The Exchanges

Q Miles Grimshaw at Benchmark, you were the first in this discussion to bring up co-pilots. Are co-pilots an incumbent strategy?

A I think Copilot is an incumbent's strategy. Incumbents own distribution, they own data, they own the UX, and they own a business model that all aligns to a Copilot. Copilot as GitHub Copilot, right, like inline suggestions. Think of it like how most go to, go to any Microsoft product right now. Every Microsoft product now is a Copilot experience instead of a sidebar, an autofill, things like that, right, where the UX, the core product is a layer on top of it, right, it's immediately added in, which is also totally a Incumbent strategy. And it's still about sort of supercharging that worker, but still where every user has a seed and every user is doing most of the work. And it works probably, you know, if you think about the evolution here, the models, most of what's rolled out might not be good enough for some of this yet, right? But that's what will come around the corner. You know, if you think back to Salesforce disrupting Siebel, Salesforce launched like five years after Netscape launched. Like it might take a moment for that to happen, but the copilot, this idea of I'm still the pilot, I'm still the user controlling everything, and it's sort of like giving me assistive suggestions like GitHub Copilot. Fits into the UX of incumbents, it fits into the business model of incumbents, and they already control all that distribution. The opportunity offered up to a startup, being …

AI assessment note: “I think Copilot is an incumbent's strategy. Incumbents own distribution, they own data”

Q Mike Maple says, like, what secret do you believe that the world doesn't see yet? And I always tend to ask that question. What did you miss with Plaid?

A This to me would be a, a story of the way the world, sometimes the way the world doesn't necessarily work the way it should work, hyper rationally. Plaid really solved OAuth for bank accounts, which was done in an arcane way before penny drops, where you have to wait six days. So it was a radical improvement. Um, but the idea that OAuth to your bank account, OAuth in general was a hard problem was like, not that hard. Like Reddit, I think probably has an OAuth system. Like they just, you can build an OAuth system. Conceptually the world, the world should work as every bank. And there's, by the way, highly concentrated in the US. So like five banks control over 50% of accounts in the UK, in the US. Um, the idea that those five banks couldn't offer an OAuth system to get account and routing numbers, that's all you need, two routing numbers, two numbers. Out of them to authenticate. In my mind, it should work. Now they have no incentive to do that because if they let you do that, it's easier to move money out, and they don't want you to move money out because they make money on you storing money there. That's how a bank works. Um, you get lock in to other products, but also just you get net interest margin on the core assets you manage. And so no bank has an incentive to actually offer this. And so again, you think user first, you know, incentives in the system. Um, it sort of sho…

AI assessment note: “I, you know, I probably got, I got too intellectual about it.”

Q Tom, it's so interesting you say there about the co-pilot strategy. It's the strategy that I see as the entry wedge for all startups today when I'm investing. Myles Grimshaw at Thrive, I know you have some thoughts on the co-pilot strategy, so hit me, how do you think about the co-pilot strategy today for startups, and how you feel about it?

A I think Copilot is an incumbent's strategy. Incumbents own distribution, they own data, they own the UX, and they own a business model that all aligns to a Copilot. Copilot as GitHub Copilot, right, like inline suggestions. Think of it like how most go to, go to any Microsoft product right now. Every Microsoft product now is a Copilot experience, instead of a sidebar, an autofill, things like that, right, where the UX, the, the core product is a layer on top of it, right, it's immediately added in, which is also totally a Incumbent strategy. And it's still about sort of supercharging that worker, but still where every user has a seed and every user is doing most of the work. And it works probably, you know, if you think about the evolution here, the models, most of what's rolled out might not be good enough for some of this yet, right? But that's what will come around the corner. You know, if you think back to Salesforce disrupting Siebel, Salesforce launched like five years after Netscape launched. Like it might take a moment for that to happen, but the co-pilot, this idea of I'm still the pilot, I'm still the user controlling everything, and it's sort of like giving me assistive suggestions like GitHub Copilot. Fits into the UX of incumbents, it fits into the business model of incumbents, and they already control all that distribution. The opportunity offered up to a startup,…

AI assessment note: “I think Copilot is an incumbent's strategy. Incumbents own distribution, they own data”

Q the single biggest thing that he's been taught by being at Thrive. This maniacal focus on who's the user and solely focusing on value derived. When you think back on your time at Thrive, I think it was six, Six years and nine months, according to the trusty LinkedIn. What were one or two of your biggest takeaways? Because it was such a meteoric journey for you and for Thrive.

