is providing a computation service, some kind of SDK, whatever, some sort of API service. And what it does is, and now there's a whole lot of privacy issues here, so you probably need the proof to be run sort of locally near the client so they're not broadcasting all their information. But you probably, your job is to build a great engineering experience that just allows the person building the model to sort of compile the model into a proof and create that proof. So you've got to team, team one is we're a med tech company. We're trying to create a more reliable way of detecting cancer. Who is their customer? Their customer is the medical profession, so some hospital, whatever, national health service in the UK, whatever it might be. And they are onward paying for some much more general platform whose job it is to create the certificate. So, you know, a kind of similar example would be back when we were all building these early websites. And you know that little lock in the corner of your screen when you're on your browser? That's an SSL certificate, right? You did not want website builders having to build their own SSL certificate. Also, there were standards that were required to make sure that a server and a client could interact with one another. So you needed that provided by some B2B business. It's that B2B business that is really the sort of company that Geometry would fund. In other words, they are building this ZK certification for all these applications that can't go wrong. And they make it, as you say, really easy for those engineering teams. So those engineering teams, they're probably product specialists. They probably do know their own market quite well. They know the medical profession. They know the defense profession. There's a whole load of professions that this could ultimately serve, as well as on-chain applications. And they are interfacing with those end users. And they simply have as an input, this thing that makes all the cryptography complication go away from them. And so, yeah, your bird's base is talking about an engineering tool. But it's a very powerful engineering tool, partly because the amount of R&D left to make this stuff more and more and more and more powerful. It's sort of almost limitless. And so there's a kind of great long-run business in being one of those ZKP providers.
Ben：So I wanted to save some time here to hear about some of the projects that Geometry has backed and why you're excited about them and what they offer to the world. So maybe we could do two examples here of interesting projects that you're excited to share with the audience.
Tom：Yeah, so kind of an interesting pairing that looks at the world in two different ways is sort of Scroll and RISC Zero, both of whom we were amongst the very earliest investors in. So as you mentioned Scroll earlier, they're doing this thing, it's like AVM. So this is addressing the idea that maybe EVM is JavaScript. And maybe therefore it has got network effects. And because the first group of people use it, the second wave of people use it, and the third wave of people use it, so everyone's going to be writing solidity code for a long time. And these are very strong arguments to say that is very likely to be true for a long period of time. RISC Zero comes in and says, "Well, we think there's also an untapped group of people who have been writing in much more formal languages for a long time, are not used to the weirdnesses of Ethereum, but they do know all these other languages. And these things all compile down into a particular architecture." So they're going after really the untapped Web2 engineers who just want to write programs for this new internet, but they don't want to learn a new language. So that's quite an interesting pairing. These companies are basically sort of operate at the same layer. They're basically scaling companies, they're sort of operating at the sort of the layer two level, but they're taking very different views of the world. What do engineers want? And probably the answer is both. There's probably a whole group of engineers who are hooked on Ethereum and that trend will continue. But there's also a lot of people who prefer to use more formal languages, C++ for us, but these sort of things that are much more formal, or that they've just sort of used sort of elsewhere for 20 years and they want to come in to build on Ethereum and so they can do that too. So that's a kind of interesting pair of companies. Again, the question for them is going to be, what is their business model? It might be something around coordinating computation. So a little bit like sort of Uber on the streets, their orchestration job is to make sure that there is always computation available to sort of basically create the proofs that allows Ethereum to sort of ingest these programs. So that's kind of one good example as if maybe a company that can sort of feed off both of those and capture value is a company like Ingonyama. So where they're building dedicated hardware and the betting there is just as we had the GPU of phones, we're going to have this CPU thing which just focuses on making zero knowledge proofs faster. So they're sort of essentially are an input to both of those companies and to many more as well. I think that probably sort of neatly encapsulates how we are trying to sort of find people who can extract value around the edges of these protocols, whose main job is going to be to sort of make the bad stuff go away. And maybe it's just one point of historical comparison. It's again a very hackneyed comparison. We've seen these kinds of businesses in the past where there is a protocol and there is a trustful business, right? And so, you know, Mongo is the one that everyone likes to use. So Mongo has this open source sort of database. It's a standard. Well, that could be very like the layer two protocol, right? It's a standard. It's a thing you know can't go wrong. But then you need these trustful providers to provide the computation to make sure your database doesn't go away. And that's obviously what Mongo does. There could be similar analogies at layer two. So there's the kind of the trustless business that's the protocol to make sure nothing goes wrong. And then the sort of the trustful business, which is the provision of computational resources to that protocol. So that's maybe a sort of good illustration of how we see some value being extracted in the long run. I find it interesting how you're, you have one central thesis, but then you're backing projects that are approaching that thesis in different directions. And they could both be correct, but just serve a different demographic group. It's almost like being LVMH and knowing that this type of luxury product serves this type of customer and this luxury product may serve a different type of customer without having any judgment on what luxury is. Precisely. And we also just don't know long run which direction the cash flow is going. I actually had, there's a great podcast called Another Podcast by- Another Not As Great But Still Great Podcast. Not as great as this one, but it is also very good. And it covers a totally different topic, by the way. And a guy called Benedict Evans, who I'm sure plenty of your listeners have come across, who was Andreessen for a while. And he gave a great example of, and I didn't know this anecdote, but of the early radio broadcasters. And they had no idea how to make money. And apparently the idea of advertising was at that point invisible to them. Presumably they had the sense that unless the advertisement is sort of visual and is sort of entertaining and visually arresting, you can't do advertising. I'm not sure what it was, but they had this idea that the cash flow at that time was only really happening in the shop where people were buying their transistor radios or whatever, their wireless sets. And so their thesis in the early days was, we monetize by recharging the radio manufacturers for a cut of the profits they're making off the end user. And that's a sort of great example, of course, that's precisely the opposite way that the cash flows ended up flowing. I think that same level of uncertainty really accords to crypto, to Web3 right now. And that's why we are trying as much as possible to sort of, to get some economic coverage of all the people who might end up being trustful providers around these otherwise trustless networks as a way to make sure that we're going to return value to investors.
Ben：What is invisible to the majority of founders and crypto investors out there that is clear as day to you in geometry?
