The Exchanges

Q theme of optimizing for learning, and one area where you mentioned a couple of times before, which is the element of crypto, you've gained huge amounts of respect and knowledge in the last few years for your work in decentralization and crypto. However, in the space where I have to admit that there's no mass market consumer decentralized product today, I'm intrigued, Adam. Why do you think this is fundamentally?

A Yeah, well, thanks. It's, it's great to know that, uh, people think I have knowledge. I must've been fooling them somehow. But, um, to your question, I think there's several reasons. I think that the first and most obvious one is that the underlying infrastructure can't support mass market dApps. So, CryptoKitties alone clogged up the Ethereum blockchain. If you think about it from sort of a higher level perspective, you The Visa network processes thousands of transactions per second, and that's only doing payments. How can you build apps on top of something that only can really do a couple dozen transactions per second? I think another reason is the lack of emphasis on user experience. So on the side of blockchain-based games, I don't think the barrier is really technology. I think it's a lack of interesting game mechanics. You could build a game like Fortnite in a centralized manner while having its cosmetic assets live on the blockchain. That's a very low And I think in other reasons that mass market consumers want to pay for DApp and for services in a unit of account that they understand. So this is why stable coins are extremely important. It's hard to have an e-commerce or e-commerce use case without a unit of account. So what this all stacks up to is that the biggest problem is that people are really excited about blockchain, and when they're building things, often tout …

AI assessment note: “the first and most obvious one is that the underlying infrastructure can't support mass market dApps”

Q Not at all, but I want to kick off today. Rarely do I get the pleasure of having someone almost as young as me on the show. So tell me, how did you make your way into the wonderful world of venture at the tender age of 22, Adam?

A Well, come on, that was almost two years ago. In essence, I was... Fortunate enough to be able to accelerate my education. So I was homeschooled until I was 10, and then I went to a community college for a few years because I really liked mathematics. So after that, I went to UC Berkeley at 13 to keep studying math. I finished at 16, but I didn't really feel ready to go out into the world. I didn't want to be a math professor or something like a quant. So I decided to go back to college and do undergrad again, this time at Stanford, and to study computer science. So, uh, big games are difficult for me, as you might imagine, but, uh, I always root for the home team. And along the way, I spent time at several research labs at Palantir, at Dropbox, and I was the first product manager at Rubrik, one of our portfolio companies in the enterprise infrastructure space. But the thing that I felt most fortunate about is that when I studied math, I focused on number theory and cryptography, and when I studied computer science, I focused on systems and distributed systems. So those two things marry very nicely together into blockchain. And although I didn't start investing professionally until a couple of years ago, I was involved in a first rounds dorm room fund when I was at Stanford. And I think that as someone with an unusual background, I have a lot of empathy for founders and realize…

AI assessment note: “I was involved in a first rounds dorm room fund when I was at Stanford”

Q theme of optimizing for learning, and one area where you mentioned a couple of times before, which is the element of crypto, you've gained huge amounts of respect and knowledge in the last few years for your work in decentralization and crypto. However, in the space where I have to admit that there's no mass market consumer decentralized product today, I'm intrigued, Adam. Why do you think this is fundamentally?

A Yeah, well, thanks. It's, it's great to know that, uh, people think I have knowledge. I must've been fooling them somehow. But, um, to your question, I think there's several reasons. I think that the first and most obvious one is that the underlying infrastructure can't support mass market dApps. So, CryptoKitties alone clogged up the Ethereum blockchain. If you think about it from sort of a higher level perspective, you The Visa network processes thousands of transactions per second, and that's only doing payments. How can you build apps on top of something that only can really do a couple dozen transactions per second? I think another reason is the lack of emphasis on user experience. So on the side of blockchain-based games, I don't think the barrier is really technology. I think it's a lack of interesting game mechanics. You could build a game like Fortnite in a centralized manner while having its cosmetic assets live on the blockchain. That's a very low And I think in other reasons that mass market consumers want to pay for DApp and for services in a unit of account that they understand. So this is why stable coins are extremely important. It's hard to have an e-commerce or e-commerce use case without a unit of account. So what this all stacks up to is that the biggest problem is that people are really excited about blockchain, and when they're building things, often tout …

AI assessment note: “the first and most obvious one is that the underlying infrastructure can't support mass market”

Q Speaking of kind of that rapid learning, different element, but kind of similar and aligned is your focus on kind of opt Optimizing for learning. Again, what does that kind of optimization around learning mean to you?

