Pat Grady - Relentless Application of Force - [Invest Like the Best, EP.378]

25:10But then when Doug was done, he would say, OK, I think this is what I heard. and then he would talk for about two minutes. And it was the most beautiful thing you have ever heard where he would perfectly capture not only what the founder was hoping to build, but everything that was wrong with the business today, which the founder may not have been so straightforward about, but Doug teased out with all of his questions. He would acknowledge those flaws. He would gently suggest how we could help with those flaws. and then he would reaffirm the vision of the future that this founder hoped to build with the company and this founder hoped to build for him or herself.

25:54And that was the most potent elixir you could possibly imagine. And maybe not 100 % of the time, but 95 % of the time after Doug unfurled one of those, the founder would be desperate to work with us. And then all we had to do is decide whether or not we wanted to. So that was a pretty magical experience. What's your version of that? Now you're the Doug. What do you do in those meetings? Obviously, everyone's style is different. How have you honed it and shaped it in your own way? Well, first off, I will never be the Doug. There is only one thing. It is not possible for there to be another Doug.

26:29One other trick that I had when I was in my 20s and early 30s and having witnessed Doug do that so many times when I would go into these meetings and that was 2007 when I joined here was the early days of the cloud transition. most of the founders that I was meeting with, they'd been in the enterprise for 20 years. And so almost every meeting I went into, the founder was meaningfully older than I was. I looked at a kid who happened to have a very nice business card, but otherwise no experience that could be relevant to them. And so I had to figure out how to get credibility. And the way that I would usually do that, mirroring what I'd learned from Doug was to start the meeting by saying something like, hey, I don't know much about your market.

27:13but it seems like you have a chance to do blah, blah, blah. And the blah, blah, blah would be whatever thesis I might've come up with based on looking at their website or talking to some of their customers or studying some of their competitors or whatever the case might be. And that was usually enough to get the benefit of the doubt and to get them thinking, okay, this guy kind of understands who I am and what I'm trying to build here. And I would say that If you fast forward to today, that's still the most powerful thing you can do is to make a founder feel like they're seen. You understand them and try to validate their ambition through the way that you describe their business.

27:52I think that's still the most powerful thing that we can do. And then beyond that, that might get you a nice in-person interaction with the founder. But then ultimately, we are in a product-led growth business. Our product is the service that we provide to our founders. and when our founders tell other founders what they think of us, that's ultimately the thing that gets them over the line. And so our objective function is to maximize net multiple money returns for our limited partners, not to maximize founder NPS. But if we can have our cake and eat it too, that's the best of both worlds. And so founder NPS and net multiple money returns are the two metrics that we probably care most about.

28:29Thinking about the emails that you would write reminds me that a bunch of people have mentioned to me how good you are at memo writing. And I know memos are taken very seriously at Sequoia, but I'd love to hear for you specifically how you define great business or investing writing, what you are seeking to do in those memos and what you respect most when you see it in other similar memos. This comes back to that expression, listening happens at the year. I think there are a lot of memos that get written, speeches that get given where the objective is to impress upon other people how smart you are.

29:05And I think when you write a memo or you give a talk or whatever, you should do so with the service mindset. My objective is not to impress upon anybody how smart I am. My objective is for them to understand. And if I actually want them to understand, I need to make it as simple as humanly possible. And if that means sacrificing a few details for the sake of clarity, not in a way that distorts the picture, but in a way that clarifies the picture because people aren't going down rabbit holes that don't represent first order issues, that's probably worth it. And so a great investment memo is three pages, not 12 pages.

29:38And when you're done with those three pages, you should have an accurate point of view on everything that is good, everything that is bad, and the so what. When we mash that soup all together, should we make this investment or not and why? And I think you can do that in three pages generally. Take a bunch of the exhibits, the cohort charts and everything else. Throw those in the appendix. You can express that in one sentence. You don't need to have everybody dig through all the data on their own. Go ahead and throw that in the appendix in case anybody wants to go back there and play with it. But keep the narrative up front.

30:09Keep it tight. Keep it crisp. And make sure that the thing you're optimizing for is clarity and understanding, not trying to impress upon people how smart you are. Do you remember reading a memo, not one you wrote, but another one, about a company that got you the most physically excited? 100%. Yes. The first one that comes to mind is in 2009, a few weeks before the market bottomed out, we got into business with Airbnb at the seed stage. And that seemed like a crazy idea, but it was working and we fell in love with Brian and Joe and Nate. And anyway, so we were lucky enough to be in business with Airbnb at the seed stage.

30:49Fast forward to 2012, the company was clearly working. Even as existing shareholders and board members, we weren't sure exactly what to make of it. And there was a growth realm coming together. So the first memo comes out. And it's a typical investment memo. It talks about the market and it talks about the numbers and it talks about the team and it talks about the competitors and all that good stuff. And there was a model at the end of it that did your typical simple linear extrapolation with decelerating growth rates and maybe some flat margins. And it got you a two and a half X return. And so you read that and you sort of yawn.

31:31Okay, two and a half X, who cares? And so we decided on that Monday not to move forward, but we're interested. So we decided to do some more work. The following Friday, a second memo comes out. page one of the memo to become a$100 billion company. And here is why. And it laid out with perfect clarity why they had a chance, not just to be a two and a half X multiple of money, but to return the fund a couple of times over because of the size of the market, because of the structural superiority of the business model, because of the creativity and passion and clarity and mission orientation of the founders.

32:14And it takes courage to do that, because most of the time when you look at a business that was probably grossly overvalued by any traditional metric at even a$2 billion entry price, and say that it will someday be a $100 billion business, you get laughed out of the room. But it turns out those are the only investments that actually matter. And if you don't have the conviction that the company has a chance to be something truly special, you shouldn't be recommending that we invest. You send me these really fascinating criteria for making an investment at the growth stage, one of which is, I can't remember the statistical term, leptocytosis or something like that.

32:52Some very fancy sounding kurtosis metric that basically is like what you just said. We're not making a growth investment to earn a three X return. That may happen a lot. We're only going to make an investment if we feel even at the growth stage that it has this crazy asymmetric upside. Why that specific thing as one of the small handful of things that was on your list of criteria? Well, first, I want to just highlight the choice of words there. So the choice of words is important. We have leptocritic return profiles. That comes from Ruloff. So Ruloff might be the only licensed actuary who is active in the venture capital business today.