Tom：The first is maybe a sort of a negative or at least a check against our assumptions, which is, I think the value extraction case for layer twos is extremely unclear at the protocol level. And I need to emphasize that. It doesn't mean you can't build huge businesses around it. And a lot of these layer two companies can. But the reason is that the impulse to make money does not sit particularly well with the impulse to stop bad things from happening for the user. So the first thing I think is going to happen in the long run is in layer twos, everyone is going to be staking ETH. In other words, the security for these layer twos will come probably from a more canonical asset, one that's better expressing security, that's better giving security, gives security more cheaply than a native token. So I think, therefore, there is a big question of whether the tokens themselves can value extract from one. Number two, and I think I have said it higher up the program, but I really think that AI has much or at least as much consequence for the investability of crypto as it does the investability of AI. With AI, it's very hard because either you've invested in a foundation model, which looks like it's in a very powerful position, but again, cash flow is not particularly clear. Or you invest in an application where everything comes down to the fine tuning and probably those applications can monetize. But are they actually always going to be at the behest of whoever built that foundation model? That's a big open question. Whereas AI has made computational systems highly mutable, highly at risk. Think of all your bank statements. I mean, I don't want to cause too much of a scare, but sort of every piece of data that you think is hard historical evidence that tells you where your assets are, what they are, is increasingly mutable and open to everything from cyber attack through to... Now, this is a very sort of gruesome view of the world, but I think it's an elevated risk that AI can now start to operate on computational systems in much more intuitive ways and mean that our collective record of history is much flimsier. And I think that means the crypto substance is now highly investible. It's a highly valuable asset to have your assets and all of your computational state recorded in this much more immutable, much harder substance that crypto provides. So I think if you want to invest in the virtues of AI, you need to at least be taking as hard a stare at crypto as AI itself. I think that's your spicy take within crypto, but I'm going to ask the same question. What is your spiciest take outside of crypto? What do you got? Well, I can now say it because it's been very... I just don't understand. It's been so trendy to disavow the role of virtual reality. And in particular, I feel a lot of scorn has been poured on Zuckerberg for his VR output so far. I don't know how many people, these people actually tried it. It's actually really good for the kinds of businesses that are now being created in very dispersed environments. So geometry has used this quite regularly in the past. For, we need a lecture room. We have 10 people all around the world, all in different cities. How do you do a whiteboarding session with those people? And VR was a great answer to that. They have completely eliminated the latency. Yeah, you get slight headaches. Yeah, okay. If you take a snapshot of what you see on the screen, it doesn't look immediately arresting versus the output of mid-journey or something. That's right. But the actual work that Facebook has done to get the experience so good, to get this sort of the acoustic replication so good, the user experience of sort of, you know, actually white writing on a whiteboard with your hand flailing in mid-air, amazing. I actually do think that this is the future of work and I don't think it's so very far away. And there are a lot of companies that have been built up in highly dispersed formats because of COVID. It's very natural. It's suddenly very natural. You notice this actually in the portfolio sort of two, three years ago, the idea of starting a company with a co-founder who's in a different country was unthinkable. Now it's the default in a lot of settings. And so this means that the, sort of the real estate of a virtual environment is highly, highly, highly valuable. Now what that then does to sort of all kinds of asset classes, I don't know, does it mean we all will find out going to the countryside because we have much less use for cities because we have, cities have much less of social draw. Or does it actually mean the flip side? We're much less sensitive to constrained space. And so actually we're very happy to live in highly built up environments because we don't need the outside space because we can just sort of push the wall open and fly out the window anytime we like in our VR. I have no idea. I think there's going to have some very interesting implications for classical real estate and other assets too, particularly some commercial real estate. But I think VR is going to be foundationally important to our lives when we're spending all our time. And if I just think people who are disavowing right now, there was a great hype cycle. Now people's excitement has gone. It's very classic. But I really think that that is one of the things that's going to be, going to imminently influence all of our working lives and all of our social life too. I'm in New York city right now. And just looking out at the amount of commercial real estate there is in so many of these big, you know, 10 story plus buildings, they must be vacant. With, it's just cheaper for the companies to allow people to work remotely and making use of an at home office. So I don't know what that means to the space, but there will have to be new use cases developed in order for that property not to be totally wasted. So I love the fact that Geometry's team is leveraging VR to coordinate a global organization. I think you guys are in seven or 10 countries. So the fact you're all able to congregate in a virtual room that has more depth to it than Zoom, I think is awesome. And yes, as you think of the Darwinian chances of survival of a company that has to hire only geographically versus the chances of a company that can hire entirely globally. And the only prior constraint was how do I put all these people in a room so that their communications don't get degraded by being split up and not having good communications. If actually ultimately VR fixes that, then yes, to me, I mean, it's easy to cycle this forward too fast, but I think it entirely replaces the role of the commercial office, totally replaces it. And so yes, I must admit, I would be interested to see what, I guess all that real estate gets repurposed as residential. And then the big question is, do people want that or do they want to actually move out into the country because they can get all of their sort of much of their meaningful social interaction through VR? I have no idea how that's going to play out.
Ben：It'll certainly be interesting to monitor. I think we've got some time now. Tom, this was awesome. Where can people learn more about you and geometry?
Tom：Yes, we have a website, www.geometry.xyz, I will say. I'm still training myself to say that. We have on there all kinds of things, sort of lists of our portfolio, a notebook, a health warning. It is very technical, but there are some sort of more digestible non-technical topics on there as well. That's where we put a lot of our research and otherwise, yep, we contact by email and also on Twitter as well. We have an increasingly popular Twitter account that started with quite a technical audience, but definitely broadening now.
Ben：Awesome. Well, thank you for coming on today's show. This was a fantastic episode and for teaching everyone about these deeply technical topics. I look forward to watching you and geometry succeed and going forward, staying at the cutting edge. Thank you for listening and we hope you enjoyed this episode of Scenius Studio. Please leave us a review and rating wherever you listen to podcasts if you love today's show. For more Scenius Capital content, check us out at sceniuscapital.substack.com and shoot me a follow on Twitter @bennypjacobs. We'll see you next time.
