The Exchanges

Q So many questions to dive on there. One, well, two, really, in particular. One is kind of stage agnosticity. So let's start with that one, actually. Walk me through the thesis, because you, you make investments at seed, series A, and even later stage. So how do you think about portfolio construction? Are they segregated to opportunity fund and initial fund for each stage? How do you think about that?

A So, so our opportunity fund is in fact contractually prohibited from chasing deals outside of our portfolio. So our opportunity fund is pure rocket fuel for things in our portfolio. That's number one. Number two, we voluntarily applied fairly strict limitations on valuation and performance metrics. Even in our own portfolio, we're not chasing deals Vaporous or Potemkin unicorns. Number three, there is an almost unitary LP position between our early funds and our opportunity funds, so there is de minimis, uh, conflict. Number four, from a portfolio formation perspective, we are investing as early as possible to achieve signal, to achieve Informational advantage. We can invest, obviously not in conflicting companies, but in companies in the same space for, uh, across almost all interesting companies in a space early for fewer dollars at risk than one beefy late A or early B investment by a larger fund, as those are traditionally done. And the ability to watch All of those teams in entire sector interact, pursue customers, develop their technology, uh, recruit, deal with incipient competition, adopt innovation elsewhere as it happens, uh, achieve product market fit, and so on and so on and so on gives us huge insights into not just which individual company is going to be a winner, but the entire sector. As opposed to sort of blindly dumping money into a sector at increasing expens…

AI assessment note: “our opportunity fund is in fact contractually prohibited from chasing deals outside of our portfolio”

Q So many questions to dive on there. One, well, two, really, in particular. One is kind of stage agnosticity. So let's start with that one, actually. Walk me through the thesis, because you, you make investments at seed, series A, and even later stage. So how do you think about portfolio construction? Are they segregated to opportunity fund and initial fund for each stage? How do you think about that?

A So, so our opportunity fund is in fact contractually prohibited from chasing deals outside of our portfolio. So our opportunity fund is pure rocket fuel for things in our portfolio. That's number one. Number two, we voluntarily applied fairly strict limitations on valuation and performance metrics. Even in our own portfolio, we're not chasing deals Vaporous or Potemkin unicorns. Number three, there is an almost unitary LP position between our early funds and our opportunity funds, so there is de minimis, uh, conflict. Number four, from a portfolio formation perspective, we are investing as early as possible to achieve signal, to achieve Informational advantage. We can invest, obviously not in conflicting companies, but in companies in the same space for, uh, across almost all interesting companies in a space early for fewer dollars at risk than one beefy late A or early B investment by a larger fund, as those are traditionally done. And the ability to watch All of those teams in entire sector interact, pursue customers, develop their technology, uh, recruit, deal with incipient competition, adopt innovation elsewhere as it happens, uh, achieve product market fit, and so on and so on and so on gives us huge insights into not just which individual company is going to be a winner, but the entire sector. As opposed to sort of blindly dumping money into a sector at increasing expens…

AI assessment note: “our opportunity fund is in fact contractually prohibited from chasing deals outside of our portfolio.”

Q Can I ask you, can I ask you a question? Is it not a case of kind of Friedman's perspective in terms of human wants and needs are unlimited, and we will just adapt to those wants and needs and find other ways to fulfill our time? Carr's Stopped horses and gave us huge amounts of efficiency increases. Will this not be the same with AI?

A Well, here's something that Tom Friedman and a handful of techno-utopian folks in the venture industry ignore. That's the issue of latency or hysteresis. So the destruction of agricultural jobs in the United States happened over 30 or 40 or 50 50 years. It's a very long time for people to be retrained to move between states for greater economic opportunity. In fact, you could almost argue that it was generational, that, you know, Ma and Pa Kettle, you know, died farming, but their children had the opportunity to go to college or trade school or were sort of sucked into the vast middle-class communities Creation engine that was World War II in the United States got a GI bill or went to work for, for one of the dominant US companies, uh, that came out of that. But the pace at which AI today can disrupt entire industries, if not even entire national economies, is measured in single digit years, low double digit at best. As I said to a colleague, even in the 19 Seventies, a suddenly unemployed welder who specialized in a particular aspect of somebody in a steel plant who specialized in a process that literally no longer had an industrial footprint in the United States by 1983 did not all of a sudden turn themselves into a IBM three 70 cobalt programmer and go get a 100,000 dollar a year job. You have to have the economic stability and the economic resource To retrain yourself, and …

AI assessment note: “That's the issue of latency or hysteresis.”