33:25And so his actuarial statistics background blessed us with that word. But leptocritic return profile, to your point, basically means... Fat right tail. Better than typical chance of a 10X plus return. Yeah. The reason we do that is because our experience has been if you quote unquote underwrite to a 3X return, you end up with a 2X. If you think something actually has a chance to be a 10X plus return, maybe you end up with a 3X or a 5X or a 7X or something shorter there. But if you don't think that that upside potential is there, it's probably just not a good enough company or it's probably just not a big enough market.

34:02And so we don't underwrite to a 10X expecting every investment to produce a 10X return. But if we can't see a 10X return, it's probably just not good enough. There's three really fascinating business criteria, which again, you're evaluating companies when you can sink your teeth into them. There's customers, there's revenue, there's sometimes profit, there's a team, there's lots that you can dig into. And I'd love to just spend a minute on each of these concepts because a lot of the things you've talked about, they're just very simple and elegant, but I'm sure that there's just tremendous amounts of nuance and depth underneath the hood.

34:33The first is the term you use is that it's an emerging market leader, which sounds like it could be a double entendre, like mean a few different things. So maybe describe that one. Well, at first, the process by which we came to this spec, so it was many years of iteration, but there was an offsite we did. I think it was called the Inn at Pelican Hill down in SoCal, where we locked ourselves in a room for about two days and spent the entire two days just debating these words on a whiteboard to come up with exactly the right criteria to define our investments. So emerging market leader, it is a little bit of a double entendre.

35:08The thing that people misunderstand about this most frequently is, oh, okay, so we need to invest in a company that is the market leader today. No, we need to invest in the company that we believe will be the market leader tomorrow. It could be two people with an idea today. It could be objectively number seven today. Google was not the first search engine. Flextronics was not the first contract manufacturer. There are lots of examples of companies that were not the first and did not start in a market leadership position, but because they had a better architecture or a better business model or a better team, they ended up being the market leaders over time.

35:44And one thing that is sort of objectively true, particularly in the world of technology, is whoever ends up number one in the market doesn't just have their proportional share of the market cap. They have a disproportionate share of the market cap. And so investing in number two or number three in a market maybe can make a little bit of money, but you're not going to produce outsized returns for your limited partners. So it's really important for us to invest in the companies that we think are going to be number one in the market. The double entendre is that we think the market is emerging and we think that the company is emerging to become the leader of that market.

36:18So the market itself might not have much of a TAM today, but we have a reason to believe that it's going to have a TAM tomorrow. And a good example of that, I remember when Okta was going public in 2017, while they were on the roadshow, Forrester published a report that said the TAM for cloud identity was$150 million. Well, at that time had, I think, a little more than$150 million of revenue. And so people look at a static moment in time and have a hard time extrapolating that to five or 10 years from now when the company has matured and trying to have that point of view about where the market is going, coupled with the point of view about what position the company is going to occupy in that market.

37:01That is what ultimately gets us to whatever that out-year revenue projection might be. It's not a function of the financial model. It's a function of the market dynamics and the company's position. The second, respecting the specificity of the word choices here is maybe the most interesting one, which is unique and compelling value proposition, which on the surface sounds like, yeah, sure, like sounds great. But I think behind each of those words is something that is incredibly important as you evaluate the businesses. So maybe describe why those specific words. Yeah. So the first one, emerging market leader is a comment on revenue scale.

37:32The second one, unique and compelling value proposition is a comment on margin structure. And so if you have a unique value proposition, that should show up in gross margin. It should show up in gross margin because if your product is truly unique, you should be a price setter, not a price taker. And if you are a price setter, you should be able to set a price that's going to provide you with nice gross margins. And so unique value prop gets you to a good gross margin. Compelling value prop is a comment on operating margin. If your product is truly so compelling, you shouldn't have to bludgeon people to death with sales and marketing to get them to try it and to get them to pay you for it.

38:13So if it is truly compelling, that should show up in the efficiency of your go-to-market organization, or maybe there's a number like new ARR divided by sales and marketing, or there's a number like LTV to CAC, or there's a number like payback period. There should be some number, or maybe 99 % of your new customers come in organically. there should be some number that basically demonstrates how compelling the value prop is that leads to low sales and marketing, which in turn leads to a high operating margin. And so if you have an emerging market leader, chances are you'll have good revenue scale.

38:45If you have a unique and compelling value proposition, chances are you'll have a nice margin structure associated with that revenue scale. And those are the ingredients that should ultimately determine a quote unquote out year financial model, as opposed to the typical linear extrapolation that you might otherwise see. I'm realizing now that the third one, which is listed as sustainable competitive advantage, is the perfect third domino, which is, okay, you get to revenue scale, you've got good margins, how do you protect them? Exactly. And so what are you thinking when you're trying to suss out the end state potential future moat or sustainable competitive advantage?

39:19How do you do that? It seems really hard to know ahead of time. This rhymes with the conversation we're having on people. And the reason this is a hotly debated term, sustainable competitive advantage versus saying moats, because moats is probably the more common vernacular. The reason it's sustainable competitive advantage and not moats, a moat implies something that has been built and will protect you forever after. Whereas a sustainable competitive advantage is a bit more dynamic. It is an advantage that you are building every single day. and the number one sustainable competitive advantage that we see out of companies.

39:54It's not a network effect. It's not an ecosystem advantage. It's not some piece of IP that's impossible for other people to replicate. It is the DNA of the team. And the canonical example of this was in 1999, 2000, the smart money would have bet on eBay, but it turns out you should have bet on Amazon. People thought the marketplace business model of eBay was so elegant and defensible and selling books online was a commodity business. It turns out that one of those companies had Jeff Bezos and the other one didn't. And it was the founder and the DNA and the culture that was created that led to the compounding advantages over time.

40:29I think a more modern example of that would be DoorDash. There were plenty of people who tried to get into this delivery business. There was only one Tony Hsu. And so our partner, Alfred, who sponsored that investment, he met Tony at the seed stage and liked him, didn't quite have conviction, kept in touch with them. And then I remember Alfred told a story about he happened to go to dinner with Tony. He was at a dinner sitting next to Tony prior to the Series A and spent the entire dinner talking with him about how DoorDash worked. And the level of detail and nuance and grasp of the business fundamentals that Tony had blew Alfred away.