大家好，欢迎来到Scenius Studio，我是Scenius Capital的 Ben Jacobs。通过这个播客，我们旨在为听众提供加密资产管理行业中最优秀和最有才华的投资者的内部访问。在每一集中，我将与一位领先的加密货币和区块链风险基金或对冲基金经理进行交流，探索在创新的前沿运营投资基金的复杂性。在这一集中，我与Geometry的创始人Tom Walton-Pocock坐下来交谈。凭借他作为Aztec Network联合创始人和首席执行官的经验，Tom推出Geometry的前提是，资深技术创始人需要能够理解他们以工程为先的业务建设方法的投资者。Geometry现在是一家领先的密码学和数学研究公司，拥有成功的早期区块链风险基金交易记录。让我们开始吧。
「Ben Jacobs是Scenius Capital Management的合伙人。Ben和本播客的嘉宾表达的所有观点都仅代表他们个人的观点，不反映Scenius Capital Management的观点。嘉宾和主持人可能在本播客中讨论的资产和基金中保持头寸。您不应将任何在本播客中表达的观点视为具体诱因来做出特定的投资或遵循特定策略，而只应将其视为他们个人观点的表达。本播客仅供信息目的。 」
Ben：大家好，欢迎来到Scenius Studio。我是你的主持人Scenius Capital的Ben Jacobs。今天的嘉宾Tom Walpoe来自英国，我们很高兴有他参加。Tom是Geometry的创始人之一。你好，Tom，近况如何？
Tom：谢谢邀请，Ben。我很好。你呢？
Ben：我很好。你现在确切位置在哪里？
Tom：我现在在伦敦南部一小时车程的一个小城市里，但伦敦是Geometry的大本营，大多数人都在那里。我们遍布全球，从旧金山、纽约到新加坡，所以我的大部分工作都是在这里完成的。
Ben：好的。今天我们有很多想要探讨的话题，例如你的背景、Geometry，以及你对一些深度技术主题的看法，这些主题我们在Scenius Studio中还没有涉及到。在我们深入讨论之前，我知道你的背景实际上是音乐。所以我想听听关于这方面的一些经历，然后也许是你如何将你音乐背景和技能与你在加密领域的现在工作联系起来。
Tom：我真的不想太夸大音乐的事情。这只是一个短暂的经历。我的本科背景是数学，然后我开始唱歌，然后这变得更加认真了，我认真对待了大约六到九个月。市场发话了。我是那些幸运的人之一，我认为这是一个糟糕的中间位置，如果你有天赋，但不是超级有天赋，你可能仍然会去从事音乐。我很幸运地处于那个级别以下。市场很快就给了答案。所以我从来没有受到任何质疑，比如我会不会威胁到考文垂花园或大都会歌剧院等地的任何人的工作。所以我觉得我很轻松。也许在作为歌手和企业家的经验方面有一些重叠，至少在心态和战斗韧性方面是如此。但我认为它们可能非常微小。
Ben：嗯，我想了解你的背景和经历，包括你上大学的地方，成长经历以及所有这些如何导致你现在在Geometry的工作。
Tom：我在一个叫做泽西岛的小岛上长大，这个名字你可能熟悉，但这是老泽西，新泽西是以它命名的。它是一个漂浮在英吉利海峡中的小小花岗岩，是个可爱的地方。它位于英国和法国之间，也许最能代表那里的生活的是你总是稍微与世界隔绝一些。互联网开始上线，我想我们可能总是稍微落后一些。所以也许一些信息流在那里的时候稍微不那么流畅。我在英国的剑桥大学学习数学本科和研究生。我还保留着研究生时的照片挂在我的墙上。很惊人的是，现在这张照片中的很多人实际上正在领导一些重要的项目，包括以太坊和WorldCoin等各种重要的研究驱动的基础设施。也许这不是什么奇怪的事，因为房间里有很多数学家。我曾短暂涉足音乐界，花了一些时间在金融行业，然后创立了一个名为Aztec的公司，这是一个L2。这是这样一种公司，它正在扩展以太坊的规模。我想在本次播客过程中我们将会谈到一些关于规模扩展的问题。它已经完成了许多工作，创造了加密技巧，使你能够使这些系统更快，而不会危及其去中心化。因此，这种新的硬共识基础，我认为将是所有未来计算的基础。你需要确保它仍然很难，仍然完全去中心化。这是一个加密技巧，由我的联合创始人Zach Williamson和我们的首席科学家Ariel Gabasson所创造的。我认为现在它正在成为以太坊现在能够享受的许多扩展的基础。我们可以稍后讨论这个问题。至于Geometry方面，Geometry是我离开Aztec后不久出现的。我与Kobi Gurkan共同创立了这个公司，希望Geometry成为一种新的投资方式，以研究为驱动的公司。我们在应用密码学方面进行了大量的研究，包括零知识证明等。我们还利用这些技术作为投资的机制。这也成为了一种方式，可以带来投资的流量，使得Geometry可以合理地承担更高的运营成本。
Ben：你认为市场上的哪些差距，以及你在Aztec的经验，是什么让你认为成立一个以研究驱动的投资基金是有机会的？
Tom：好的。所以我想这可以分成两个问题。一个是为什么特别感兴趣，另一个是为什么要建立一个研究型基金？我认为，也许先回答为什么要建立一个研究型基金。我注意到，当技术公司在寻求融资时，他们往往很难与投资者建立关系，因为投资者缺乏对公司深层次技术的了解，这导致投资者和公司之间存在一种缺乏共鸣的情况。投资者通常会去寻找最懂技术的人。我认为投资者一直在努力找出如何在日益激烈的投资环境中证明自己的立场。他们如何证明自己应该在资本市场上，而不是其他人？很多时候，投资基金可能会做出承诺，“哦，我们会向我们的人群介绍其他人。”我认为这可能对非常强大的投资者有用，但除此之外我认为你只需要一个独特的卖点。那么，我们为什么特别选择零知识证明和密码学呢？实际上，这听起来可能是一种非常狭窄的兴趣，但实际上它可能会占据我们正在构建的未来互联网上的所有计算。零知识证明基本上处于我们现在正在积极推进的新去中心化验证互联网上所要做的任何事情的核心位置。因此，这是一个非常普遍的感兴趣的领域，因为每当你想要与这个互联网互动时，就会有这个零知识证明被创建、执行和验证。因此，从战略的角度来看，它真的允许你看到未来市场的广泛部分，我们感到这是一个很好的起点。此外，我认为这也是我和Kobe之间的共同点。这是将我们的兴趣联系在一起的东西，似乎是一个非常自然的开始。我们仍然想保持密切关注研究，这不仅仅是为了投资原因，还是出于对知识产出最肥沃的领域的兴趣。显然，我们非常地偏执，以至于我们思考这个问题时无时无刻不在接触、学习这方面的知识，但我真的认为这是计算历史上最肥沃的时期之一。已经有过几次这样的时期，但我认为这真的具有这种重要性水平。因此，我们希望保持密切关注研究。