Q Can I ask you, can I ask you a question? Is it not a case of kind of Friedman's perspective in terms of human wants and needs are unlimited, and we will just adapt to those wants and needs and find other ways to fulfill our time? Carr's Stopped horses and gave us huge amounts of efficiency increases. Will this not be the same with AI?

A Well, here's something that Tom Friedman and a handful of techno-utopian folks in the venture industry ignore. That's the issue of latency or hysteresis. So the destruction of agricultural jobs in the United States happened over 30 or 40 or 50 50 years. It's a very long time for people to be retrained to move between states for greater economic opportunity. In fact, you could almost argue that it was generational, that, you know, Ma and Pa Kettle, you know, died farming, but their children had the opportunity to go to college or trade school or were sort of sucked into the vast middle-class communities Creation engine that was World War II in the United States got a GI bill or went to work for, for one of the dominant US companies, uh, that came out of that. But the pace at which AI today can disrupt entire industries, if not even entire national economies, is measured in single digit years, low double digit at best. As I said to a colleague, even in the 19 Seventies, a suddenly unemployed welder who specialized in a particular aspect of somebody in a steel plant who specialized in a process that literally no longer had an industrial footprint in the United States by 1983 did not all of a sudden turn themselves into a IBM three 70 cobalt programmer and go get a 100,000 dollar a year job. You have to have the economic stability and the economic resource To retrain yourself, and …

AI assessment note: “here's something that Tom Friedman and a handful of techno-utopian folks in the venture industry ignore”

Q of increased timescale in terms of availability of AI to replace human resources, Simple human functions, as we said, truck driving jobs, I'm sure assistant roles. Um, so, so what is the future of work with such potential mass scale unemployment? How does that look to you? Uh, don't worry about being overly negative or positive. What are your thoughts on kind of the future of work in that respect?

A Well, so we are trying consciously as a firm to invest in things that augment human intelligence and augment Human capability. So that even as people may become unemployed, the people who are creative or have an interesting idea may find themselves lifted up by even this simpler AI so that the, you know, 10,000 hours required today to become an expert in something can be vastly reduced. I, I know, you know, uh, Samil, uh, Shah, he has a portfolio company that uses AI in combination with Heretofore unavailable because of both expense and size precision motors that can turn even a complete klutz like me into something approximating a master cabinet maker. It's a set of hand tools that are self-guided across a hugely complex set of operations on wood and, and, and metal and other things that go into, um, cabinetry and other piecework like that. So that's the kind of stuff that makes me hopeful. On the other hand, if that is too expensive or is somehow restricted, uh, in its sale or use because of trade union rules or, uh, or other political black swan emergence of restrictions, then you're going to have a lot of people out of work and their route to reemployment and fulfilling lives may be very grim.

AI assessment note: “we are trying consciously as a firm to invest in things that augment human intelligence”

Q You said there about early stage investors, um, being predicated towards kind of favoring, uh, business models and startups, which they know they can easily pass through to follow on funding. How do you extricate yourself from that kind of funnel process thinking and think outside of the box in terms of your investing thesis then?

A Well, you, you raise this question later and I don't want to steal your thunder, but that's one of the reasons That for many years now, we have had an opportunity fund. When we see a young company showing exponential progress, when we believe that it could be a potential monopoly or member of an oligopoly and have appropriate impact and returns that goes along there with, we want to have the freedom to back that company. It's also the reason that We made a very conscious decision from the very beginning of our first institutional fund not to be a micro VC, not to be in the 30 to fifty million dollar fund size category, because if you disagree with conventional wisdom about a series A and you want to write two or three or five or even seven million dollars in a check yourselves through a company that you believe is extraordinary, you need to have the capacity to do that. Otherwise, you are perpetually the handmaiden to the enthusiasms, or lack thereof, of the folks upstream.