41:06And it was coming out of that that Alfred came back and said, we have to make this investment, not because of the business model, not because of the market, but just because of the founder. This is the sort of person who's going to create compounding advantages forever. Yeah. Having interviewed Tony, his command of that business, which is a very complicated business, is truly unbelievable. Just a special human. I've also loved the way you've articulated how you do interviews, reference checks, and just evaluate a person. You mentioned some of your favorite questions to ask on the long walk, ask about their family, ask about how they make decisions, all these sorts of things.

41:44There was a couple in there that I'd love to actually take and turn on you. And one of them, which I liked a lot, was if you had this magic wand that you could change something about yourself, what you would change. I'm curious what your answer is to that specific question. I was afraid you're going to do this. And I really should have prepared to answer my own questions because I haven't. But I have always admired, I guess it would be overly simplistic to say extroverts. People who are charming can command a room, natural networkers. Our partner, Carl Eschenbach, is a good example. Carl is now spending 99.9 % of his time as CEO of Workday, but he still helps us out on stuff from time to time.

42:25And Carl was here at Sequoia for about seven years. But when he walks into a room, it's as if there's this hushed voice that follows him just whispering, executives. And it's just the presence that he has is unbelievable. And his wife's the same. His wife Anna's the same way. Carl and Anna, just they light up a room. Actually, my wife is like that. She lights up a room as well. I don't light up a room. I've never been somebody who lights up a room. I'm probably hiding in the corner, hoping that somebody I already know comes over and talks to me so I'm not forced to go network. And that's something that I have worked on over and over and over again and forced myself into awkward situations over and over and over again.

43:05And as much as I've done it, it's still not comfortable. I'm still not good at it. And so that's probably the magic wand is to make myself a little more extroverted, a little more charming, a little more able to light up the room the way that some other people can. Well, we have to lead into our strengths. And you said earlier that you had this brute force mentality. And I love this line about keeping going until nothing surprises you when you're investigating a person doing reference checks, I guess, maybe even investigating a business. Maybe say a little bit more about what that actually takes and means to keep going until nothing surprises you.

43:37I just think it's like a really nice heuristic. We get the question a lot from founders, what are you looking for? And what they want is a simple answer. They want to hear, well, we're looking for your first 10 POCs to convert, or we're looking for X million of revenue growing Y percent year over year. You know, that's the sort of stuff that they want. The answer that I give them is a very frustrating answer, but it's the real answer, which is the thing that we're looking for is not perfection. The thing that we're looking for is clarity. whatever story you tell needs to be internally consistent.

44:11Whatever evidence is available to support that story needs to support the story. It can't be disconfirming with important aspects of the story. I mean, it would be Snowflake and Zoom. So we were lucky enough to get into business with Eric at Zoom when the company was close to 100 million in revenue. It was probably 85, 90 in that neighborhood. And at that scale, Zoom was already 80 % plus gross margin. And as much as they were trying to hire more people and burn cash, they just couldn't do it. The money from customers was coming in too fast. And so they were trying to burn cash, but they kept generating cash every single quarter.

44:44And so Zoom had perfect margins across the board, exponential growth. Snowflake, we were lucky to get into business with when they were just shy of 50 million of ARR. So not quite the same scale, but similar. Snowflake would have been more of the workday. Snowflake at that time had maybe 50 % gross margins and burning a whole ton of cash. And when we went from our first investment in Snowflake to our second investment in Snowflake, which is only about six months later, what triggered it was bad news. So we do these semi-annual portfolio reviews where our portfolio companies send us a bunch of information.

45:18The information comes in from Snowflake. They're behind the revenue plan. Gross margins are worse than expected. Free cash flow is worse than expected. And we got that and said, oh man, I'm in big trouble here. Maybe we shouldn't have made this investment. And so I got some time with Brad Floring, who is still the VP of FP &A at Snowflake, and asked him to walk me through what was going on in the numbers. And it turns out it was all good news. It turns out the reason it was all good news, which is hard to appreciate, why was revenue behind? Well, revenue is behind because they're landing much bigger deals than they were expecting to land.

45:53And those much bigger deals take more time to ramp up. And it's the ramping up of that stuff that determines the revenue, not the landing of the deals. Why is the gross margin behind? Well, the gross margin is behind because they're getting pulled globally faster than they expected, which means opening up availability zones in regions that are going to be underutilized around the world, which means that COGS utilization is not as high as you might expect it to be. The gross margins were also down because they're starting to get asked to do these full-scale Teradata replacements in the enterprise far earlier than they had anticipated, which means staffing up on professional services, which goes into Cox.

46:27And then operating margins are down because it turns out that the per rep productivity was much higher than they were expecting. And so they were loading up on the sales organization because the reps were just producing way faster and at a way higher level than they were expecting. And so all of the reason that the numbers were bad turned out to be good reasons. In that situation, we ended up having clarity on why the numbers were the numbers and what was actually happening in the business. And as a result, we went from an initial$50 million investment to another$200 million investment a couple months later.

47:00Anyway, the moral of the story is you just keep asking questions until the picture that's in your head becomes clear. And it doesn't have to be perfect. It does have to be clear. If it's not clear, you're not going to have a good understanding of what risks you are taking and what return you can expect in exchange. Do you have a specific goal in your mind when you're doing a reference check, that specific unit of investigation? Yes. If it's a reference check on a person, I want to understand the vector that is that human being. We're lucky to be surrounded by people who are pretty creative and good thinkers.

47:33And so this is a framework that we learned from Elon Musk about the output of an organization is the vector sum of its individuals. And the point is that a vector has both magnitude and direction. So you want to hire people with high magnitude, but then you have to make sure they're all pointing in the same direction. And so the thing that I want to get out of a reference check is the magnitude and the direction. And the magnitude is almost a top grading exercise. How good has this person been at each step of their life? Were they the best person in their high school? Were they the best person in their college?

48:02Were they the best person in job number one? Were they the best person in job number two? And that's a fairly blunt way to look at it. There are plenty of dimensions beyond did you have the highest GPA or did you have the best performance reviews that indicate exceptional performance. And so this is in part where the direction component comes in. Maybe they were not the best person in their high school because the thing they cared most about in high school was building businesses on the side. And maybe they were incredibly successful in that endeavor. Or maybe they just fell in love with coding and it turned out that they were a phenom in the open source world while they were failing in their history class.