Ben：关于你提到的具有深厚技术背景的创始人的问题，我想进一步讨论。投资公司最难理解这些创始人的哪些方面？Geometry是如何解决这个问题并确保这些深厚技术背景的创始人，尽管可能没有传统的商业经验，但能够成功地启动一个成功的商业协议等等呢？
Tom：我认为这基本上归结于创始人的动机。你知道，回顾历史并挑选出像Steve Jobs、Bezos 和其他伟大的创始人，这是一件非常陈词滥调的事情。很明显，将他们联系在一起的是对用户和易用性的极度痴迷，以及是尽可能地满足主流用户。可以说，计算机和整个计算机产业成长于业余爱好者的支持下，这些公司才开始关注可能存在的主流用户。但此后，一切都关注主流用户，并极度关注他们。我认为可能困扰深度技术创始人的有趣之处在于，他们带着对知识产生的渴望和热情进入这个行业。他们可能直接从大学系里走出来，可能之前在一些深度技术创业公司工作过，但通常是在研究的前沿，不一定需要解决如何实际找到真正的用户这样的重大问题。你可以说，整个加密行业被一个问题所困扰，即惊人的经济学和惊人的数学相遇，但到目前为止，它确实很难找到一个长期的用户。它是由业余爱好者的支持建立起来的。这些业余爱好者到目前为止实际上主要是出于财务驱动，现在它真的在寻求长期的最终用户。所以问题在于，有人进来说：“我们有这个伟大的东西，多方计算，同态加密，或零知识证明，或者其他任何东西。”于是他们会发现，不是灯泡瞬间，“啊，这个用户需要这个。技术在哪里？”它往往是：“我有这个神奇的东西，我的老加速器曾称之为球，这个神奇的技术驱动着他们和他们所有的朋友。”然后他们花了几年时间，就像Aztec曾经做过的那样，我们也确实曾经挣扎过，你有了这个神奇的东西，但这个神奇的东西对用户有什么用？所以我们总是确保至少有一位创始人的DNA里有对未来最终用户的执着冲动，即使此刻的最终用户可能没有明确的定义，可能过于笼统。我们总是努力确保他们有关注最终用户的能力，这样当公司发生阶段性变化时，即研究、产品，产品将接管，产品不会受到其余创始团队渴望运营大学而非企业的愿望的阻碍。那么，对于您刚刚提到的那些创始人来说，运行一个以研究驱动的VC基金意味着什么？在发展Orb的同时，您如何在商业目的周围提供支持？是的，我们有几种方式。首先，我们通常会在投资公司之前进行研究合作。实际上，通常情况下，这至少偶尔是建立关系的方式。我们经常为创始团队的技术部分提供一些早期突破所需的支持。最近一个值得注意的例子可能是Anika Monblatt在Sangria上的工作，这基本上是一种使这些证明更快的方法。我们可以深入探讨，但基本上它在某种程度上是我们以前做过的事情的重现，但它是加密栈的不同层。因此，我们有这种非常纯粹的研究驱动互动。我想我更多地为团队带来的是，我已经承受了三、四年的生存恐惧，努力让马拉着车走。因此，我花了更多时间试图处理业务与加密系统或业务与这个新小部件之间的连接，并尝试概括出第一个用户的定义，因为由于研发看起来像地狱一样，是一场艰难的攀登，有很多性能提升要做，因此通常会在讨论中丢失。因此，我认为团队在两个方面都能提供帮助，平衡得相当不错。
Ben：在过去三到四年中，我们思考成功的项目，很多都是基于代币激励的，只是玩财务游戏而已。你们在Geometry支持的这些更加技术性的协议中，有哪些商业模式让你们确信可以应用？我想知道有哪些不同的例子可以让这些深度技术性的工作变现并发展成真正的业务？
Tom：关于这些协议和密码学的开发，有趣的一点是，它需要很多信任残留物，这些信任残留物实际上是你通常在一个企业中寻找的，以证明你能够为此收费的能力，对吧？你需要粘性，如果你离开我越困难，我就能收你更多的钱，或者某个参与者可能会对你的信息造成不良影响。所以，如果我是Oracle，我可以忘记我在数据库中存储的所有信息，或者我可以变异它。所以你准备为Oracle支付一大堆服务费，但其中一个核心服务是请不要出错。这种请不要出错主义是从在Web3上做事情的风险中提取出来的，而Web3的整个使命是试图减少对中心化人员的依赖，这对商业模式也不总是有益的。所以，这就是我们寻找的地方，这些想法在这个惊人的工作完成后，哪些信任残留物可能仍然存在？所以可能有一些很好的例子。首先，我可能相信，有一小撮协议附加的代币可以在没有现金流的情况下具有价值。我认为原因是，它们正在争取成为宏观资产的候选人。它们希望成为一种储存价值的资产，可以与黄金、货币或其他东西并列。所以，比特币、以太币，也许还有几个其他经典的第一层资产可以做出这种主张。所以它们基本上可以变成像货币一样的东西。它们可以在疯狂的现金倍增下坐着。这没关系，因为全世界都同意它们是一种价值储存。它仍然留下了一篇信仰文章，但没关系。然后是我们所谓的L2。可能我们也应该谈谈以太坊在结构上如何不同于其他网络。我们可以做到这一点。但本质上，它借用了以太坊的安全性。从长远来看，它与以太坊一样安全。它允许一些操作来加快交易量并增加以太坊可以承载的计算量。这些L2也附带了代币。现在，我认为对于这些代币是否能够拥有未来的现金流权益，还需要观察。有两个方面需要考虑。第一个是，它们是否有提取现金的机制？这会引入一系列问题，例如，从监管的角度来看，这是一种证券还是属于版税？它有一些特殊的地位。它需要产生现金吗？我认为可能需要。我不确定L2在整个以太坊金融机构中是否拥有与第一层相同的重要作用。以太坊在整个以太坊金融机构中处于非常核心的位置。对于第二层代币，是否可以说同样的话呢？不确定。但是有很多人正在L2周围，他们将帮助其运作，建立信任关系，并且有一些基本的输入成本是无法消失的。比较好的例子是，我们投资了一家名为Ingeniawa的公司。有几家公司正在加速这种计算类型，这种计算类型将驱动未来的互联网。