AI assessment note: “Otherwise, you are perpetually the handmaiden to the enthusiasms, or lack thereof”

Q then to say, you mentioned obviously capital efficiency and increasing life cycles, uh, of investments there. So, so I'm intrigued as to your thoughts on the life cycle of current venture. You've said before, probably contrarian thinking again, that you think it's potentially very wrong, uh, the current fund life cycles. So what are your thoughts on this and the kind of inherent inefficiencies that currently persist in the industry?

A We gently tease our limited partners. They, they are by and large, extremely patient, extremely farsighted people whose trust in us and whose generosity we are incredibly grateful for. But we tease them occasionally when, uh, as, as Chris Duvos, uh, one of our LPs says, uh, where's the mula in the kula about mild cognitive dissonance between LP and Chest beating over 20, 40, 7100 year time horizons, and also being anxious that they haven't gotten all of their money back inside three years. And again, our, our LPs are actually very, very good about this, but industry-wide, that sort of mental conflict is a real one. The fact is, in deep tech, in sustainable franchises, not consumer companies, Uh, not that consumer companies are bad, and we all may still be using Facebook in a hundred years, but companies with 40, 5100 year durability, whether it's a GE or an Intel or a, um, uh, Qualcomm, which seems to be marching towards that, Cisco and Microsoft, uh, which are getting very close to, uh, uh, to hitting those kind of durability milestones many, many decades. And tens of billions of dollars of value, those aren't fully realized inside a ten-year fund life, 12 at best with, um, uh, two petitions for one-year extensions, uh, by the manager to LPs. And with companies, especially the kind of deep tech companies and companies in general, whether it's, uh, Uber or Airbnb, running out t…

AI assessment note: “those aren't fully realized inside a ten-year fund life”

Q I absolutely love that, but today we're going to deep dive on all things AI, so I want to start on a seriously meta topic. So Matt, what's the fundamental problem of AI?

A Well, one of the biggest problems is that we don't know what we don't know, uh, in a sort of conflation of a quote from the great physicist John Archibald Wheeler and the equally great astrophysicist Carl Sagan, we're sort of Children playing on a beach on an island of ignorance surrounded by a vast sea. Uh, we really are relatively ill-equipped to understand how much knowledge we need to explore that sea, and to Wheeler's point, the more, uh, the more we explore, the greater the boundaries of our ignorance actually become. One concrete example of this is that for every wonderful AI breakthrough Whether it's embodied in hardware, like self-driving trucks, or whether it's embodied in software that goes into the bowels of government or corporations, there are huge potential unintended side effects. So with self-driving trucks, they're, uh, you're talking about potentially vastly reducing, if not completely disemploying, I guess to coin a term, Three or four million of the last non-college educated middle class people in the United States, and quite a large similar number in Europe. So yes, you vastly reduce highway fatalities, pollution, because many of these trucks could be electric, they could drive more slowly, they could auto plug into refueling stations, so on and so on and so forth. But you also just made three million people surplus and caused Great anxiety and economic ha…

AI assessment note: “one of the biggest problems is that we don't know what we don't know”

Q I mean, do you think it's actually possible for founders who are all in on a startup to focus on getting eight hours sleep a night? I've never met a founder who could do that.

A There are a handful of founders who Who are not only superhumanly brilliant, so they process and execute on information an order or more of magnitude faster than you and I. Uh, I'm happy to have a few of these founders in our portfolio. They also tend to be supremely, uh, disciplined people. I think of them as kind of, uh, early examples of, you know, homo superior from a, uh, from a science. And, and a lot of them, a lot of People in that tiny class do wonders and still get seven or eight hours of sleep. Some of the really scary ones are short sleepers and are that smart and that capable and that relentless, and they sleep only four hours a night, which is really, really scary. I hope they give me a job as a pool boy when they take over the world. Or ordinary, ordinary humans like us can't respond to the demands of Of a truly successful startup, one that has this hard, sharp arc of takeoff and sleep eight hours and cook our own organic breakfast and get in a brisk six mile walk and then write a little bit of poetry for the significant other before ambling off to work.