48:35And so understanding what they actually care about and trying to figure out whether they've been exceptional at the things they really care about, that defines the direction of the vector and the magnitude of the vector. And that's the thing I'm trying to suss out. I know you studied physics in undergrad. The vector thing reminded me. And there's a lot of physics envy in investing, a want for formulas and variables. We've even done it a little bit today. Here's our three things. Where does the physics background help you? And where does physics in general should be left behind or fall short in the world of investing?

49:08My Catholic guilt compels me to acknowledge that I was only a physics major for two years of college. I ended up switching to economics and finance with a concentration in math. And so I just want to be clear, but I appreciate it. One way that I think about this, I'll use the analogy of the two critical ingredients in an investment are the people in the market. The market determines how big the company can get and the people determine how big the company will get. And I think similarly, when you approach things with the view of physics, you're sort of understanding the rules of the system, but it's the individual agents who are operating in that system that will ultimately determine the outcome.

49:53And so I think the physics point of view is very complementary to a much more human point of view. And if you can get both and you understand the system level dynamics, but then you also understand the individual actors within that system, that's where I think you end up with the highest likelihood of making a good decision. Can we apply all of that to the world of AI today, both the market and the sorts of people that you're beginning to see thrive in a frontier, in a Wild West feeling part of the world and incredibly exciting, but uncertain new technology? I would love you to just frame up first how you think about the market and your shorthand for what the opportunity is here and how to think about it.

50:35And then I'd love to talk about some more specifics, but maybe just starting broad strokes, What has you excited? What has you pausing? What has you most interested in the world of AI? As context for this, my first 10 years or so at Sequoia, we're basically focused on the cloud transition. And that was 2007 to 2017, which is a pretty wonderful time to have that as a focus. Come 2017-ish, it felt like the vast majority of first-class market opportunities in the world of cloud software had already been occupied. If you went to the other major tectonic shift that was happening at the time, which was mobile, the top 10 apps in the app store had been pretty static for a number of years.

51:16And so it felt like we were getting fairly late cycle as far as these technology platform shifts go. A bunch of us here started trying to think about what the next major platform shift might be. And at the time, our shorthand was data. And the reason our shorthand was data is because we just observed that the best application experiences we were seeing tended to be fueled by a pretty healthy dose of machine learning. And so it felt like the companies that were making use of all the data that was available were just creating better experiences and creating better businesses than companies that were not.

51:50And so we had this loose hypothesis that the next major platform shift was maybe something related to data. That was part of what informed the Snowflake investment. That was certainly what informed the Confluent investment, the dbt investment some of the other things in the modern data stack but it also led us toward natural language processing natural language understanding which is almost the predecessor term for what we now think of as llms led us to hugging face and led us to open ai it eventually led us to a bunch of different application or developer companies around that whole theme but i mentioned that because the thinking that's gone into the ai theme for us really began in earnest many years ago when we were seeing the maturation of the cloud and mobile cycle and trying to figure out what might be next.

52:34I'd say the reason that we have conviction in the AI theme and sort of what it is we've actually thought about kind of has to do with the precedent conditions coupled with just what we're observing in the environment. And when I talk about the precedent conditions, the idea of a neural net has been around for literally 70 or 80 years, but it hasn't been possible given compute, given bandwidth, given data, given talent. It hasn't been possible to put it into practice the way it is today until very recently. And the major accelerant, of course, was the release of ChatGPT, which we think will end up being this generation's Netscape moment.

53:14You know, if you go back to 1996, when the browser first came out of Netscape, that opened the eyes of everybody to the power of the internet. And I think similarly, when ChatGPT came out in the fall of 2022, it opened everybody's eyes to what was going on with LLMs or AI more broadly and gave people a visceral sense for what could be done. And so that was sort of a step function increase in the activity in this area. Earlier that summer, we had stable diffusion. If you remember, summer of 2022, stable diffusion came out. All of a sudden, people are creating these fantastical images and sending them around on Twitter.

53:47that took the AI market from researchers to researchers plus machine learning engineers. When ChatGPT came out that fall, it was another step function increase and the people were paying attention. And it went from ML engineers to all engineers, product managers, founders, consumers, boardrooms of Fortune 500 companies. And all that energy that's been focused on this has started to lead to some pretty interesting applications. How big do you think this would be if I froze the current frontier model capabilities? If I said, we're never going to get anything better than GPT-40 or CLAWD-3 or 4, the best ones that are out there today, do you think that it's still exciting or is most of the excitement dependent on continued and successful scaling of the quality and the capability of these things?

54:38I think that's a fantastic question. And we think about that a lot. And the short answer is, I think if you froze capabilities today, and the only thing that you invested in was optimization, making it cheaper, making it faster, making it easier. If you did that, you would revolutionize almost every industry on earth. I think the capabilities that exist today are so unbelievably powerful and have only just begun to be harnessed. There's an interesting question. If you were Sam Altman, what would you do? I think what he's doing currently is probably the right move, which is let's continue to be at the very bleeding edge.

55:21Let's continue to produce the very best models. and because we have an advantage in aggregating capital and talent, let's press that advantage and use it to stay on the absolute bleeding edge. There's an alternative version of the world, and I'm not recommending this, but I'm saying it's possible. There's an alternative version of the world where you say, okay, we think we're starting to see diminishing returns to scale, which means maybe we've squeezed about as much juice out of this architecture as we can squeeze, which means we're going to change our attention to a few other things. Number one, we're going to have a small team of true geniuses trying to figure out new architectures.

55:59Number two, we're going to move some of the compute from training into inference. And so instead of spending a bunch of compute building the model, we're going to spend a little bit more running the model, which is what people talk about when they talk about planning and reasoning, which basically means that the model can do more sophisticated things when you are asking it questions. And then the third thing that you could do, if you believed that returns to scale, were starting to diminish. The third thing that you could do is just straight up optimization and try to make it fast, try to make it cheap, try to make it easy.

56:30And in doing so, just run away with the developer ecosystem. So I think you could do that and you would have a cash generative business overnight. You would still have a dominant market position, but you would be taking the risk that we have not started to see diminishing returns to scale or that those returns have not diminished to the point where it invalidates the investment. So it's an interesting alternative in a version of the AI world to think about, but it's not the one that we're living in. How would you handicap the scale question about whether or not we've hit that scale wall, whether we can come up with creative ways of gathering more novel data or just other means of breaking through this?