他们正在构建称为零知识证明的东西。这基本上意味着，任何要进入系统的程序都必须经过这个过程，转化为简洁的证明。它之所以必须经过这个过程，是因为以太坊无法每秒检查数十亿个交易。但是，简洁的证明是一种减轻以太坊负担的方法，它只需要检查一些内容，就可以知道所有内容都是正确的，而不必检查每个细节。因此，每个软件、任何代码、任何东西都最终将通过这个零知识证明创建系统运行。因此，有特定的硬件供应商专门负责构建这种非常重要的范例，就像CPU、GPU专注于图形一样。这个叫做ZPU或ZPU的东西将专注于构建这种重要的范例，因为它将无处不在。不管您运行的是什么代码，与什么计算机程序交互，都需要将其转换为一组KPI。这可能是一个很好的例子，需要跟踪现金流动。你需要弄清楚第二层是否可以赚钱。目前还没有定论。L2另一个问题是它可能随着时间的推移而冻结，这与公司的做法有些相反。公司显然需要不断创新以降低成本。这些区块链基础设施有一个有趣的特点，随着它们托管的资金越来越多，依赖它们的人越来越多，它们越来越无法在空中重建。以太坊刚刚完成了这个过程，很可能是最后一次。进一步调整以太坊可能现在只是微调。L2可能最终也会遵循这个模式。因此，这可能会进一步质疑，是否有任何冲动从中提取现金？是否有任何机制或愿望从用户那里提取超过运营成本的费用，从而在协议层面获得良好的回报？但围绕此可以建立许多业务，例如硬件，甚至可能是帮助工程师维护其软件的软件服务。请记住，如果未来运行智能合约，区块链程序的威胁可能比经典程序或经典网站高一个数量级。因此，不断监控区块链程序威胁的软件的角色可能可以从工程团队中提取大量价值。这可能会创造许多机会，但也会删除所有商业模型。当然，我们在技术领域看到这种情况很多次了，所以这并不一定会让我们感到惊讶。
Ben：我觉得现在是探究今天 Merriam-Webster 部分的好时机。我们从 ZK Proof 开始，因为它是 Geometry 有趣的论文的核心，而且现在很多创新都朝着这个方向进行。它似乎是解锁区块链下一波浪潮的关键基元。ZK Proof 是什么？它的历史是什么？它为什么重要？
Tom：ZK Proof 首先，你第一次遇到这个东西时，似乎很不可能存在。它是一件非常奇怪的事情。那么我们在尝试做什么呢？我们尝试让以太坊，比如说，或者某个第一层区块链能够同时处理数千万、数亿、数十亿的交易。你不能这样做，对吧？因为它只能运行得像检查系统的最慢计算机一样快。因此，ZK Proof 为区块链提供了什么，实际上是一种简洁证明，而 ZK Proof 这个词有些不准确，它实际上应该被称为简洁证明。历史上为什么我们使用了这个不太好的单词呢？零知识有点误导性。这基本上是一种方式，通过这种方式，可以进行一种计算，其中有 1 亿笔交易，这是一项巨大的计算。而不是让以太坊网络上的每个节点（基本上是参与以太坊网络的每个人）去检查这个东西，你可以进行更大量的计算。在这个做这个大论证的基础上，生产一个微小的证明，证明那些数百万笔交易都是正确的，这是一个疯狂的事情。换句话说，我可以说服你，从以太坊上的一个块到下一个块的信息更改的结果是由，假设是一百万或一千万或任何数量的交易，都是真实更新的，而你所需要做的工作不超过检查几笔交易的代价。我确实获得了几乎无限的扩展能力。现在，这实际上继承自上世纪八十年代开始的一系列工作。我在另一个播客中解释过这一点，我会尝试找到另一种解释方法，但基本上，数学经历了一些存在危机。在这个播客中，我应该说数学或数学，它经历了一次存在危机。因此，数学证明变得非常冗长。人们基本上不得不，有一个叫做四色定理，它基本上说你可以用四种颜色给任何地图上色。数学家无法验证它是否正确，因为它已经被简化成了成千上万个配置，500页的证明，所以没有人能够检查它。所以，这就像是区块链一样。区块链有太多的东西要检查。因此，数学界无法检查这些证明，这是一个问题。因此，在MIT开展了一系列的研究工作，他们试图弄清楚是否有可能让某人验证计算是否已经正确进行，且没有错误，而不必实际查看原始计算。这就是它的起源。区块链真正为它提供了经济刺激，从理论上来说，这是可能的，但谁会使用它，到实际应用。它实际上最初是在隐私的背景下出现的。所以，也许我以前在播客中没有解释清楚的一件事，我认为这非常有趣，隐私和扩展性是非常密切相关的现象。原因是什么呢？如果我想在以太坊上进行私人交易，我做一个零知识证明，这意味着我想向你展示交易是真实的，但我不想与你分享这些信息。而扩展性则意味着我想证明交易或批量交易是真实的，但我不想把所有信息都发给你，因为你不可能全部检查。现在这些是不同的动机，但是它们是相同的范例。我不会把所有信息都发给你，但我仍然希望你能够验证它。因此，隐私和扩展性都需要同样的数学要求。我们实际上首先在隐私方面看到了它，2013年Zcash，2013年或2014年左右。但它现在实际上成为以太坊扩展机制的救星。是的，这些零知识证明无处不在，它们非常重要。如果没有它们，我不知道我们如何扩展以太坊，并且它没有任何希望赶上Visa的激增能力，后者可以每秒处理高达60,000或70,000个交易。如果没有这些零知识证明，我们就没有任何希望。隐私和扩展性之间的密切相关性非常有趣。因此，我想通过现在正在推出的许多ZK EVM（以太坊虚拟机）扩展解决方案来使它具体化。
Ben：我们看到Polygon推出了ZK EVM，Scroll以及基本上所有没有建立在乐观的Rollup技术之上的东西。你能解释一下，我们以Scroll为例，Scroll是做什么的？为什么它很重要？以及它的EVM组件是什么？