A Maybe I'd say three things that, uh, that jump out to me as lessons from Josh. There's lots of, um, skills and the like investing, but sort of maybe ways of looking at the world and being in the world, um, are the lessons that you really carry with you the longest. One from Josh would really be That kindness and competitiveness don't have to be oil and water, um, in, uh, in, in business. And sometimes the cliches treat those as, um, you know, opposite poles of a magnet. You're either, you know, a pushover or you're, you know, doggedly competitive. And I think, um, he really showed that, that, that they can be in harmony and, and, um, and that's a really special thing. So kindness and competitiveness. Another would be, um, sort of this idea, like, yes, you can, um, And I think belief in someone's potential and, and someone's possibilities is such a, an amazing gift. Um, he suddenly gave me that gift, um, and I've watched him give many others that gift, many founders that gift of like, yes, you can, a real belief in their potential. And it's a, it's a beautiful thing to give someone and it's an amazing way to go through, um, the world and keep, keep, keep with you.

AI assessment note: “Maybe I'd say three things that, uh, that jump out to me as lessons”

Q Who's the best board member that you've sat on a board with?

A My partner, Eric. Um, why Eric? I think there are two things that, um, two, two expressions in venture that I don't like and think differently about. Um, and I think he embodies them the way I think you should think differently about them in, in, in, in the context of a board too, in a relationship. One is founder friendly. Like, I, I don't think you should be founder friendly. Not that you shouldn't be friendly, obviously. But I think that the higher purpose is like founder respect, um, and friendly can imply like cheerleading, yes, person, et cetera. And I think to honor and respect a great founder, um, you know, means sharing truths, um, maybe not saying the easiest thing to be saying, maybe saying the uncomfortable thing to say. And I think that that is like respect, um, you know, and, and so I kind of like founder respect and founder friendly more. The, and, and the other would be, I think first call is a very reactive mindset. So Venture always talks about wanting to be the founder's first call. Um, I think of it as first to call and being proactive, right? Um, obviously not obnoxious. It's not your company, right? But being proactive versus just reactive first, first call implies lots of reactivity. I think Eric is, um, you know, if you, if you looked up his phone and looked at who he's called, um, Uh, in the last, you know, couple days or whatever, he'd have caught a lo…

AI assessment note: “My partner, Eric. Um, why Eric?”

Q Who's the best board member that you've sat on a board with?

A My partner, Eric. Um, why Eric? I think there are two things that, um, two, two expressions in venture that I don't like and think differently about. Um, and I think he embodies them the way I think you should think differently about them in, in, in, in the context of a board too, in a relationship. One is founder friendly. Like, I, I don't think you should be founder friendly. Not that you shouldn't be friendly, obviously. But I think that the higher purpose is like founder respect, um, and friendly can imply like cheerleading, yes, person, et cetera. And I think to honor and respect a great founder, um, you know, means sharing truths, um, maybe not saying the easiest thing to be saying, maybe saying the uncomfortable thing to say. And I think that that is like respect, um, you know, and, and so I kind of like founder respect and founder friendly more. The, and, and the other would be, I think first call is a very reactive mindset. So Venture always talks about wanting to be the founder's first call. Um, I think of it as first to call and being proactive, right? Um, obviously not obnoxious. It's not your company, right? But being proactive versus just reactive first, first call implies lots of reactivity. I think Eric is, um, you know, if you, if you looked up his phone and looked at who he's called, um, Uh, in the last, you know, couple days or whatever, he'd have caught a lo…

AI assessment note: “My partner, Eric. Um, why Eric?”

Q the single biggest thing that he's been taught by being at Thrive. This maniacal focus on who's the user and solely focusing on value derived. When you think back on your time at Thrive, I think it was six, Six years and nine months, according to the trusty LinkedIn. What were one or two of your biggest takeaways? Because it was such a meteoric journey for you and for Thrive.

A Maybe I'd say three things that, uh, that jump out to me as lessons from Josh. There's lots of, um, skills and the like investing, but sort of maybe ways of looking at the world and being in the world, um, are the lessons that you really carry with you the longest. One from Josh would really be That kindness and competitiveness don't have to be oil and water, um, in, uh, in, in business. And sometimes the cliches treat those as, um, you know, opposite poles of a magnet. You're either, you know, a pushover or you're, you know, doggedly competitive. And I think, um, he really showed that, that, that they can be in harmony and, and, um, and that's a really special thing. So kindness and competitiveness. Another would be, um, sort of this idea, like, yes, you can, um, And I think belief in someone's potential and, and someone's possibilities is such a, an amazing gift. Um, he suddenly gave me that gift, um, and I've watched him give many others that gift, many founders that gift of like, yes, you can, a real belief in their potential. And it's a, it's a beautiful thing to give someone and it's an amazing way to go through, um, the world and keep, keep, keep with you.