57:07The bitter lesson seems to be the most interesting written piece about this, that you just need more data and more scale of data and the thing will keep getting better. But we've used all the data we have on the internet or in the written word or whatever. So what do you think the odds are that wall exists versus us just finding a way because we're humans and this is our way to push through it? I think people will always find ways to push through it. One of the data points that I find interesting, and different people have framed this in different ways. There's a guy named John Carmack who may be the world's greatest living engineer.

57:37Some people think that he is. And he has sequestered himself with a couple of other geniuses in the middle of nowhere, reading old research papers, trying to figure out if a better architecture already exists. It just hasn't been assembled in just the right way. And the reason he's doing that is because state-of-the-art LLMs are about four orders of magnitude less efficient than your brain. If you think about the basic input as energy and the basic output as computation, your brain is 10 ,000 times more efficient than a state-of-the-art LLM. Now, Andre Karpathy has actually had the same observation, but he thinks that it's six orders of magnitude.

58:19So one way or the other, these models are dramatically less efficient than the human brain. And so the reason that's important is that nature has shown us that a better architecture exists. And I have to imagine long before we get to anything that is universally agreed upon as AGI, we're going to end up with just a dramatically better architecture that's going to be far more energy efficient. It's not going to come out of just optimizing transformers. It's probably not going to come by just putting planning and reasoning on top. It's probably going to be some different base architecture. And I'm sure at some point, somebody will figure out what that is.

58:56This feels like the thing you said earlier, where if you bet on the system, you get the two and a half X Airbnb. If you bet on human ingenuity of the collective founder of humanity, you get the a hundred billion dollar Airbnb or something. Totally. Ravi said you explained this to him using a unit called math per Cheeto. What does that mean? Yeah, Ravi and I were talking about this. And I don't know if I've done a good job of this today, but one of the things that I tend to be known for here is trying to make things as simple as humanly possible. Math per Cheeto, if the input is energy, one fairly efficient source of energy for human beings is a bag of Cheetos.

59:34The output is computation. Math is a form of computation. And so Ravi and I were saying that the metric for the efficiency of an LLM should be maths per Cheeto. And that at the moment, humans can do way more maths per Cheeto than your best LLM. Maybe use Harvey as an example of, okay, let's just zoom. All this stuff is exciting. And I hope we break through all these walls and all the ways you described. And that would be so cool. But in the version of the world where we just have what we have today and just got to build a useful application and a business on top of it. But Harvey seems like a great example to just double click on, describe what it does and how it's using the model.

1:00:10And I would love you to just explain what you've seen so far, what lessons that business and product has taught you. Just zoom in on a real tangible example. Yeah. So I think Harvey's a good example because I think it is the first and best example of a new wave of application companies that will come out of this AI tectonic shift. They got started at exactly the right time. So they were the very first company to get access to GPT-4 to start building on top of it. And the founders come from the perfect background. Winston comes from the world of law. Gabe comes from AI research. And so together, they understand both the problem and the solution.

1:00:52And so it's a great example of founder market fit. It's a great example of the why now and being in exactly the right place at the right time. And what they've built over the last 18 months or so is the very best legal assistant. So it is not an AI lawyer at the moment. It is a legal assistant that can basically do the work of a first-year associate at a big law firm. And partners at big law firms have actually A-B tested the Harvey assistant versus the associate. and the Harvey assistant is just as good and immediate. And so a task that might've taken six hours instead takes six seconds. And so it's a pretty darn powerful assistant.

1:01:37The ambition for the company is to eventually use that to democratize the world of law. If you think about the legal world today, it's a rich person product. It's very expensive. And so whether you're a company or an individual, either you have a lot of money to spend on it or you're probably not going to get a very good legal service or you're probably not going to get any legal service. And it turns out that with AI and the fullness of time, we can provide world-class legal services and we can do it at a tiny fraction of the price. The idea for Harvey is never to replace the human beings. It's to dramatically expand the market to a whole bunch of people who don't have access to legal services today.

1:02:21And so for the very high-end law firms and that sort of thing, we're going to be an assistant. For the rest of the world, we're going to be the service, hopefully. And if we can pull that off, I think it has a chance to be an incredibly important company. When you're evaluating one of these products, because there's just not that many of them that are built and fleshed out the stuff's a year old, what are the things you're looking for that are distinct from the same things you might look for in what I'll call a non-AI product? Or Or is it just all the same stuff that it's just solves a problem efficiently and elegantly?

1:02:51And it's just the same, that there's just something different under the hood. Unique and compelling value prop. It comes back to that. Market by market, there are different pros and cons to the different technical approaches. But at the end of the day, the technical approach only matters to the extent it does something unique and compelling for the customer. And so we try to spend less time underwriting the architecture and more time underwriting the customer and just really understanding what problem this solves for them, why it's unique and compelling, how durable it is, how else they might solve that problem, where they see it going in the fullness of time, all that good stuff.

1:03:29I actually think a mistake that a lot of investors make, there are a lot of investors who are very technical and strongly weight their personal opinions of the architecture of the product. And that's a useful input. You just have to weight it appropriately because again, it's only as good as its impact on the customer. And so we try to be more customer oriented and less tech out more customer back. In the world of venture, it seems like there is this almost magic pixie dust that certain firms have that founders seek out. There's a handful. We could probably name them on the call pretty easily together.

1:04:03And everyone's guess that these names would be the same. Sequoia is certainly one of them. When it comes to the maintenance of that magic pixie dust that a few firms seem to have, where the winners keep winning and winning begets itself because that brand grows and the reputation grows, the role of the platform, the way you talk about Sequoia's platform seems to play a key role in the odds that that pixie dust will persist into the future. Can you describe the platform strategy to building an investment firm like Sequoia in a way that maybe others building investment firms might be able to borrow some of those concepts that have been effective for you?

1:04:36And first, I'll define what we mean by platform. So when I joined Sequoia, we had 14 people on the investment team and two people who I would call front office operators. We had one person in talent, one person in marketing. So we had 14 and two. If you fast forward to today, we have 27 people on the investment team and probably about 65 people who I'd say are front office operators, meaning marketing, talent, engineering, product, data science, design, and a handful of other things, customer partnerships. And so that group of operators is really what we mean by the platform. That's the bulk of what we mean by the platform.

1:05:21There are two key advantages we get out of that group. One is they dramatically amplify the efforts of the investment team. So one concrete example of that is the amount of information that we have. We have a homegrown CRM system powered by a homegrown data science system. The information that we have available in that system for a company that we've never met is more than the information we would have had on the same company 15 years ago at the time of making a final investment decision. And so that's a massive amplification of our ability to source and pick and work things through the funnel that leads to an investment.