A Yeah, so throughout my life, I changed from math to computer science and engineering to product management to venture capital, and I did that so that I could learn as quickly as possible. I think that math is a great building block, and it was really only computer science that let me build things, and I think that product management lets you see sales and marketing engines get built, and customer support organizations, and you get to bring a customer perspective to engineering discussions. But I ultimately picked venture because it's a job that nobody can ever master. There will always be a new space or a paradigm shift to learn about. And, you know, that's sort of what we were talking about earlier, and that's why I gravitate towards founders who have, they operate in a growth mindset. They digest anything that they can. The ones who, who really can take any sort of beating from customers or from their employees with difficult situations, but iterate and come back and improve.

AI assessment note: “throughout my life, I changed from math to computer science... to learn as quickly as possible”

Q I am too intrigued that you mentioned they're obviously kind of funding themselves through the selling of their own tokens. We've also seen companies raise Traditional venture funding, and often from some pretty brilliant cryptologists. I'm intrigued. How do you think about the choice to take early venture funding versus the selling of your own tokens?

A Yeah, so I think that a lot of this question comes down to where value accrues, and value will usually accrue into either equity or a token network. It's very rarely both, but if it is both, it's probably like a 95%, five percent split. So we haven't released it yet, but we are in processing Of writing up a template term sheet for both fungible token and non fungible token projects at the formation stage. And the very high level of that is the easiest way to do the initial round is to have a price to equity round that has a modified pro rata clause that grants a proportional amount of the tokens held by the company. So if say there's 100% token supply overall, and you hold 50% for the company itself, For investors, for the founders, strategic deals, etc., and we own 10% of the company, we get 10% of that 50%. I think this aligns the two structures in a nice way and lets founders focus on the technology and not whether they should be selling tokens or equity.

AI assessment note: “a lot of this question comes down to where value accrues”

Q the tokens and what it allows in terms of funding. I do want to dig into one kind of real-world example, though, that we've seen most prominently in the news, and it's kind of often the case discussed Being Telegram, obviously an ICO that got an immense amount of investor appetite and attention. On that note, Adam, I have to ask, first, why do you think they got such attention?

A Yeah, so I mean, there's sort of a lazy answer, which is that they raised a lot of money, but the more interesting answer is that there's two things that are really important for building a blockchain DAT platform, and those two things are distribution and scaling. So we, we spoke about scaling earlier. Maybe we can focus on distribution here. Telegram has hundreds of millions of users. But most importantly, they don't have a perceived affiliation with a government. And I think in a world where owning your own data becomes more important to users by the day, they're extremely valuably positioned because they're the largest such platform that sort of meets that criteria. You think about Facebook, you think about WeChat, they're perceived differently. They're perceived as affiliated to a government, whether that's true or not. And I think the other really exciting thing is Telegram can build front ends for dApps within their platform. They can lower the bar and To entering the ecosystem while still maintaining a very, very powerful user experience. They can be the Dapp store, which is a very powerful jumping off point. They can have first party wallets. They can have first party exchanges and so on.

AI assessment note: “Telegram has hundreds of millions of users. But most importantly, they don't have a perceived affiliation”

Q Not at all, but I want to kick off today. Rarely do I get the pleasure of having someone almost as young as me on the show. So tell me, how did you make your way into the wonderful world of venture at the tender age of 22, Adam?

A Well, come on, that was almost two years ago. In essence, I was... Fortunate enough to be able to accelerate my education. So I was homeschooled until I was 10, and then I went to a community college for a few years because I really liked mathematics. So after that, I went to UC Berkeley at 13 to keep studying math. I finished at 16, but I didn't really feel ready to go out into the world. I didn't want to be a math professor or something like a quant. So I decided to go back to college and do undergrad again, this time at Stanford, and to study computer science. So, uh, big games are difficult for me, as you might imagine, but, uh, I always root for the home team. And along the way, I spent time at several research labs at Palantir, at Dropbox, and I was the first product manager at Rubrik, one of our portfolio companies in the enterprise infrastructure space. But the thing that I felt most fortunate about is that when I studied math, I focused on number theory and cryptography, and when I studied computer science, I focused on systems and distributed systems. So those two things marry very nicely together into blockchain. And although I didn't start investing professionally until a couple of years ago, I was involved in a first rounds dorm room fund when I was at Stanford. And I think that as someone with an unusual background, I have a lot of empathy for founders and realize…

AI assessment note: “I was involved in a first rounds dorm room fund when I was at Stanford.”