AI assessment note: “One from Josh would really be That kindness and competitiveness don't have to be oil and water”

Q VCs do though? You know, often people say on the show, VCs do more harm than good. Do you think that's fair? I do worry often bluntly with my advice. I'm the first to say, I actually don't know. I've never done this, but let me introduce you to someone who has, because I'm very nervous of what I say, because people do put weight on it as a founder.

A I'd say two, one, you should always have the Hippocratic oath. Do no harm. For sure. I've been in some, some board meetings where investors don't really get the product, don't really get the strategy that has to be pursued. They're trying to apply like procedures and checklists, you know, versus strategies, um, specific to the instance. It becomes a reporting structure. It becomes a distraction. It becomes a management burden for the founder. And so yeah, Hippocratic Oath. The second is though, I think that it's not that you always have the answer. Part of it is that a founder trusts you to say, Hey, I've got, I'm working on this and come to you with it and you go work on it together. And so I don't think you have to always, um, uh, in fact, it would be dangerous, I think, to assume you've always got like the perfect answer and opinion on everything so much as to say with that context, with that trust, with the, with high velocity, you know, um, communication, you, you can go work through it.

AI assessment note: “one, you should always have the Hippocratic oath. Do no harm. For sure.”

Q Well, speaking of like going on that journey, how do you bring the partnership along with you and your excitement in a deal? Do you tag teaming up at Benchmark where you say, hey, Chathan, let's get on this together. And then you discuss it as a partnership, but you and Chathan are the leads. How do you bring everyone with you in your conviction for a company?

A Yeah, we do. We do a lot of that. Like, hey, come, come, come join this. Hey, let's go meet together. There's ultimately one person who is, takes a board seat, right? Manifested that way. Um, but it is a, again, we're an equal partnership, so it's a whole firm commitment. No person is incentivized to, to, to do more, help more one thing versus another. A founder really gets the whole partnership. Um, but obviously they've, they have a principal relationship A dominant relationship with one person, but we're, we're constantly pulling each other and actually just before I came here, a founder who works with Bill Gurley, I'd spent some time with recently and sent me a text, like remembering something I'd said to him in a line because he was using it again and again. And so we're, you know, we're always pulling each other into current investments, future investments, um, sort of in a very fluid way.

AI assessment note: “Yeah, we do. We do a lot of that. Like, hey, come, come, come join this.”

Q specific trajectory of the company, but an example for me would be, respectfully, Clubhouse. If I were to do reserves, I would have concentrated all my reserves into Clubhouse. Actually, that would have been the wrong decision. Or for Docker, you would have not invested in subsequent rounds of Docker because of traction misleading you. How do you think about reserves management today, given data can mislead you very often?

A We really don't think about it very much. If you would hit the spreadsheet of our investments, um, almost all the capital we invest is the very first, um, uh, very first time we partner. And yeah, we have reserves, you know, because tough times happen, you know, who knows what, right? Like we, we certainly have reserves, but we don't reserve, um, with a mindset of buying more Um, later. Um, we don't reserve with a mindset of, hey, we started, we got some ownership now, we're gonna try and buy, um, some later. Um, we're either fully in, and it's a whole commitment, and that's, that's what we're doing, or we're not in. And so we don't reserve with a, um, with a future investment orientation.

AI assessment note: “We really don't think about it very much.”

Q Miles Grimshaw at Benchmark, you were the first in this discussion to bring up co-pilots. Are co-pilots an incumbent strategy?

A I think Copilot is an incumbent's strategy. Incumbents own distribution, they own data, they own the UX, and they own a business model that all aligns to a Copilot. Copilot as GitHub Copilot, right, like inline suggestions. Think of it like how most go to, go to any Microsoft product right now. Every Microsoft product now is a Copilot experience instead of a sidebar, an autofill, things like that, right, where the UX, the core product is a layer on top of it, right, it's immediately added in, which is also totally a Incumbent strategy. And it's still about sort of supercharging that worker, but still where every user has a seed and every user is doing most of the work. And it works probably, you know, if you think about the evolution here, the models, most of what's rolled out might not be good enough for some of this yet, right? But that's what will come around the corner. You know, if you think back to Salesforce disrupting Siebel, Salesforce launched like five years after Netscape launched. Like it might take a moment for that to happen, but the copilot, this idea of I'm still the pilot, I'm still the user controlling everything, and it's sort of like giving me assistive suggestions like GitHub Copilot. Fits into the UX of incumbents, it fits into the business model of incumbents, and they already control all that distribution. The opportunity offered up to a startup, being …

AI assessment note: “I think Copilot is an incumbent's strategy.”