Tom：我通常将其类比为微芯片半导体行业，稍微回顾一下。我们从什么开始？我们从第一层开始。它给我们带来了什么技术？它为我们带来了一种全新的计算物质，一种非常坚固和不可变的东西，在未来几年中将变得非常重要。我稍后会解释为什么。它需要带着一个关于如何编写计算机程序的协议。这就是微芯片所做的事情。当然，它具有这些指令，每个你运行的程序，如 Microsoft Word，Microsoft Excel 或其他任何东西，都会编译成相同的有限指令集。通常只有几十个或几百个指令，每个计算机程序都可以由这些指令构建而成，就像字母表一样。EVM 是第一个在区块链上出现的这样的指令集，作为编码所有程序的一种方式。现在许多人不确定它是否是最佳设计，但在当时他们面临了许多约束。它仍然变得非常重要，现在你可以在其他 L 1 上到处看到 EVM。因此，作为旨在扩展以太坊的 L2 网络的自然之举是说，已经编写了许多安全程序。例如 Aave 和 MakerDAO，它们都是以这种以太坊脚本编写的，已知是安全的。这是一组以太坊的汇编集。因此，我们需要找到一种方法来扩展这些已知安全的程序，它们已经经过了严格的测试，承载着数十亿美元，因此我们需要将这个东西放入零知识证明中，以便我们现在可以在 L2 中使用相同的计算范式执行大量的这些相同的交易。这正是 Scroll 所做的，以及 ZK-Sync 所做的，它们将这个指令集放入了零知识证明中。现在事实证明，这是一件非常困难的事情。不幸的是，EVM 的设计存在一些历史选择，我不会详细解释，但这使得它很难适合于零知识证明。这是因为当设计 EVM 时，没有人意识到零知识证明是解决方案，所以在 EVM 的设计中做了一些东西，实际上导致当你需要将 EVM 放入零知识证明中以扩展它时，它需要更多的脑力。
Ben：那么，零知识证明和 Scroll 的推出对于终端用户有什么影响呢？例如，这些技术为对链上感兴趣的普通人带来了哪些好处？
Tom：短期内，它不会使你的交易便宜十倍，因为现在你的交易被与许多其他人的交易捆绑在一起，以太坊只需检查一个综合证明即可知道所有这批交易都是安全的。现在有一些原因需要做一些额外的假设才能获得更好的扩展性，但简单回答是您只需为相同的东西支付更少的费用。我认为远远更重要的是，这现在打开了工程师创新的方式篮子，因为那些更复杂的程序以前无法在以太坊主网上运行得起，现在它们将便宜一个数量级。因此，从投资的角度来看，更有趣的事情是现在可以构建哪些以前无法构建的程序，因为它们太复杂了。一个非常简单的例子是订单簿和交易所。在自托管的经典交易所环境中，需要对所有买卖双方的竞价和要价进行搜索的量太大了，以至于将其放在区块链上是不可行的。这可能是一个很好的例子，说明由于零知识证明的原因，现在可以在EVM或其他计算范式上构建的东西，这是不可行的。顺便说一句，有趣的是，正是因为它是不可行的，我们才得到了Uniswap，对吧？我们得到了一个计算上更简单的东西，就像这样在海底深处孕育出的生物。有一些类型的生物可以在海洋最低深度的非常高压下生存。因此，你得到非常特殊的类型的生物，它们可能不是最复杂的生物，但它们是非常坚韧的生物。但随着区块链上的成本降低，你可以获得更复杂的经济生物，就像更复杂的软件一样。
Ben：与Uniswap的比较很有意思，它是在技术还不成熟的时候，出于对简单的需要而产生的。作为Geometry的领导者，你的工作是展望未来，走在潮流前沿，确保团队正在研究最前沿技术的方向。以你的智慧，你会引领这艘船向何处驶向？为了确保Geometry始终关注区块链和加密技术的需求，超越了现在已成为区块链基础设施核心组成部分的零知识证明，我们应该预期下一个重大的飞跃是什么？
Tom：这是一些人现在开始研究的事情。我不想听起来像是那些突然看到AI从稳定的状态突然飞跃的人之一。AI带来了一些有趣的机会，既从研究方面，也从投资角度。首先，我认为AI使加密货币具有投资价值。我认为它也使得AI具有投资价值，但在我看来，它确实使得加密货币具有投资价值，因为长期以来，我听到一位投资者描述了他们对加密货币的看法，那是一个非常哥特式的观点。你必须相信很多东西会出错，才能让加密货币成为普遍相关而不仅仅是处理这些宏观资产。我认为AI可能是答案，您现在有这些程序，它们不仅可以决定如何执行，还可以决定何时或是否执行。这意味着我们所有人依赖的存储历史的存储，例如文件、您完成的工作，所有这些东西都越来越可能是可变的。因此，加密货币作为世界上的计算后端，在新互联网中的角色变得越来越重要。它提供的冗余和验证水平非常重要。所以这是要说的一件事情。从研究角度来看，这非常有趣，这意味着，顺便说一下，对于至少未来12个月来说，将吸引Geometry的研究方向在主题上相当相似，即零知识证明现在需要变得更快。如果现在的互联网需要越来越多的验证，那么当人们运行人工智能，或者说更低层次的机器学习模型时，如果世界正在从这种编译计算环境中移动，其中所有程序都是由工程师硬编码的，并且它们都进入这个指令集，而越来越多的在线工具是更具机器学习生成的，它们在这个意义上更加直观，即直观计算而不是经典计算。在它们对你的状态进行操作时，知道它们是否被正确执行将非常重要。如果现在的区块链状态是您的资产，那么从长远来看，这可能是你存储的一切、在互联网上所做的一切，从根本上备份，以在这种去中心化网络上执行。这意味着次要证明需要变得更快。这些机器学习模型非常巨大，因此，以前我们需要能够做的是做一个unisort，好吧，它非常简单，它是x乘以y，它需要一个非常简单的等式，现在你必须验证一个巨大的模型，数十亿或数万亿的参数，数字流动到处，这实际上是非常简单的，可以得到很大的改进。因此，机器学习和加密两个世界相遇的地方可能会为我们之前一直使用的改进系统带来巨大的需求和改进要求。我喜欢考虑最终用户和两个不同主要群体是谁。有普通消费者，还有公司、机构、企业等。
Ben：为什么零知识证明、验证、安全、隐私对于那些第二组，即企业、公司非常重要，以便他们可以开始在链上更多地参与？
Tom：首先，我认为第一个受益者不在链上。因此，这些零知识证明在链下非常重要。链上基本上只是记录交易发生的顺序，但你可能需要检查计算是否发生过，但你不在链上。这里有一个很好的例子。所以，作为这个基本输入，你必须考虑不知道机器学习模型是否被执行的代价到底在哪里特别高。在哪里有巨大的财务成本或合规成本，等等？这就是你必须盯着的地方，以便确定在哪里应用这项技术。一个非常好的例子是医疗科技。越来越多的情况下，我们将看到比如肿瘤学等领域。到目前为止，医生一直盯着扫描结果，试图确定这个人是否有肿瘤或身体上的问题。但是，越来越多的情况下将会向AI或机器学习方向发展。你实际上运行了一些模型，它能够检测到人眼无法从输入数据中检测到的所有额外特征。因此，这项技术的一个很好的例子是：你能否想象如果无法证明一份纸质文件追踪，并且可能有人只是没有运行正确的模型或经过授权的模型，这是诊断癌症最佳模型，将会导致多么大的医疗风险。因此，证明你执行了机器学习模型的一个很好的例子就是肿瘤学。因此，伴随着扫描结果，不论它去到哪里，都将有一个证明，证明这是最佳模型。这是一份简明的证明，可以检查确保模型实际上运行了。这是一个非常好的例子，说明产生证明的成本与没有证明的医学上正确使用早期版本的AI或ML相比所产生的巨大成本相比是不高的。