AI assessment note: “a lot of People in that tiny class do wonders and still get seven or eight”

Q I absolutely love that, but today we're going to deep dive on all things AI, so I want to start on a seriously meta topic. So Matt, what's the fundamental problem of AI?

A Well, one of the biggest problems is that we don't know what we don't know, uh, in a sort of conflation of a quote from the great physicist John Archibald Wheeler and the equally great astrophysicist Carl Sagan, we're sort of Children playing on a beach on an island of ignorance surrounded by a vast sea. Uh, we really are relatively ill-equipped to understand how much knowledge we need to explore that sea, and to Wheeler's point, the more, uh, the more we explore, the greater the boundaries of our ignorance actually become. One concrete example of this is that for every wonderful AI breakthrough Whether it's embodied in hardware, like self-driving trucks, or whether it's embodied in software that goes into the bowels of government or corporations, there are huge potential unintended side effects. So with self-driving trucks, they're, uh, you're talking about potentially vastly reducing, if not completely disemploying, I guess to coin a term, Three or four million of the last non-college educated middle class people in the United States, and quite a large similar number in Europe. So yes, you vastly reduce highway fatalities, pollution, because many of these trucks could be electric, they could drive more slowly, they could auto plug into refueling stations, so on and so on and so forth. But you also just made three million people surplus and caused Great anxiety and economic ha…

AI assessment note: “one of the biggest problems is that we don't know what we don't know”

Q you have some thoughts on and that is fun sizes. Um, and before you said that all fun size models are wrong, With kind of current bifurcation of VC into angels, micro VCs, small C to A funds, et cetera, et cetera. So why do you think this is suboptimal? First from a returns perspective, and then let's do a societal. So why is this suboptimal from a return perspective?

A I sincerely believe, and there's a large but quiet group of other experienced VCs with successful track records that Uh, who also believe that the strict segregation of funds by size has an information destroying or signal reducing aspect. Today, the earliest investors in the company, whether they're the, the micro VC or small seed or, or a fund who know the most about the company have to basically hold a blind auction at each node or locus of Capital formation as the company moves from stage to stage. What that means is that the later stage funds, because the early investor has overwhelming need, along with the founders of the company, to complete the round to essentially hold as fully populated and as aggressive an auction as possible. The later stage investors are now competing with each other on as short a time frame As the early investor, uh, can possibly make happen, which means their ability to do complete diligence and to have the most informed possible judgment is reduced. So that puts LP money in a later stage fund at risk. In a hot environment, they're competing on the order of weeks. In a slow environment, for a credible deal, it's still a handful of months at the very most. Meanwhile, the early stage investor Because they have to consider this capital formation risk is at best unconsciously and at worst very consciously biasing the nature of the investments that th…

AI assessment note: “strict segregation of funds by size has an information destroying or signal reducing aspect”

Q You said there about getting in really early and the signaling that that entails and allows you to see. How do you avoid, then, the negative signaling function of not Following on with the availability of the opportunity fund. How does that work?

A Well, first of all, I don't think that our peers expect us to always follow on, uh, from the opportunity fund because sometimes companies do well enough that they don't need money from the opportunity fund, or sometimes they're growing at a pace where, uh, traditional financing is sufficient. Uh, we'll get into this in terms of the life cycle. These, uh, deep tech companies are Across sort of the horizon of the vintage. And sometimes folks show up at our doorstep. We're friendly with almost everybody and co-invest quite happily up and down Sand Hill Road and say, look, XYZ company crossed our threshold. We've done some work. We really want to be the folks to write a 20 or 25 or thirty million dollar check. And since we've already achieved Reasonable ownership upfront. There's no economic urgency to extract additional equity from the entrepreneur out of the opportunity fund just for the sake of doing so. So I don't think the opportunity fund carries any signaling risk. A fair question is the obverse, which is, is there signaling risk when you make a seed investment and then you don't follow on? I would say in our case, that is relatively minor. We are brutally frank with all parties, including the entrepreneurs and the folks upstream about how we make an investment decision. Also, when we're not interested in following on, we're extremely transparent. And I think people understa…

AI assessment note: “I don't think the opportunity fund carries any signaling risk.”