Q a lot from successes and failures. You mentioned your lattice, your bench, and God, I remember you introduced me to Sajith. Like, years and years ago when they were not very well known at all. Um, but Benchling, Lattice, Monzo, there's many more. But when you think about your biggest wins, what have been your biggest lessons or takeaways when you reflect on those wins? Because they do ingrain lessons.

A I think if you looked at Realized, it would probably GitHub or Segment. If you look at Unrealized, maybe Benchling or Monzo, Um, and so Benchling or Monzo, it's funny, you learn two very different things. Um, Benchling really is, is, um, a visceral felt story of, of the power of someone perfectly matched to a totally new market and a story of what could maybe seem like a small market because it's a vertical market, um, getting opened up and being the leading provider to grow with that market. If you want a positive feedback loop to small but growing markets with, you know, high degrees of market share in them, um, and a founder perfectly matched to, to, to dominate that market, uh, and what could be possible. It's a great, it gives you a feedback loop for that lived reality. Monzo is the exact opposite in many ways in that is banking. Outside the UK audience, um, they, they might not have barely heard of Monzo because it's a UK phenomena, but in the UK there are something like 60, seventy million core checking account users in the UK. Monzo now has seven million accounts. It's like a 10th of the British population has signed up to this thing.

AI assessment note: “Benchling really is, is, um, a visceral felt story of, of the power”

Q You said about new variants there. The thing I do want to ask is, you know, obviously you led the Langchain deal, one of the hottest and most competitive deals, I think, when you led it, and then subsequent round. Um, my question to you is, And actually Harrison's. How did you first get into AI and LLMs when you did, which was considerably before everyone else actually?

A Yeah, so I was spending time sometime in twenty-twenty-two, probably early, like not that early in the grand scheme of things, right? Um, but I guess earlier than the marginal- Zeitgeist. We're captivated. You saw the first versions of things which were, um, uh, just sort of spitting out content, right? Um, but you started to see in mid-twenty-twenty-two or so, this idea that the, the models had some amount of reasoning Capacity. There's, there's two things I might point to in that one. One is they can do things like know which colors are like each other, which things are, the ability to do math. Um, now in some sense that's learned from the language and it's just statistics at the end of the day. Um, but the fact that those properties were kind of there was fascinating. And then you've got things like the react paper, which was sort of this reasoning and action and the model kind of giving it feedback to itself. And you started to see, at least I started to see when I saw those is like, There might be so much more here. It might be, um, a sort of new computing paradigm in a way. And there's more power in this than just sort of marketing copy. And you combine that with, um, another thing, another aspect that I felt could be possible and is proving out to be true is that ML had been in many ways, mostly kind of just like, um, reserved for ML wizardry. Right? It was sort of the, …

AI assessment note: “I was spending time sometime in twenty-twenty-two, probably early... but you started to see”

Q Yeah. Uh, my favorite is actually, um, if I have more time, I would have written a shorter letter. I love that. What was the third one? Sorry, I cut you off there.

A The third one, um, would be this, um, again, The greatest gift of things you sort of carry with you as ways of being would be what I might call, um, impatiently patient. Again, sort of a little, a duality that almost shouldn't exist, um, uh, which is why not now? Why not yesterday? Like hyperactive all the time, but still a really high standard and really patient for the long term. And so that combination of like, Being fully present, like fully engaged, new, um, and, and, and, um, and active, but also like high standards and not rushed. And that duality, I think, um, he, he really lives. And, um, you know, we were having a mentor like that, um, early on who can show you that through that lived experience is a, is a really powerful thing to carry with you.

AI assessment note: “The third one, um, would be this... what I might call, um, impatiently patient.”

Q a lot from successes and failures. You mentioned your lattice, your bench, and God, I remember you introduced me to Sajith. Like, years and years ago when they were not very well known at all. Um, but Benchling, Lattice, Monzo, there's many more. But when you think about your biggest wins, what have been your biggest lessons or takeaways when you reflect on those wins? Because they do ingrain lessons.