Q Before we dive into the quickfire, I do want to just quickly touch on the documents that really Present the quality or potentially lack of quality in a crypto opportunity being the mythical white paper. How do you evaluate the current state of white papers, Adam? And are they really an efficient presentation mode, do you think?

A So I think that white papers pay respects to the origins of the space, which were pretty academic, to be honest. I think if you look again at the traditional centralized world, people used to write business plans, and business plans eventually evolved into executive summaries and eventually evolved into decks. And I think that Something similar is already happening here. Decks are a very efficient way to communicate, and you still need to iron out all the details of what you're doing, and you, you should have a white paper, but you need to have a friendlier entry point for investors, employees, and partners alike, and I think that decks will differ from white papers and some other key aspects as well, so in the traditional world, you see pro formas. You see estimates of, of profit and loss over the coming years. I think that we will start to expect more in-depth Discussions of how you plan to fund your operations over time. I think that the days of saying we're going to reserve 30% of the tokens and that will fund our operations for perpetuity or over, I think it will look a lot more like traditional models. You'll say, we're raising this much in our initial token sale, and we are holding this amount of our security token, which will entitle us to this amount of cash flow. That might mean that we expect to have a 100,000 users with this sort of customer acquisition Some certain…

AI assessment note: “you should have a white paper, but you need to have a friendlier entry point”

Q Before we dive into the quickfire, I do want to just quickly touch on the documents that really Present the quality or potentially lack of quality in a crypto opportunity being the mythical white paper. How do you evaluate the current state of white papers, Adam? And are they really an efficient presentation mode, do you think?

A So I think that white papers pay respects to the origins of the space, which were pretty academic, to be honest. I think if you look again at the traditional centralized world, people used to write business plans, and business plans eventually evolved into executive summaries and eventually evolved into decks. And I think that Something similar is already happening here. Decks are a very efficient way to communicate, and you still need to iron out all the details of what you're doing, and you, you should have a white paper, but you need to have a friendlier entry point for investors, employees, and partners alike, and I think that decks will differ from white papers and some other key aspects as well, so in the traditional world, you see pro formas. You see estimates of, of profit and loss over the coming years. I think that we will start to expect more in-depth Discussions of how you plan to fund your operations over time. I think that the days of saying we're going to reserve 30% of the tokens and that will fund our operations for perpetuity or over, I think it will look a lot more like traditional models. You'll say, we're raising this much in our initial token sale, and we are holding this amount of our security token, which will entitle us to this amount of cash flow. That might mean that we expect to have a 100,000 users with this sort of customer acquisition Some certain…

AI assessment note: “Decks are a very efficient way to communicate, and you still need”

Q No, absolutely. I mean, I interrupted you with my rudeness there. I was too intrigued. In terms of the development of token economics, what do you think would make you more comfortable, Adam?

A Yeah, so I think that solving convexity in a meaningful way In the fungible world is really important. So every blockchain project that's building a network in some sense has two phases. The first phase is from inception to network launch. And the second phase is from network launch to sort of perpetuity. And I think that there aren't that many approaches that have been ironed out today that make phase two economically work. If you're holding 40% of your tokens and that progressively dwindles over time, at some point you're going to run out of money. The growth of the value of your token is not going to indefinitely meet your cashflow requirements. So one interesting approach that I've seen that's used by at least one of the projects I've invested in, although unfortunately it's one that hasn't announced, is to have two separate tokens. One that's a payment coin, just for the utility of the service itself, with the expectation that, 50 years out, the value of that token trends downward, but then separately have a security token that Siphons off a little bit of the cash flows from that token, and the separation of these two instruments lets you indeed have a reserve over time, because now you're capturing cash flows from your customers. Does that sort of make sense?

AI assessment note: “solving convexity in a meaningful way In the fungible world is really important”

Q Speaking of kind of that rapid learning, different element, but kind of similar and aligned is your focus on kind of opt Optimizing for learning. Again, what does that kind of optimization around learning mean to you?