Ben：我认为你的工作是帮助引导这些非常技术性的团队考虑最终用户。肿瘤科医生肯定对肿瘤学有深入的了解，但可能对加密或零知识证明一无所知。因此，用户体验，如加密和数学的复杂性，需要被抽象到后台。鉴于这些主题的非常技术性质，团队如何考虑用户体验，这是Geometry 和其他投资者可以介入并真正推动强调该主题的地方吗？
Tom：确实。我的意思是，作为加密团队，你可能不会直接向医疗专家销售。因此，你需要考虑自己的平台化。你真正应该做的是提供计算服务，某种SDK，任何API服务等。它的作用是，现在有很多隐私问题，所以你可能需要证明在客户端附近本地运行，以便他们不会广播所有信息。但你的工作是建立一个优秀的工程体验，使建模人员能够将模型编译成证明并创建证明。因此，你有一个团队，第一团队是我们是医疗技术公司。我们正在尝试创建一种更可靠的癌症检测方法。谁是他们的客户？他们的客户是医学专业人员，如医院，英国国家医疗服务等。他们向其支付更普遍平台的费用，该平台的工作是创建证书。因此，一个类似的例子是，当我们都在建立这些早期的网站时。你知道当你在浏览器上时，在屏幕角落有一个小锁吗？那是SSL证书，对吧？你不希望网站建设者不得不建立自己的SSL证书。此外，还需要标准来确保服务器和客户端可以相互交互。因此，你需要由某些B2B企业提供。这是真正需要几何投资的B2B企业。换句话说，他们正在为所有这些不能出错的应用程序构建这个ZK证书。并且正如你所说，他们让这些工程团队的工作变得非常容易。这些工程团队可能是产品专家。他们可能非常了解自己的市场。他们了解医疗行业，了解国防行业。这些应用程序最终可以为许多行业提供服务，包括链上应用程序。他们正在与最终用户进行交互。他们只需将这个使所有加密复杂性消失的东西作为输入即可。所以，你所说的是一个工程工具。但这是一个非常强大的工程工具，部分原因是因为剩下的研究和开发可以使这些东西变得更加强大。这种工具的商业前景非常广阔，成为ZKP提供商是一个长期的好生意。
Ben：为了节省时间，我想听一些Geometry支持的项目的介绍，以及为什么你对它们感到兴奋，以及它们为世界提供了什么。也许我们可以举两个有趣的例子来与听众分享。
Tom：第一个例子是Scroll和RISC Zero。我们是最早的投资者之一。正如你之前提到的，Scroll正在做的是类似于AVM的事情。这是为了应对EVM可能是JavaScript的想法。可能因此它具有网络效应。因为第一批人使用它，第二批人使用它，第三批人使用它，所以每个人都将长时间编写solidity代码。这些都是非常有力的论据，表明这很可能长期存在。RISC Zero则认为，“嗯，我们认为还有一些从事更正式语言编程的未开发的人群，他们不习惯以太坊的怪异之处，但他们确实知道所有这些其他语言。这些东西都编译成一个特定的架构。”所以他们针对的是真正未开发的Web2工程师，他们只想为这个新的互联网编写程序，但他们不想学习一种新语言。这是一个非常有趣的组合。这些公司基本上在同一层次上运作。它们基本上是扩展公司，它们在第二层级别上运作，但它们对世界的看法却非常不同。工程师想要什么？答案可能是两个都要。可能有一整群工程师沉迷于以太坊，这一趋势将继续下去。但也有很多人更喜欢使用更正式的语言，比如C++，但这些语言更为正式，或者他们在其他地方使用了20年，他们想来建立以太坊，所以他们也可以这样做。所以这是一个有趣的组合。对于它们来说，问题在于它们的商业模式是什么？它们可能会围绕计算进行协调。就像在街上的Uber一样，它们的编排工作是确保始终有计算可用，以基本上创建证明，使以太坊能够接受这些程序。所以这是一个很好的例子，也许一个可以从这两个公司中吸取价值并捕捉价值的公司是Ingonyama。因此，他们正在构建专用硬件，押注的是就像我们拥有手机GPU一样，我们将拥有这个CPU设备，它专注于使零知识证明更快。因此，他们基本上是这两个公司以及许多其他公司的输入。我认为这可能很好地概括了我们如何试图找到那些可以在这些协议的边缘提取价值的人，他们的主要工作将是消除糟糕的东西。也许只是一个历史比较的一个观点。这又是一个非常陈词滥调的比较。我们过去看到过这样的企业模式，就是一个协议和一个值得信赖的企业，对吧？所以，你知道，Mongo是每个人都喜欢使用的那个。因此，Mongo有这个开源的数据库。它是一个标准。这可能非常像第二层协议，对吧？它是一个标准。你知道它不会出问题。但是你需要这些值得信任的提供者来提供计算资源，以确保你的数据库不会消失。这显然就是Mongo所做的。L2可能存在类似的比喻。因此，有一种无需信任的企业是协议，以确保不会出错。然后还有一种值得信任的企业，即向该协议提供计算资源。因此，这可能是我们看到长期中一些价值被提取的好例子。我觉得有趣的是，你有一个中心论点，但是你支持以不同方向接近该论点的项目。它们都可以是正确的，但是服务于不同的人群。就像是LVMH，知道这种奢侈品服务于这种类型的客户，而这种奢侈品可能服务于另一种类型的客户，而不对奢侈品做出任何判断。正是如此。而且，我们长期来看也不知道现金流向哪个方向。我曾经听过一个名为Another Not As Great But Still Great Podcast的很棒的播客。顺便说一下，它涵盖了完全不同的话题。一个名叫Benedict Evans的人曾经在Andreessen工作过，他讲了一个很好的例子，我以前不知道，就是早期的广播电台。他们不知道如何赚钱。显然，广告的想法在那时是看不见的。他们可能认为，除非广告是视觉上引人注目的和娱乐性的，否则就无法进行广告。我不确定是什么原因，但他们在当时的想法是，我们通过从终端用户获取利润的一部分来重新收取广播制造商的费用来获得收益。这是一个很好的例子，当然，现金流的流向最终完全相反。我认为这种不确定性在加密货币和Web3中非常明显。这就是为什么我们尽可能地试图获得所有可能成为这些否则信任网络周围值得信任提供者的经济覆盖，以确保我们将向投资者返回价值的方式。