A I think if you looked at Realized, it would probably GitHub or Segment. If you look at Unrealized, maybe Benchling or Monzo, Um, and so Benchling or Monzo, it's funny, you learn two very different things. Um, Benchling really is, is, um, a visceral felt story of, of the power of someone perfectly matched to a totally new market and a story of what could maybe seem like a small market because it's a vertical market, um, getting opened up and being the leading provider to grow with that market. If you want a positive feedback loop to small but growing markets with, you know, high degrees of market share in them, um, and a founder perfectly matched to, to, to dominate that market, uh, and what could be possible. It's a great, it gives you a feedback loop for that lived reality. Monzo is the exact opposite in many ways in that is banking. Outside the UK audience, um, they, they might not have barely heard of Monzo because it's a UK phenomena, but in the UK there are something like 60, seventy million core checking account users in the UK. Monzo now has seven million accounts. It's like a 10th of the British population has signed up to this thing.

AI assessment note: “Benchling or Monzo, it's funny, you learn two very different things.”

Q Can I ask, you said before that the comparison to mobile, which many have made, many have said that actually this is similar to mobile and actually the incumbents will need to shift to it like they did to mobile. You said before maybe that's not the right analogy. Why do you think the comparison to mobile is not the right analogy?

A Shifts happen. We want to analogize and no analogy is going to be perfect in many ways. Um, I'll compare it more to the idea of what the internet enabled, the mobile enabled, But like both of those were also new distribution channels, and this is not a new distribution channel. Um, so both of them are fundamentally flawed in that way. I'd also argue that if you think about distribution is closer to the internet in that the internet was a distribution channel. It also catalyzed, um, the sort of buying, it catalyzed the, the idea of sort of buying an internet native solutions, right? You, you think back to the story of Microsoft and, and, um, 95 Bill Gates wrote sort of the Paul Harbour memo of like, we need an internet native Kind of like strategy to it, right? It was a catalyst for, um, consumption and sort of change within those organizations, more so than mobile. But none of them are perfect. The reason why I think mobile is, is a more flawed way to think about it is that mobile really was, um, the same sort of architecture and business model as, as what was before it. Um, the way we architected applications didn't really change. Um, the business models associated with them was pretty much the same. And those things can really lead to Um, disruption can lead to new startup competing. And most of the incumbents before mobile were the successes in mobile. What the new people th…

AI assessment note: “mobile really was, um, the same sort of architecture and business model”

Q When you say co-pilot, can you just explain that?

A Think, think of co-pilot as GitHub co-pilot, right? Like inline, um, suggestions, um, think of it, uh, like how most, Go to, go to any Microsoft product right now. Every Microsoft product now is a copilot experience instead of a sidebar, an autofill, things like that. The core product is a layer on top of it, right? It's immediately added in, um, which is also, um, you know, totally incumbent strategy. And it's still about sort of supercharging that worker, but still where every user has a seed and every user is doing most of the work. Um, and it works probably, you know, if you think about the evolution here, um, the, the models, Um, most of what's rolled out might not be good enough for some of this yet, right? But that's what will come around the corner. You know, if you think back to Salesforce disrupting Siebel, Salesforce launched like five years after Netscape launched. Like it might take a moment, um, for, for that to happen. The copilot, this idea of I'm still the pilot, I'm still the user controlling everything. And it's sort of like giving me assistive suggestions like GitHub copilot. Um, it fits into the UX of incumbents. It fits into the business model. They have incumbents and they already control all that distribution. And so I think, uh, the opportunity offered up to a startup, um, being a co-pilot for something else, like probably won't be that amazing. And the…

AI assessment note: “Think, think of co-pilot as GitHub co-pilot, right? Like inline, um, suggestions”

Q There's some things I just have to ask because there's a lot of like tropes that venture investors are saying about AI, which I don't often know how to respond to. A lot of like, They're all just cheap rappers on top of large language models and just rappers on GPT layers. How do you respond to that?