A Yeah, so throughout my life, I changed from math to computer science and engineering to product management to venture capital, and I did that so that I could learn as quickly as possible. I think that math is a great building block, and it was really only computer science that let me build things, and I think that product management lets you see sales and marketing engines get built, and customer support organizations, and you get to bring a customer perspective to engineering discussions. But I ultimately picked venture because it's a job that nobody can ever master. There will always be a new space or a paradigm shift to learn about. And, you know, that's sort of what we were talking about earlier, and that's why I gravitate towards founders who have, they operate in a growth mindset. They digest anything that they can. The ones who, who really can take any sort of beating from customers or from their employees with difficult situations, but iterate and come back and improve.

AI assessment note: “I changed from math to computer science and engineering to product management to venture capital”

Q the tokens and what it allows in terms of funding. I do want to dig into one kind of real-world example, though, that we've seen most prominently in the news, and it's kind of often the case discussed Being Telegram, obviously an ICO that got an immense amount of investor appetite and attention. On that note, Adam, I have to ask, first, why do you think they got such attention?

A Yeah, so I mean, there's sort of a lazy answer, which is that they raised a lot of money, but the more interesting answer is that there's two things that are really important for building a blockchain DAT platform, and those two things are distribution and scaling. So we, we spoke about scaling earlier. Maybe we can focus on distribution here. Telegram has hundreds of millions of users. But most importantly, they don't have a perceived affiliation with a government. And I think in a world where owning your own data becomes more important to users by the day, they're extremely valuably positioned because they're the largest such platform that sort of meets that criteria. You think about Facebook, you think about WeChat, they're perceived differently. They're perceived as affiliated to a government, whether that's true or not. And I think the other really exciting thing is Telegram can build front ends for dApps within their platform. They can lower the bar and To entering the ecosystem while still maintaining a very, very powerful user experience. They can be the Dapp store, which is a very powerful jumping off point. They can have first party wallets. They can have first party exchanges and so on.

AI assessment note: “the more interesting answer is that there's two things that are really important”

Q I am too intrigued that you mentioned they're obviously kind of funding themselves through the selling of their own tokens. We've also seen companies raise Traditional venture funding, and often from some pretty brilliant cryptologists. I'm intrigued. How do you think about the choice to take early venture funding versus the selling of your own tokens?

A Yeah, so I think that a lot of this question comes down to where value accrues, and value will usually accrue into either equity or a token network. It's very rarely both, but if it is both, it's probably like a 95%, five percent split. So we haven't released it yet, but we are in processing Of writing up a template term sheet for both fungible token and non fungible token projects at the formation stage. And the very high level of that is the easiest way to do the initial round is to have a price to equity round that has a modified pro rata clause that grants a proportional amount of the tokens held by the company. So if say there's 100% token supply overall, and you hold 50% for the company itself, For investors, for the founders, strategic deals, etc., and we own 10% of the company, we get 10% of that 50%. I think this aligns the two structures in a nice way and lets founders focus on the technology and not whether they should be selling tokens or equity.

AI assessment note: “lets founders focus on the technology and not whether they should be selling tokens”

Q I couldn't agree with you more, especially from a white paper, but I do want to move, before we move into the quickfire, to discuss something that we've chatted about before, and it's that you said there are Four key trade-offs that are found and must contemplate in the world of crypto. Tell me, what are those key trade-offs in your mind, Adam?

A Yeah, and I think there's honestly probably way more than four, even. The ones that I was talking about are really because this is such a new world, founders can make choices that benefit themselves in the short term, but harm in the longer term. And like we talked about, there's sort of these, these two phases to a blockchain project that actually make these decisions pretty difficult. So one example is how you handle the fundraising docs and your regulatory strategy. I think that if you wait too long for the regulatory environment to settle, and you wait for the perfect set of docs, one of your competitors might suck the air out of the room with a huge fundraise. That's scary. At the same time, it's a little scary on the other side to have to push things along and maybe have a little bit extra regulatory risk just to get money in earlier. Another example is sort of how you set up your team. I think that a lot of teams are extremely marketing heavy in this world, and that's generally the wrong approach for long-term success, but that may generate a lot of hype in the short term. I generally bias towards engineering heavy teams because that's how you build a scalable protocol that has longevity. But if you think about it, the trade-off's not all that simple, right? If you look at an enterprise infrastructure company, in the early days, people hire a lot of engineers so they can…

AI assessment note: “One example is how you handle the fundraising docs and your regulatory strategy.”