Q you have some thoughts on and that is fun sizes. Um, and before you said that all fun size models are wrong, With kind of current bifurcation of VC into angels, micro VCs, small C to A funds, et cetera, et cetera. So why do you think this is suboptimal? First from a returns perspective, and then let's do a societal. So why is this suboptimal from a return perspective?

A I sincerely believe, and there's a large but quiet group of other experienced VCs with successful track records that Uh, who also believe that the strict segregation of funds by size has an information destroying or signal reducing aspect. Today, the earliest investors in the company, whether they're the, the micro VC or small seed or, or a fund who know the most about the company have to basically hold a blind auction at each node or locus of Capital formation as the company moves from stage to stage. What that means is that the later stage funds, because the early investor has overwhelming need, along with the founders of the company, to complete the round to essentially hold as fully populated and as aggressive an auction as possible. The later stage investors are now competing with each other on as short a time frame As the early investor, uh, can possibly make happen, which means their ability to do complete diligence and to have the most informed possible judgment is reduced. So that puts LP money in a later stage fund at risk. In a hot environment, they're competing on the order of weeks. In a slow environment, for a credible deal, it's still a handful of months at the very most. Meanwhile, the early stage investor Because they have to consider this capital formation risk is at best unconsciously and at worst very consciously biasing the nature of the investments that th…

AI assessment note: “strict segregation of funds by size has an information destroying or signal reducing aspect.”

Q I mean, do you think it's actually possible for founders who are all in on a startup to focus on getting eight hours sleep a night? I've never met a founder who could do that.

A There are a handful of founders who are not only superhumanly brilliant, so they process and execute on information an order or more of magnitude faster than you and I. Uh, I'm happy to have a few of these founders in our portfolio. They also tend to be supremely Uh, disciplined people. I think of them as kind of, uh, early examples of, you know, homo superior from a, uh, from a science. And, and a lot of them, a lot of people in that tiny class do wonders and still get seven or eight hours of sleep. Some of the really scary ones are short sleepers and are that smart and that capable and that relentless. And they sleep only four hours a night. Uh, which is really, really scary. I hope they give me a job as a pool boy when they take over the world. Or ordinary, ordinary humans like us can't respond to the demands of a truly successful startup. One that has this hard, sharp arc of takeoff and sleep eight hours and cook our own organic breakfast and get in a brisk Six mile walk and then write a little bit of poetry for the significant other before ambling off to work.

AI assessment note: “a lot of people in that tiny class do wonders and still get seven or eight”

Q You said there about early stage investors, um, being predicated towards kind of favoring, uh, business models and startups, which they know they can easily pass through to follow on funding. How do you extricate yourself from that kind of funnel process thinking and think outside of the box in terms of your investing thesis then?

A Well, you, you raise this question later and I don't want to steal your thunder, but that's one of the reasons That for many years now, we have had an opportunity fund. When we see a young company showing exponential progress, when we believe that it could be a potential monopoly or member of an oligopoly and have appropriate impact and returns that goes along there with, we want to have the freedom to back that company. It's also the reason that We made a very conscious decision from the very beginning of our first institutional fund not to be a micro VC, not to be in the 30 to fifty million dollar fund size category, because if you disagree with conventional wisdom about a series A and you want to write two or three or five or even seven million dollars in a check yourselves through a company that you believe is extraordinary, you need to have the capacity to do that. Otherwise, you are perpetually the handmaiden to the enthusiasms, or lack thereof, of the folks upstream.