A I think there are some obviously ones that are, that are not that already. Right. And so, um, think about what's happened in some point, mostly probably to, um, imagery, things like mid journey or runway ML and video and some of these other things. Right. Um, clearly kind of not rappers in those ways. A lot of what has Happened in SAS and certainly what we see the most of in the news because incumbents are doing is the copilot. And I think copilot is an incumbent strategy. Is that a wrap around the model? Sort of in a way, right? I think they're doing a bunch of other stuff to it, right? For, for this new application paradigm, but it's, it's that, you know, we at Benchmark are starting to see, and what we're really excited about is, um, what we were just talking about, this idea of this new architecture, this idea of sell the work, not the software that you have an SLA on outcomes, not, you know, reliability. You know, UX for the manager, not the worker, et cetera, right? This sort of fundamental change. And we're starting to see those, those ideas really percolate. Um, uh, a lot of it though is things that are, um, kind of just becoming available. You need, you do need better models for it. You probably need some fine tuning around it. Fine tuning available to startup is only just starting to happen, right? You could have been a savant and maybe take it open source models, but…

AI assessment note: “I think there are some obviously ones that are, that are not that already.”

Q No, I totally get you. Um, can I ask you, when it came to Lionchain, how did that deal go down?

A More than half of what we do is the first partner, you know, probably a third kind of an inception, um, uh, as one was exploring more of what was possible with things like sort of reacting the agent ideas and this idea that, um, every developer would, would become an AI developer versus just the ML wizardry, you know, Harrison, um, very definitely, uh, and, um, uh, but like without grandiosity, you know, put out an open source library with the beginnings of abstractions for, for, for making, Agentic behavior and now a lot more around retrieval and other things easier. Um, and we just started spending a lot of time together. Um, we probably, Probably three or so months, um, before we ended up actually investing. We, you know, talk about what was happening. We talk about his framework. We talk about, you know, sequence, strengths to be building now, et cetera, and just build that relationship. I never pressured him to say, you've got to start a company. There were some people who pressured him to say, you've got to start a company, but like, that's got to be a founder's calling. Um, uh, you know, you can't unfound once you found, like it's a real responsibility. And so he then said, Hey, I want, I want to go do this. And I said, great. Like let's partner. And it's been a, it's been a privileged journey since.

AI assessment note: “Probably three or so months, um, before we ended up actually investing.”

Q Did Fenton and Gurley give you any advice on joining?

A I wouldn't, Say there was, um, ever, and you know, in some sense we're always, um, as we go through it, like giving each other thoughts and feedback and, and, and nagging each other to be better, obviously. And so there wasn't, um, there wasn't sort of like some entry indoctrination or, you know, or whatever. In many ways it was, um, you be you, like you be, be, You totally kind of unfiltered and as authentic as you are, and the best version of you will be the best investor. And I think in many ways, when you get rid of structure, um, in this, there's no like level to be had or, you know, no, nothing like that. Um, the, the, just be yourself is in some sense the hardest, hardest thing.

AI assessment note: “In many ways it was, um, you be you”

Q Yeah. Uh, my favorite is actually, um, if I have more time, I would have written a shorter letter. I love that. What was the third one? Sorry, I cut you off there.

A The third one, um, would be this, um, again, The greatest gift of things you sort of carry with you as ways of being would be what I might call, um, impatiently patient. Again, sort of a little, a duality that almost shouldn't exist, um, uh, which is why not now? Why not yesterday? Like hyperactive all the time, but still a really high standard and really patient for the long term. And so that combination of like, Being fully present, like fully engaged, new, um, and, and, and, um, and active, but also like high standards and not rushed. And that duality, I think, um, he, he really lives. And, um, you know, we were having a mentor like that, um, early on who can show you that through that lived experience is a, is a really powerful thing to carry with you.

AI assessment note: “The third one, um, would be this, um, again... impatiently patient.”

Q You said about new variants there. The thing I do want to ask is, you know, obviously you led the Langchain deal, one of the hottest and most competitive deals, I think, when you led it, and then subsequent round. Um, my question to you is, And actually Harrison's. How did you first get into AI and LLMs when you did, which was considerably before everyone else actually?

A Yeah, so I was spending time sometime in twenty-twenty-two, probably early, like not that early in the grand scheme of things, right? Um, but I guess earlier than the marginal- Zeitgeist. We're captivated. You saw the first versions of things which were, um, uh, just sort of spitting out content, right? Um, but you started to see in mid-twenty-twenty-two or so, this idea that the, the models had some amount of reasoning Capacity. There's, there's two things I might point to in that one. One is they can do things like know which colors are like each other, which things are, the ability to do math. Um, now in some sense that's learned from the language and it's just statistics at the end of the day. Um, but the fact that those properties were kind of there was fascinating. And then you've got things like the react paper, which was sort of this reasoning and action and the model kind of giving it feedback to itself. And you started to see, at least I started to see when I saw those is like, There might be so much more here. It might be, um, a sort of new computing paradigm in a way. And there's more power in this than just sort of marketing copy. And you combine that with, um, another thing, another aspect that I felt could be possible and is proving out to be true is that ML had been in many ways, mostly kind of just like, um, reserved for ML wizardry. Right? It was sort of the, …

AI assessment note: “I was spending time sometime in twenty-twenty-two”

Q Well, speaking of like going on that journey, how do you bring the partnership along with you and your excitement in a deal? Do you tag teaming up at Benchmark where you say, hey, Chathan, let's get on this together. And then you discuss it as a partnership, but you and Chathan are the leads. How do you bring everyone with you in your conviction for a company?