Q I couldn't agree with you more, especially from a white paper, but I do want to move, before we move into the quickfire, to discuss something that we've chatted about before, and it's that you said there are Four key trade-offs that are found and must contemplate in the world of crypto. Tell me, what are those key trade-offs in your mind, Adam?

A Yeah, and I think there's honestly probably way more than four, even. The ones that I was talking about are really because this is such a new world, founders can make choices that benefit themselves in the short term, but harm in the longer term. And like we talked about, there's sort of these, these two phases to a blockchain project that actually make these decisions pretty difficult. So one example is how you handle the fundraising docs and your regulatory strategy. I think that if you wait too long for the regulatory environment to settle, and you wait for the perfect set of docs, one of your competitors might suck the air out of the room with a huge fundraise. That's scary. At the same time, it's a little scary on the other side to have to push things along and maybe have a little bit extra regulatory risk just to get money in earlier. Another example is sort of how you set up your team. I think that a lot of teams are extremely marketing heavy in this world, and that's generally the wrong approach for long-term success, but that may generate a lot of hype in the short term. I generally bias towards engineering heavy teams because that's how you build a scalable protocol that has longevity. But if you think about it, the trade-off's not all that simple, right? If you look at an enterprise infrastructure company, in the early days, people hire a lot of engineers so they can…

AI assessment note: “So one example is how you handle the fundraising docs and your regulatory strategy.”

Q The other element, though, that you did mention is the development or lack of development, so to speak, of the token economy. Tell me, Adam, what would you like to see happen in terms of token economics to make you maybe more comfortable?

A Yeah, I think what's really interesting about the blockchain world right now is we're in a place where projects are not only innovating on technology, but also their business model, and that's fundamentally pretty scary. But I'd say that one of the most interesting problems I've seen is the convexity one, which is that projects are most exposed to their own token in the beginning, but they gradually sell it over time to fund their operations. So if you think about it, you know, many projects keep 20 to 50% of their token, which are intended to fund the project in perpetuity. And that's really tough because if the project's successful, the value of the token goes up over time. And that means you want to sell later and not earlier. So I think some non-fungible token projects have overcome this in a pretty interesting way. The ones that are minted by companies actually could gain exposure over time. So if you're doing something like CryptoKitties or something similar that's IP-based, if your product's a success, 10 years down the line, you could be selling 10 times as much product rather than depleting a reserve that you have.

AI assessment note: “some non-fungible token projects have overcome this in a pretty interesting way”

Q No, absolutely. I mean, I interrupted you with my rudeness there. I was too intrigued. In terms of the development of token economics, what do you think would make you more comfortable, Adam?

A Yeah, so I think that solving convexity in a meaningful way In the fungible world is really important. So every blockchain project that's building a network in some sense has two phases. The first phase is from inception to network launch. And the second phase is from network launch to sort of perpetuity. And I think that there aren't that many approaches that have been ironed out today that make phase two economically work. If you're holding 40% of your tokens and that progressively dwindles over time, at some point you're going to run out of money. The growth of the value of your token is not going to indefinitely meet your cashflow requirements. So one interesting approach that I've seen that's used by at least one of the projects I've invested in, although unfortunately it's one that hasn't announced, is to have two separate tokens. One that's a payment coin, just for the utility of the service itself, with the expectation that, 50 years out, the value of that token trends downward, but then separately have a security token that Siphons off a little bit of the cash flows from that token, and the separation of these two instruments lets you indeed have a reserve over time, because now you're capturing cash flows from your customers. Does that sort of make sense?

AI assessment note: “one interesting approach that I've seen... is to have two separate tokens”

Q Adam, I know you well. I know that you wouldn't have looked at this half-heartedly. So on the other side of the coin, what were the reasons for your concern? There's a lot of skepticism around the opportunity also, as well as excitement. Where do you think there's kind of reason to dig deeper, so to speak?

A Yeah, so I, I think that they advertise some really aggressive timelines and, and really impressive technical goals. As with any project, you need to make sure that the team can execute on it. So if you look at the team here, Pavel Durov and Nikolai Durov, they've built two apps, VK, which is the most popular social network in, in Russia and the surrounding countries, and Telegram itself, both of which have hundreds of millions of users. That gives comfort that they can build something that scales, but also deliver a powerful user experience. Okay. And I think that it's also really exciting that they have an impressive engineering team that's complemented by many math and computing Olympiad winners. That being said, it's, it's really hard to fully validate a complicated technological approach to one of the most challenging problems known today.