Ben：在Geometry中，什么是对于大多数创始人和加密货币投资者来说看不见的东西，而对你们来说却很清楚？
Tom：第一点也许是对我们的假设的一种否定或至少是检查，那就是我认为在协议层面上，对于二层的价值提取案例非常不清楚。我需要强调这一点。这并不意味着你不能在其周围建立巨大的业务，很多这些二层公司可以做到。但原因是赚钱的冲动并不特别适合阻止用户遭受不良事件的冲动。因此，我认为长期来看，二层中的每个人都会押注ETH。换句话说，这些二层的安全性可能来自于一种更加规范的资产，这种资产更好地表达了安全性，以更便宜的价格提供了安全性，而不是一种本地代币。因此，我认为有一个重要的问题，即这些代币本身能否从中提取价值。第二点，我认为我在节目的前面已经提到过了，但我真的认为AI对于加密货币的可投资性具有同样多的影响。对于AI来说，这很难，因为你要么投资于一个基础模型，看起来它处于非常强大的位置，但现金流不是特别清晰。要么你投资于一个应用程序，在这种情况下，一切都归结于微调，可能这些应用程序可以实现货币化。但是，它们是否总是会受到建立该基础模型的人的支配？这是一个重要的未知问题。而AI已经使计算系统高度可变、高度危险。想想你所有的银行对账单。我不想引起太多的恐慌，但每一份数据都是你认为是硬性历史证据，告诉你你的资产在哪里、是什么的证据，它们越来越易变，并且容易受到从网络攻击到...的一切威胁。这是一个非常可怕的世界观，但我认为AI现在可以以更加直观的方式在计算系统上操作，这意味着我们的历史记录集合变得更加脆弱。我认为这意味着加密货币物质现在是非常具有投资价值的。在这个更加不可变、更加坚固的物质中记录您的资产和所有计算状态是非常有价值的资产。因此，我认为如果你想投资于AI的优点，你至少需要像投资于AI本身一样认真看待加密货币。我认为这是你在加密货币领域最激进的看法，但我要问同样的问题。在加密货币领域之外，你最激进的看法是什么？你有什么看法？嗯，我现在可以说出来了，因为这已经很……不知道为什么现在很时髦否认虚拟现实的作用。特别是，我觉得对于扎克伯格迄今为止的虚拟现实产品，有很多鄙视的声音。我不知道有多少人实际上尝试过它。实际上，它非常适合正在创建的那些分散环境下的业务。因此，Geometry在过去经常使用它。例如，我们需要一个讲堂。我们有来自世界各地、不同城市的10个人。如何与这些人进行白板会议？虚拟现实是一个很好的答案。它完全消除了延迟。是的，你会有轻微的头痛。是的，没错。如果你拍下屏幕上所看到的画面，与旅途中间输出的画面相比，它并不会立即引起注意。没错。但是Facebook所做的实际工作，使得体验非常好，声音复制也非常好，用户体验非常好，你知道，用手在空中挥舞在白板上写字，这是惊人的。我真的认为这是未来的工作方式，而且我认为它并不遥远。由于COVID-19的原因，许多公司已经建立了高度分散的格式，这是非常自然的。你会发现这在两三年前的组合中，与不同国家的联合创始人共同创立一家公司的想法是不可想象的。现在在许多情况下，这已经成为了默认设置。这意味着虚拟环境的房地产非常有价值。现在这对于各种资产类别的影响，我不知道，这意味着我们所有人都会去乡村吗？因为我们几乎不再需要城市，因为城市的社交吸引力减少了很多。还是说反过来？我们对受限空间的敏感度降低了，所以我们很愿意住在高度建筑密集的环境中，因为我们不需要外部空间，我们可以随时在我们的VR中推开墙壁和窗户飞出去。我不知道。我认为这将对经典的房地产和其他资产产生一些非常有趣的影响，特别是一些商业房地产。但我认为VR将成为我们生活中的基础，并且当我们花费所有的时间时，VR对我们的工作和社交生活将有即刻的影响。我现在在纽约市。只要看看这里有多少商业房地产，就知道很多这些大型建筑物，10层以上的楼房，必须是空置的。对于公司来说，让人们远程工作并利用家庭办公室是更便宜的。所以我不知道这意味着什么，但将必须开发新的用例，以便这些物业不会被浪费。所以我喜欢Geometry团队利用VR来协调全球组织的方式。我认为你们在七个或十个国家。所以你们都能聚集在一个有更深度的虚拟房间里，比Zoom更好，我认为这很棒。当你考虑到一个公司只能雇佣当地人的生存达尔文机会与可以全球招聘的公司之间的差别时，前一个限制是如何将所有人放在一个房间中，以便他们的沟通不会因分散和沟通不佳而受到损害。如果VR最终解决了这个问题，那么对我来说，我觉得它完全取代了商业办公室的角色，完全取代了它。因此，我必须承认，我很想看看那些房地产如何被重新用作住宅。然后一个大问题是，人们想要这样吗，还是他们想要实际搬到乡下，因为他们可以通过VR获得他们的大部分有意义的社交互动？我不知道这会如何发展。
Ben：这肯定是一个有趣的趋势，我们需要些时间来观察。Tom，非常感谢你的分享。请问人们如何了解你和Geometry呢？
Tom：是的，我们有一个网站www.geometry.xyz，我得承认我还在适应这个名字。我们在网站上有各种东西，包括我们的投资组合列表、笔记、健康警告。这些内容很技术性，但也有一些比较易懂的非技术性话题。我们在那里发布我们的研究成果，此外，我们也通过电子邮件和Twitter联系。我们的Twitter账户越来越受欢迎，开始是比较技术性的观众，但现在正在逐渐扩大范围。
Ben：太好了。感谢你今天的节目，非常棒，也让我们了解到这些深度技术话题。期待你和Geometry的成功，并继续保持在前沿。感谢大家收听本期Scenius Studio节目，如果您喜欢今天的节目，请在您收听播客的平台上给我们留下评论和评分。如果您想了解更多Scenius Capital的内容，请关注我们的Substack账号sceniuscapital.substack.com，同时欢迎在Twitter上关注我@bennypjacobs。我们下期节目再见。