A Yeah, we do. We do a lot of that. Like, hey, come, come, come join this. Hey, let's go meet together. There's ultimately one person who is, takes a board seat, right? Manifested that way. Um, but it is a, again, we're an equal partnership, so it's a whole firm commitment. No person is incentivized to, to, to do more, help more one thing versus another. A founder really gets the whole partnership. Um, but obviously they've, they have a principal relationship A dominant relationship with one person, but we're, we're constantly pulling each other and actually just before I came here, a founder who works with Bill Gurley, I'd spent some time with recently and sent me a text, like remembering something I'd said to him in a line because he was using it again and again. And so we're, you know, we're always pulling each other into current investments, future investments, um, sort of in a very fluid way.

AI assessment note: “Yeah, we do. We do a lot of that. Like, hey, come, come, come join this.”

Q Can I ask, you said before that the comparison to mobile, which many have made, many have said that actually this is similar to mobile and actually the incumbents will need to shift to it like they did to mobile. You said before maybe that's not the right analogy. Why do you think the comparison to mobile is not the right analogy?

A Shifts happen. We want to analogize and no analogy is going to be perfect in many ways. Um, I'll compare it more to the idea of what the internet enabled, the mobile enabled, But like both of those were also new distribution channels, and this is not a new distribution channel. Um, so both of them are fundamentally flawed in that way. I'd also argue that if you think about distribution is closer to the internet in that the internet was a distribution channel. It also catalyzed, um, the sort of buying, it catalyzed the, the idea of sort of buying an internet native solutions, right? You, you think back to the story of Microsoft and, and, um, 95 Bill Gates wrote sort of the Paul Harbour memo of like, we need an internet native Kind of like strategy to it, right? It was a catalyst for, um, consumption and sort of change within those organizations, more so than mobile. But none of them are perfect. The reason why I think mobile is, is a more flawed way to think about it is that mobile really was, um, the same sort of architecture and business model as, as what was before it. Um, the way we architected applications didn't really change. Um, the business models associated with them was pretty much the same. And those things can really lead to Um, disruption can lead to new startup competing. And most of the incumbents before mobile were the successes in mobile. What the new people th…

AI assessment note: “mobile really was, um, the same sort of architecture and business model”

Q So we mentioned kind of, uh, Benchmark there, uh, from me, when you made the move, I'm sure you had a lot of people call you and just give you advice. People obviously built an incredible set of friends, community around you. What advice did you get when you joined Benchmark?

A People in many ways, I spent a lot of time and not a lot of time, but a decent amount of time in venture. And so knew of, uh, different firms in many ways. Um, so I didn't really, uh, go and get advice on making the decision. The only person I really asked, um, was, uh, was my wife, which were, and I was like, are you really up for moving to California? Um, and both of our families are on the East coast and family's important to both of us. She'd grown up entirely on the East coast, didn't have friends as many ways out there and didn't have a life out there. And we wanted Children. So we set roots out here. And so, uh, the main question was, are you really up for that? We'd only been married in many ways, also like six months during the pandemic. Um, and she, she was fully up for it. And so, um, that was the, uh, that was the, the biggest question.

AI assessment note: “I didn't really, uh, go and get advice on making the decision.”

Q Just on that point, I'm always in awe of his ability to spot talent. When you look at your Chris Pikes, when you look at obviously you, but your Will Gabriks, ah, you know, even Vince now and Karim who've scaled incredibly well internally. What do you think makes Josh so good at talent detection on the investor side? Because I want to be as good as he is.