AI assessment note: “we made a very conscious decision from the very beginning of our first institutional fund”

Q You said there about getting in really early and the signaling that that entails and allows you to see. How do you avoid, then, the negative signaling function of not Following on with the availability of the opportunity fund. How does that work?

A Well, first of all, I don't think that our peers expect us to always follow on, uh, from the opportunity fund because sometimes companies do well enough that they don't need money from the opportunity fund, or sometimes they're growing at a pace where, uh, traditional financing is sufficient. Uh, we'll get into this in terms of the life cycle. These, uh, deep tech companies are Across sort of the horizon of the vintage. And sometimes folks show up at our doorstep. We're friendly with almost everybody and co-invest quite happily up and down Sand Hill Road and say, look, XYZ company crossed our threshold. We've done some work. We really want to be the folks to write a 20 or 25 or thirty million dollar check. And since we've already achieved Reasonable ownership upfront. There's no economic urgency to extract additional equity from the entrepreneur out of the opportunity fund just for the sake of doing so. So I don't think the opportunity fund carries any signaling risk. A fair question is the obverse, which is, is there signaling risk when you make a seed investment and then you don't follow on? I would say in our case, that is relatively minor. We are brutally frank with all parties, including the entrepreneurs and the folks upstream about how we make an investment decision. Also, when we're not interested in following on, we're extremely transparent. And I think people understa…

AI assessment note: “So I don't think the opportunity fund carries any signaling risk.”

Q There's studies that say kind of a AGI is actually 2050. You know, it's kind of 35 years realistically out. Do you think that's a fair assumption?

A Well, first I'll say you don't need AGI to unemploy 30% of all Pink collar and white collar workers in the G-Twenty within the next 10 years. Presuming that it has to be AGI, and therefore we have some kind of buffer zone, is I think a dangerous assumption. Uh, the next thing is, to my point about being kids playing on the beach, while I am relatively sanguine that AGI that is so super intelligent that it views us not malevolently, but indifferently, the way we may or may not Kill a gnat that is fluttering around our monitor on a late summer night. You know, that, that might be many, many decades, but near parity for machine vision with human beings was a matter of, you know, a few hundred lines of code that one brilliant person had an insight into. And as low level AI becomes better, that human insight into pushing the boundaries of AI Begins to become exponentially more productive. The assistant AI for discovering algorithm spaces becomes exponentially better, and you might have something that, even if it's not full AGI, is weaponized semi-AGI. And when I say weaponized, I don't mean the Skynet of the Terminator films. I mean something more like what happened to Lee Sedol facing DeepMind AlphaGo. In many of the matches, DeepMind had Already beaten Lisa doll, but he and a panel of brilliant human observers had no idea that they were already beat in the moment of the game. Deep…

AI assessment note: “that might be many, many decades, but near parity for machine vision”

Q There's studies that say kind of a AGI is actually 2050. You know, it's kind of 35 years realistically out. Do you think that's a fair assumption?

A Well, first I'll say you don't need AGI to unemploy 30% of all Pink collar and white collar workers in the G-Twenty within the next 10 years. Presuming that it has to be AGI, and therefore we have some kind of buffer zone, is I think a dangerous assumption. Uh, the next thing is, to my point about being kids playing on the beach, while I am relatively sanguine that AGI that is so super intelligent that it views us not malevolently, but indifferently, the way we may or may not Kill a gnat that is fluttering around our monitor on a late summer night. You know, that, that might be many, many decades, but near parity for machine vision with human beings was a matter of, you know, a few hundred lines of code that one brilliant person had an insight into. And as low level AI becomes better, that human insight into pushing the boundaries of AI Begins to become exponentially more productive. The assistant AI for discovering algorithm spaces becomes exponentially better, and you might have something that, even if it's not full AGI, is weaponized semi-AGI. And when I say weaponized, I don't mean the Skynet of the Terminator films. I mean something more like what happened to Lee Sedol facing DeepMind AlphaGo. In many of the matches, DeepMind had Already beaten Lisa doll, but he and a panel of brilliant human observers had no idea that they were already beat in the moment of the game. Deep…

AI assessment note: “Presuming that it has to be AGI... is I think a dangerous assumption.”