1:06:02So that's one concrete example. So one thing that we get out of the platform is an amplification of our efforts as investors. The second thing that we get out of the platform is advantages that have a chance to compound over time. So historically, the only compounding advantage that you get in a venture capital business is your brand and your culture and your network. But all those things are somewhat ephemeral. One bad decision can tarnish your brand. The platform team is building things that can compound over time. And one tangible example of that is we have a clever way that one of our talent partners came up with to collect signals on people.

1:06:40We now have a couple hundred thousand people in our database on which we've collected these proprietary signals that are not available anywhere else. And so our ability to take a look at a company and pretty quickly get a good sense for the talents inside the building and how well they've hired based on the signals that are already in our system, that's an advantage that's going to just keep on compounding. We have a couple hundred thousand today. Over time, theoretically, we could have just about everybody in the technology world in that database. One other point that's worth mentioning here, the reason we decided to invest in our platform has to do with what we saw happening outside the building and a strategic choice that we made.

1:07:19So what we saw happening outside the building was the democratization of the means of production. And what I mean by that is any founder anywhere can now go online to educate themselves about the basics of technology and building a business and become an internet entrepreneur overnight. And as a result, the volume, variety, and velocity of startups has increased dramatically. But if we were still just 14 people or today 27 investors trying to do our jobs, we wouldn't be able to cover the universe of opportunities. We wouldn't be able to make our way through them efficiently. And so the strategic choice that we made was we could have taken the path of, OK, well, let's not have 27 investors.

1:07:59Let's have 270 investors. If we have a big team, we can cover everything. And the reason we specifically decided not to do that is because at the end of the day, there are only two things that you need a human being to do in the world of investing. Everything else can be automated, but the two things that you have to have a human being do. Number one, build the relationship with the founder. And number two, make the decision. It doesn't matter how many inputs you have. Somebody has to take those inputs and make the decision. The declarative statement, we should invest because. And so if the two things that we have to have human beings do are a relationship with a founder and make the decision, if we disperse the knowledge and experience of the partnership across a couple hundred people, any given one of them is not going to be all that special.

1:08:47If we concentrate the knowledge and the experience of the partnership on the smallest possible number of people, we have a chance for each one of those people to grow into something really special. And if we're hunting outlier founders, they don't want to deal with people who are just okay. They want to deal with people who are outliers themselves. And if we can hire people who already have outlier characteristics and then supercharge them with concentrated experience and knowledge, we have a chance to produce the next Doug Leone, the next Rulak Bota, the next Alfred Lynn. What Mike Moritz story most stands out in your memory where the story taught you something interesting?

1:09:30April 2010. You answer these crazy fast. It's very impressive. him. Well, this one does get on my head. So April of 2010, we had this partner named Chris Olson, who found at that time, a young man named Sebastian over in Stockholm. And Chris built a relationship with Sebastian and started to fall in love with a company named Klarna. And Chris asked Michael Moritz to parachute in to help him win this competitive investment. At that time, it was already a big deal in Europe and it was competitive. And so Chris and Michael Moritz end up securing the opportunity to invest in Klarna. And then they bring me along for a week to try to do all the diligence and meet the team and polish up the final investment recommendation.

1:10:17So I'm in Stockholm with Michael Moritz and Chris Olson for a week. We're spending all day at the company and Michael has not said a word. Chris is leading the conversation. I'm chipping in from time to time. And Michael is just sitting there silent, just listening, taking it all in. And Chris and I are desperate to know what he's thinking, particularly because asking him to spend a week in Stockholm, kind of a big ask. We want to make sure we're not wasting his time. So we finally get to dinner on night two or three. And at that point, Chris and I thought that the major issues in the investment were things like what's going to happen with interest rates, because remember, this is a bank with a balance sheet in the wake of the global financial crisis.

1:10:55We're very concerned about what's going to happen with interest rates. We were concerned about whether or not they'd ever be able to make it into Germany. At that time, they had a pretty strong position in the Nordics and Germany was the big market that they were trying to enter. And we go to dinner with Michael Moritz. Chris works up the nerve to say, OK, you know, what did you think? And Moritz in typical Moritzian fashion, there's an exhale and a long pause. And he says, the question is whether they can get to a few hundred million of net income and the answer will come down to the strength of the engineering team.

1:11:33Now, Chris and I, I don't know if we were showing a few hundred million of revenue in the model that we had built. And we certainly had not asked that many questions about the strength of the engineering team. And it turned out that Michael was exactly correct. And if you look at the company today, it's an absolute behemoth. and the strength of the engineering team was so critical because the value prop for this product was very strong for merchants and very strong for consumers. So it's been no brainer, except it was a pain in the butt to implement. And so the key was going to be, could you deal with the complexity of all the different e-commerce systems and all the different payment mechanisms and all the different preferences of the customers, could you deal with that complexity in an elegant way that is product-driven, not brute force-driven, to reduce the friction for people to deploy this product?

1:12:33And if you could, you were going to become ubiquitous. And if you couldn't, you weren't. And it was going to come down to the strength of the engineering team. That was a lesson for me because Chris and I, we had planned for this trip. We had our long list of all the different questions we wanted to ask. We were frantically scurrying about trying to do all of our work on the investment. And Michael got it down to the very simplest possible thing, which turned out to be exactly the right thing. And so I guess the lesson you taught me was you got to zoom out and make sure that you're operating at the right level of ambition.

1:13:05That was his 300 million net income thing. And that you're actually focused on the first order issue, which is the strength of the engineering team thing. When you think about the sensations in both your body and your mind of the feeling of being desperate to win, how would you describe what that feels like? It's funny you ask this because I've been concerned that as I've gotten older, I've lost some of the edge or some of the killer instinct. And then I was comforted by the camping trip that we went on with our founders a couple of weeks ago, where I felt like glimmers of it were still there.

1:13:39and they showed up in the silliest possible way, which was we did this set of activities, one of which was axe throwing. And you were on a clock and you had to get as many bullseyes as possible before the time expired. And I was on a team with a couple of our founders and we realized that one of them was better than the rest of us at throwing the axe. And so we ended up doing division of labor where my job was to sprint and retrieve the axes that had been thrown. And his job was to keep throwing them. And at one point, he got a few bullseyes in a row, and he turned around and raised his arms in victory.