AI assessment note: “really hard to fully validate a complicated technological approach”

Q No, I, I get that, and I love the alignment there to consume a brand. I'm intrigued in staying on the theme of trade-offs. Are there other trade-offs where you think make crypto opportunities, and certain crypto opportunities in particular, really uninvestable, be it security, be it the scalability element? What are those that make it uninvestable for you from the trade-off perspective?

A Yeah, I, I think that some of these trade-offs are really false dichotomies. I think something like security or scalability has I don't think if we were talking about a centralized application, you would give that sort of leeway. I would really, first and foremost, just give yourself a test. If you're building something on the blockchain, is it as interesting if you just use the word database instead of ledger or blockchain? And if you do, then you don't need to be building on the blockchain. It's sort of the same thing around how a lot of companies are built on top of AI. You could substitute out the word statistics for AI, and that's an interesting litmus test, too. So I think that there certainly are trade-offs, but at the extreme ends of these trade-offs, it really means you're using the wrong technology.

AI assessment note: “some of these trade-offs are really false dichotomies”

Q The other element, though, that you did mention is the development or lack of development, so to speak, of the token economy. Tell me, Adam, what would you like to see happen in terms of token economics to make you maybe more comfortable?

A Yeah, I think what's really interesting about the blockchain world right now is we're in a place where projects are not only innovating on technology, but also their business model, and that's fundamentally pretty scary. But I'd say that one of the most interesting problems I've seen is the convexity one, which is that projects are most exposed to their own token in the beginning, but they gradually sell it over time to fund their operations. So if you think about it, you know, many projects keep 20 to 50% of their token, which are intended to fund the project in perpetuity. And that's really tough because if the project's successful, the value of the token goes up over time. And that means you want to sell later and not earlier. So I think some non-fungible token projects have overcome this in a pretty interesting way. The ones that are minted by companies actually could gain exposure over time. So if you're doing something like CryptoKitties or something similar that's IP-based, if your product's a success, 10 years down the line, you could be selling 10 times as much product rather than depleting a reserve that you have.

AI assessment note: “rather than depleting a reserve that you have.”

Q No, I, I get that, and I love the alignment there to consume a brand. I'm intrigued in staying on the theme of trade-offs. Are there other trade-offs where you think make crypto opportunities, and certain crypto opportunities in particular, really uninvestable, be it security, be it the scalability element? What are those that make it uninvestable for you from the trade-off perspective?

A Yeah, I, I think that some of these trade-offs are really false dichotomies. I think something like security or scalability has I don't think if we were talking about a centralized application, you would give that sort of leeway. I would really, first and foremost, just give yourself a test. If you're building something on the blockchain, is it as interesting if you just use the word database instead of ledger or blockchain? And if you do, then you don't need to be building on the blockchain. It's sort of the same thing around how a lot of companies are built on top of AI. You could substitute out the word statistics for AI, and that's an interesting litmus test, too. So I think that there certainly are trade-offs, but at the extreme ends of these trade-offs, it really means you're using the wrong technology.

AI assessment note: “at the extreme ends of these trade-offs, it really means you're using the wrong technology.”

Q Adam, I know you well. I know that you wouldn't have looked at this half-heartedly. So on the other side of the coin, what were the reasons for your concern? There's a lot of skepticism around the opportunity also, as well as excitement. Where do you think there's kind of reason to dig deeper, so to speak?

A Yeah, so I, I think that they advertise some really aggressive timelines and, and really impressive technical goals. As with any project, you need to make sure that the team can execute on it. So if you look at the team here, Pavel Durov and Nikolai Durov, they've built two apps, VK, which is the most popular social network in, in Russia and the surrounding countries, and Telegram itself, both of which have hundreds of millions of users. That gives comfort that they can build something that scales, but also deliver a powerful user experience. Okay. And I think that it's also really exciting that they have an impressive engineering team that's complemented by many math and computing Olympiad winners. That being said, it's, it's really hard to fully validate a complicated technological approach to one of the most challenging problems known today.

AI assessment note: “it's really hard to fully validate a complicated technological approach”