Q No, I mean, I mean, that, that, that itself is very interesting in terms of how important you feel thesis driven venture firms are. Uh, so, so let's do that a quick one there before we dive into fun cycles. Um, what, how important do you think thesis driven firms are? Do you think they need to be thesis driven now instead of stage driven?

A We're a little unusual in that We have a meta thesis through which we, we view the world, but to your, uh, to your point, we don't have an industry or sector thesis. Our meta thesis, the thing that we want to see in any company is that they are exemplifying the use of compute and novel, highly defensible algorithms to themselves be capital efficient, Whether they involve hardware or not, so that they have longer runway, more bites on the apple, more control of their own destiny, and for their customers, are capital efficient, so they deliver higher ROI, they have more control over margin, they can survive downturns that afflict every sector, and that they are doing so in a way that addresses, ideally, an unloved contrarian market Certainly a massive one. We've done stuff in computational biology, in agricultural genomics and informatics and robotic automation, and also in fraud and risk reduction for some of the largest financial institutions in the world, as well as a company, uh, Planet Labs is delivering massive informatics Advantage to its customers by virtue of having become the second or third largest space power in the world measured in satellites. We're agnostic about sector, but how the company gets there, how they defend their edge, what their internal economics are, are incredibly important to us. We regretfully either say no up front or don't follow on in the kind o…

AI assessment note: “We have a meta thesis through which we, we view the world”

Q I mean, do you think it's actually possible for founders who are all in on a startup to focus on getting eight hours sleep a night? I've never met a founder who could do that.

A There are a handful of founders who Who are not only superhumanly brilliant, so they process and execute on information an order or more of magnitude faster than you and I. Uh, I'm happy to have a few of these founders in our portfolio. They also tend to be supremely, uh, disciplined people. I think of them as kind of, uh, early examples of, you know, homo superior from a, uh, from a science. And, and a lot of them, a lot of People in that tiny class do wonders and still get seven or eight hours of sleep. Some of the really scary ones are short sleepers and are that smart and that capable and that relentless, and they sleep only four hours a night, which is really, really scary. I hope they give me a job as a pool boy when they take over the world. Or ordinary, ordinary humans like us can't respond to the demands of Of a truly successful startup, one that has this hard, sharp arc of takeoff and sleep eight hours and cook our own organic breakfast and get in a brisk six mile walk and then write a little bit of poetry for the significant other before ambling off to work.

AI assessment note: “a lot of People in that tiny class do wonders and still get seven or eight”

Q of increased timescale in terms of availability of AI to replace human resources, Simple human functions, as we said, truck driving jobs, I'm sure assistant roles. Um, so, so what is the future of work with such potential mass scale unemployment? How does that look to you? Uh, don't worry about being overly negative or positive. What are your thoughts on kind of the future of work in that respect?

A Well, so we are trying consciously as a firm to invest in things that augment human intelligence and augment Human capability. So that even as people may become unemployed, the people who are creative or have an interesting idea may find themselves lifted up by even this simpler AI so that the, you know, 10,000 hours required today to become an expert in something can be vastly reduced. I, I know, you know, uh, Samil, uh, Shah, he has a portfolio company that uses AI in combination with Heretofore unavailable because of both expense and size precision motors that can turn even a complete klutz like me into something approximating a master cabinet maker. It's a set of hand tools that are self-guided across a hugely complex set of operations on wood and, and, and metal and other things that go into, um, cabinetry and other piecework like that. So that's the kind of stuff that makes me hopeful. On the other hand, if that is too expensive or is somehow restricted, uh, in its sale or use because of trade union rules or, uh, or other political black swan emergence of restrictions, then you're going to have a lot of people out of work and their route to reemployment and fulfilling lives may be very grim.

AI assessment note: “their route to reemployment and fulfilling lives may be very grim”

Q I mean, do you think it's actually possible for founders who are all in on a startup to focus on getting eight hours sleep a night? I've never met a founder who could do that.