1:14:14And there was still a minute or two left on the clock. And like a crazy person, I ran back to him yelling, no, no, no, there's more time. Keep throwing. Keep throwing. Keep throwing. So I guess the feeling is you get a little bit carried away with yourself. And actually, I think this is one of the things that makes Doug Leone so special, because he lives his entire life this way. You go into a mode where you are purely driven by the objective function. Whatever the thing is that you are trying to achieve, that is the only thing that you can think about and nothing else enters your consciousness.

1:14:45And in that case, I had lost track of the social graces of yelling at somebody to put their arms down and throw more axes because the objective function was the only thing that I could see. And then the analogy to Doug is one of the things that I think makes Doug so special. anytime you ask him to do anything. Personal discomfort, personal risk does not enter his calculation at all. If it is physically possible for him to do the thing that is required to achieve the mission, he will do it. And I think about that as it's the ultimate humility, not caring about himself, his ego, his comfort at all.

1:15:29It's the ultimate service mentality. And it's the ultimate mission orientation, where the only thing that you can see is the mission and what needs to be done to achieve the mission. And everything else just doesn't register. When you go into that zone where the only thing that you can see is the mission and everything else doesn't register, I think that's where that killer instinct comes in. If you ever do retire and you're at a retirement party, what do you hope people say about you? I think the themes have been consistent. I mentioned with Boston College, there's the Jezevin motto, men and women for others, and then the BC motto, ever to excel.

1:16:04So you have that concept of teamwork and that concept of performance. And at Sequoia, very explicitly, the two things we care about most are teamwork and performance. So I think the thing that I would hope to hear is that I was a top performer, but also a top teammate. One of the ways that most manifests on a day-to-day basis, the performance thing, I feel like we've covered. You're an animal and you want to win and you do whatever it takes. On the teammate side, where does that most commonly manifest? And what have you learned about it 15 plus years into doing this? Our partner, Andrew Reid, had this good line the other day, which was sometimes you need less leadership and more leadership.

1:16:44And what he meant by that was sometimes people think that leadership means telling other people what to do. but sometimes leadership actually just means doing the work so that people can see how it is supposed to be done. And because we're in an apprenticeship business, I think a lot of what we need to do is to just do the work, just do the basics of blocking and tackling, doing the job, and that that's more helpful to the other people on the team than any amount of one-on-ones or mentorship or structured feedback or whatever else. And so I think what being a good teammate means, when we construct a team to go after an investment, there are two roles.

1:17:31There's the sponsor and there's the wing person. And the sponsor's job is to secure the investment and to make the case internally. The wing person's job is to support the sponsor. I don't have to be the sponsor. I can be the wing person. There can be an investment that somebody else is sponsoring and my role is just to support them. And that might mean that I'm the one building the financial model and writing the memo and calling the customers. And they're the one doing the fun stuff of romancing the founder and making the case in the partner meeting and that sort of thing. And so I think what it means to be a good teammate comes back to that mission orientation that we were talking about earlier.

1:18:07Whatever the mission is, whatever the job is that needs to be done, just do the job. It doesn't matter what your specific role is in achieving the mission. It just matters that we achieve the mission. If you think about the landscape of this style of investing, it's mature. There's lots of firms. When you started, it was much smaller, both in people, firms, assets, investments, et cetera. Do you still think there are open zones of opportunity to try new concepts and reinvent the game a little bit from the investing side? I guess asked differently, if I forced you to go start a new firm with none of the benefits of the existing firm, just you, how you would approach that challenge?

1:18:46where the goal was to win and be successful and back great companies. How would you go to market as a new investor in this more mature environment? It's hard for me to think of anything other than what my wife, Sarah, is doing with her firm, Conviction. Tell me about it. I actually think what she's doing is exactly right. And I'll give you the specific example, but I can also generalize from there. So Sarah was a partner at Greylock for about a decade and then left just under two years ago to start a new firm, which is called Conviction Partners. And the reason she started it was because she saw this new crop of what she calls software 3.0, which is basically AI-driven companies starting to emerge and wanted to build a firm that could be built from the ground up to service that new crop of entrepreneurs.

1:19:34The thing that I think is so effective and so special about what she's doing is that unlike a lot of people in the venture capital world, she believes that being small is a weapon and that you don't get advantages out of scale in the venture capital business. You get advantages out of quality. And so she kept her first funds much smaller than it could have been. She's kept her team much smaller than it could have been. She's kept her portfolio much smaller than it could have been. And in each step of the way, she's optimized for quality. and the benefit that you get from optimizing for quality is that if you achieve high quality, the growth comes to you.

1:20:16You look for growth, you're not likely to reverse engineer quality. If you look for quality, you're going to have plenty of choices about how much you want to grow. And so she's assembled an exceptional portfolio. She's now starting to think about fund number two, and it's going to be the easiest thing in the world to raise it. She's going to keep it smaller than it needs to be because, again, she's not optimizing for assets under management. She's optimizing for quality. And I think that's a great way to build a business. And then the way I would generalize that is not just the point on quality, but also she is currently known for one thing, early stage AI companies, early defined as series A or earlier, and AI defined as AI.

1:21:01If you are an early stage AI company, there's a pretty decent chance that you're going to think of her. If you are not an early stage AI company, there's a pretty decent chance that you're not. I was on a board with a guy named Jeff Richards from GGV, and he referred to this as the chicken issue. At some point, some genius at Chick-fil-A probably said, hey, if we put burgers on the menu, we can attract more customers. And somebody else said, yeah, but the thing we're known for is chicken. If you want a burger, you should go somewhere else. Most companies have a chicken issue where they want to do the chicken and the burger.

1:21:31And just focusing on the chicken is important. Yeah, it's funny. It's turtles all the way down. It's probably the same advice for a new technology software company, too. You got to do one thing and do it really well to get going. And that just seems to be a universally good advice for sure. Is there anything else about your whole world that you wish was meaningfully different than it was like system settings or just ways of doing things or just norms that armed with that magic wand, you would change drastically? Yes. I probably won't articulate this in the best possible way, but my partner Ruloff expresses this in a pretty good way where he says, look, venture capital is not an asset class.

1:22:10What he means by that is less than 1 % of the companies that get started end up accounting for 99 % of the market cap that's created. And I don't know if Those are the exact right numbers, but it's something like that. And so if you want to approach this as an asset class and buy an index of all the startups, you're going to get drowned out with noise. And the vast majority of those investments are going to be no good because the vast majority of those companies don't need to exist. They're not solving an important problem or they're not doing it in a unique and compelling way. And so when you approach venture capital as an asset class, you end up with companies that don't need to exist, funds that are not going to perform, and people who are attracted to maybe the fame of being a popular founder, maybe the perceived riches of having a successful exit, but they're not attracted to this necessarily for the right reasons.

1:23:11And when I say right reasons, everything is relative, right in terms of what I think of as right. It's not objectively right. It's just my personal opinion or my personal point of view. But if you say, okay, well, what are the right reasons? I think if you are a founder who really cares about some problem in the world that is not being solved in just the right way, and you want to dedicate the next couple of decades of your life to solving that problem, that's a great reason to go build a business. If you are an investor who believes that entrepreneurship, more so than any other force, shapes the future of the world that we get to live in, and you want to dedicate your life to serving those entrepreneurs so that they can realize the maximum possible impact of their dream, that's a pretty good reason to be an investor.

1:24:09If you're an entrepreneur who just wants to get invited to the fancy parties and conferences and issue press releases about your latest funding round and tell your friends you're a unicorn, that's not a great reason to be a founder. And if you're an investor who wants to maximize assets under management so you can milk the fee stream and spend your time on Twitter pontificating about the future direction of AI so that you can show up in news reports, that's not a great reason to be an investor. And so if I had a magic wand that I could change one thing about the industry, I would try to slice off some of the people who are participating without the most pure motives and increase the concentration of founders who really care about their customers and investors who really care about their founders.

1:24:57Hey, man. What does it take to achieve legendary potential? I love that term you use all the time. That word is really important to Sequoia, legendary. It implies effort and scope of ambition and all the things that we've talked about and that you just talked about. I would love you to just sum it all up with what you think it takes and what you've watched it take. This isn't theoretical. You've seen it happen many times. What does it take to achieve that sort of potential? Why is that so motivating to you? So we had Max Rhodes, who is the founder and CEO of FAIR, at our offsite maybe last year.

1:25:34and he particularly in the early days of fair was just legendary for his work ethic and he still is but he was very legendary once upon a time and so we asked him what is it that keeps you going when all of your friends are out having fun or when you're exhausted and frustrated and just want to quit what is it that keeps you going and he said it was a voice echoing in the back of his head happened to be the voice of our former partner, Michael Moritz, who he once asked this question, what is it that separates the truly legendary companies from all the rest? And the voice was saying, relentless application of force.

1:26:15And I think that's it. Of course, the question is, what is it that causes you to relentlessly apply the force? and that gets to the core of who you are and what you care about and why you're building this company to begin with. And to my earlier comments, if your motivation is to release press releases about how you're now a unicorn, that's probably not a durable enough motivation to really keep going when things get tough. And your motivation doesn't have to be obsession around the customer problem. Maybe your motivation is you just really love building things or you really prioritize craft and you really want to build just a beautiful, amazing product that people are going to love.

1:26:56Or maybe you got into it for the wrong reasons, but now that you have a couple hundred employees and you realize that people are really counting on you, you feel a sense of responsibility and you really want to do right by them. And so there are plenty of reasons that might cause you to relentlessly apply that force. But I would say the thing that probably separates the legendary from the rest is in fact the relentless application and force. I think you might know my traditional closing question for everybody, which I love and is a very appropriate one, given a lot of the values you've talked about and just your own past.

1:27:29What is the kindest thing that anyone's ever done for you? I had a hard time coming up with a good answer to this question, knowing that you're going to be asking it. And the best thing that I could come up with was, it's going to sound very generic, but I'll make it a little bit more specific. And the very generic form is giving me a chance. the thing that's a little more specific is I had this nice scholarship in college. And I remember the wife of the director of the scholarship program who oversaw all of us to make sure we weren't losing our way and just an incredibly sweet lady. I remember one time she made the comment, you know, you're kind of rough around the edges, but you clean up pretty nice.

1:28:11And And I think that's a fair comment. I think particularly high school, college, earlier in my career, rough around the edges might have been a generous statement. I know Doug refers to the early version of himself as insufferable. Maybe I was somewhere close to that. I was certainly a bit prickly and a bit full of myself. And so I guess the kindest thing that anybody's ever done for me is to see through that and to see whatever goodness or whatever positive attributes might have been hiding inside and to help those things come out and flourish over time. I'd start with my parents who were strict, but not hard on me as a kid and gave me the room to figure out who I was.

1:28:51And then in college, the people who were kind enough to give me a scholarship or spend time with me, my first job, and I mentioned John Carroll. It's funny, I heard after the fact that there was general agreement at Summit Partners that I should be hired, but no one person actually wanted me on their team. and he was the one who said, well, okay, I'll take him. And then I think here with Doug, as I mentioned, I was the youngest person we'd ever hired. I was going to experiment. I was far from perfect and I almost got fired multiple times after I got here, but Doug was the one who took a risk on hiring me and Jim Getz was the one who stood up for me when I was not doing so well.

1:29:27Why did you almost get fired? What was the closest? It wasn't that I did something egregious. It wasn't a thing that caused me to get fired. In the context of me joining, we were just making growth investing a first-class citizen. And I was hired from Summit Partners, which is a really good growth equity firm. And I misunderstood my job as teach Sequoia how to invest like Summit. And the thing that I should have done was understand who Sequoia is and then extend Sequoia into growth. And so when I first got here, it was kind of like oil and water where I was just trying to rinse and repeat with the stuff that I've earned at Summit, which is not the right stuff to do as part of Sequoia.

1:30:06I wasn't learning fast enough. And at one point, apparently five of six general partners said that I should be let go. And Jim Getz was the one who said, over my dead body. The reason Jim threw his body across the tracks when everybody else wanted to fire me was it was a little bit skills or attributes or whatever. But the biggest thing was intentions. He could see that I desperately wanted to do the right thing. Fascinating. Pat, This has been a total blast and pleasure. I've learned a lot. Thank you so much for your time. Awesome. Thank you. If you enjoyed this episode, check out joincolossus.com.

1:30:40There you'll find every episode of this podcast complete with transcripts, show notes, and resources to keep learning. You can also sign up for our newsletter, Colossus Weekly, where we condense episodes to the big ideas, quotations, and more, as well as share the best content we find on the internet every week.

1:31:11Outro Music