A There are a handful of founders who are not only superhumanly brilliant, so they process and execute on information an order or more of magnitude faster than you and I. Uh, I'm happy to have a few of these founders in our portfolio. They also tend to be supremely Uh, disciplined people. I think of them as kind of, uh, early examples of, you know, homo superior from a, uh, from a science. And, and a lot of them, a lot of people in that tiny class do wonders and still get seven or eight hours of sleep. Some of the really scary ones are short sleepers and are that smart and that capable and that relentless. And they sleep only four hours a night. Uh, which is really, really scary. I hope they give me a job as a pool boy when they take over the world. Or ordinary, ordinary humans like us can't respond to the demands of a truly successful startup. One that has this hard, sharp arc of takeoff and sleep eight hours and cook our own organic breakfast and get in a brisk Six mile walk and then write a little bit of poetry for the significant other before ambling off to work.

AI assessment note: “a lot of people in that tiny class do wonders and still get seven or eight hours”

Q No, I mean, I mean, that, that, that itself is very interesting in terms of how important you feel thesis driven venture firms are. Uh, so, so let's do that a quick one there before we dive into fun cycles. Um, what, how important do you think thesis driven firms are? Do you think they need to be thesis driven now instead of stage driven?

A We're a little unusual in that We have a meta thesis through which we, we view the world, but to your, uh, to your point, we don't have an industry or sector thesis. Our meta thesis, the thing that we want to see in any company is that they are exemplifying the use of compute and novel, highly defensible algorithms to themselves be capital efficient, Whether they involve hardware or not, so that they have longer runway, more bites on the apple, more control of their own destiny, and for their customers, are capital efficient, so they deliver higher ROI, they have more control over margin, they can survive downturns that afflict every sector, and that they are doing so in a way that addresses, ideally, an unloved contrarian market Certainly a massive one. We've done stuff in computational biology, in agricultural genomics and informatics and robotic automation, and also in fraud and risk reduction for some of the largest financial institutions in the world, as well as a company, uh, Planet Labs is delivering massive informatics Advantage to its customers by virtue of having become the second or third largest space power in the world measured in satellites. We're agnostic about sector, but how the company gets there, how they defend their edge, what their internal economics are, are incredibly important to us. We regretfully either say no up front or don't follow on in the kind o…

AI assessment note: “we don't have an industry or sector thesis. Our meta thesis”

Q then to say, you mentioned obviously capital efficiency and increasing life cycles, uh, of investments there. So, so I'm intrigued as to your thoughts on the life cycle of current venture. You've said before, probably contrarian thinking again, that you think it's potentially very wrong, uh, the current fund life cycles. So what are your thoughts on this and the kind of inherent inefficiencies that currently persist in the industry?

A We gently tease our limited partners. They, they are by and large, extremely patient, extremely farsighted people whose trust in us and whose generosity we are incredibly grateful for. But we tease them occasionally when, uh, as, as Chris Duvos, uh, one of our LPs says, uh, where's the mula in the kula about mild cognitive dissonance between LP and Chest beating over 20, 40, 7100 year time horizons, and also being anxious that they haven't gotten all of their money back inside three years. And again, our, our LPs are actually very, very good about this, but industry-wide, that sort of mental conflict is a real one. The fact is, in deep tech, in sustainable franchises, not consumer companies, Uh, not that consumer companies are bad, and we all may still be using Facebook in a hundred years, but companies with 40, 5100 year durability, whether it's a GE or an Intel or a, um, uh, Qualcomm, which seems to be marching towards that, Cisco and Microsoft, uh, which are getting very close to, uh, uh, to hitting those kind of durability milestones many, many decades. And tens of billions of dollars of value, those aren't fully realized inside a ten-year fund life, 12 at best with, um, uh, two petitions for one-year extensions, uh, by the manager to LPs. And with companies, especially the kind of deep tech companies and companies in general, whether it's, uh, Uber or Airbnb, running out t…

AI assessment note: “realization of maximum value from a successful investment and traditional fund lifetimes are beginning to invert”