A Um, I think, I think the, the track record obviously is phenomenal in that regard. A big part, he talked about this recently on a podcast, you know, um, uh, in many ways is sort of probably, uh, uh, A combination of a really an appreciation of will over skill. Some people will hire for, um, proven ability. You, you know, you already have the knowledge or, or, or the like. And again, stuck to this idea of like, yes, you can. Um, and a belief in someone, uh, it's really a gift of a, a, a confidence in someone, someone's work ethic and determination. And yes, they have to have the mental capacity and the intellectual capacity and all those things. Right. But you, you can know that in someone. But the person who will, um, love it and is passionate about it, you know, I have six younger siblings, and sometimes I think about, you know, what, what lessons I'd like to potentially impart as I go sort of on them. And, and, and, uh, you think about these days, everyone talks about follow your passion. And I kind of believe that for sure, but I believe it for the reason, not where everyone else necessarily talks about it, but you should follow your passion because you'll just work harder. And if you think about compounding 10 extra hours a week, 50 hours a year, 500 hours, you know, not far into the future, um, you've just, you've, you've learned more, you've caught up, you've, you know, y…

AI assessment note: “A combination of a really an appreciation of will over skill.”

Q Do you worry about too much money too soon? I know they obviously subsequently raised another round pretty quickly after yours. I saw it and I was like, well done. That's all the cash. Do you worry about too much money too soon?

A Definitely. Um, uh, you know, we, we, we probably could have taken more, probably could take more now. And so, but just because you could say no to more doesn't mean you shouldn't say no to anything, you know, kind of thing. I think also about the person, right? Um, uh, some folks would let that all go to their head, would spend frivolously like, and so it's situation specific, right? The culture that has to be imbued in, in the company of which, um, more capital makes it more important to make sure you imbue that, but it, and it almost poses the foil and the tension, but it helps clarify it and, and, and, and, and you still need to keep focused on it. But if you force, Clarify it or strengthen it, um, is that it's all investment. It's got to be fruitful investment and not wasteful investment, right? I think you can run the risk, but I don't think it's definitive that that happens. And so, um, you know, at LangChain, obviously, like we have more capital than we need right now. Um, uh, but the bottleneck's been like hiring great people and, and the initiatives we want to work on and, and, and that's driving investment and pace of investment versus We have the capital, so let's, why not?

AI assessment note: “Definitely. Um, uh, you know, we, we, we probably could have taken more”

Q When did you get to know Figma? Because there was that intersection with the A when John O'Leary did the A, And it wasn't a live product. I mean, there was nothing, and it was kind of in this kind of desert light zone. The product was developing, but there was no, it wasn't released. Yeah.

A I spent time with Dylan, um, was fortunate to spend time with him, um, at the round that Kleiner ultimately did. Um, and then also again at, um, at when Sequoia invested. It was definitely the one I, um, am most mad at myself for. And my hesitation was, um, I didn't think designers was a terribly, Big opportunity, a big market. Like the N of designers is not, certainly relative to, let's say, engineers is tiny. If you go looked at most of your companies and got a head count of designers, it's a very small N. Um, at any later stage company, there's, I'm almost certain there'd be like more accountants than there are designers, right? And so if you're thinking about seed software, it's, it's not many. I'll think about the hesitation I had was that I didn't think designers was a Particularly large N. I think if you went and, uh, asked how many designers were in many of your companies, certainly relative to engineers would be a tiny fraction. Um, I think N of designers relative to accountants at some later stages of startups would be tiny.

AI assessment note: “I spent time with Dylan, um, was fortunate to spend time with him, um, at the round that Kleiner ultimately did.”

Q When did you get to know Figma? Because there was that intersection with the A when John O'Leary did the A, And it wasn't a live product. I mean, there was nothing, and it was kind of in this kind of desert light zone. The product was developing, but there was no, it wasn't released. Yeah.

A I spent time with Dylan, um, was fortunate to spend time with him, um, at the round that Kleiner ultimately did. Um, and then also again at, um, at when Sequoia invested. It was definitely the one I, um, am most mad at myself for. And my hesitation was, um, I didn't think designers was a terribly, Big opportunity, a big market. Like the N of designers is not, certainly relative to, let's say, engineers is tiny. If you go looked at most of your companies and got a head count of designers, it's a very small N. Um, at any later stage company, there's, I'm almost certain there'd be like more accountants than there are designers, right? And so if you're thinking about seed software, it's, it's not many. I'll think about the hesitation I had was that I didn't think designers was a Particularly large N. I think if you went and, uh, asked how many designers were in many of your companies, certainly relative to engineers would be a tiny fraction. Um, I think N of designers relative to accountants at some later stages of startups would be tiny.

AI assessment note: “I spent time with Dylan... at the round that Kleiner ultimately did.”

Q co-pilot for anyone in the legal industry, but then through the proprietary data acquisition that they get from, uh, you know, different user inputs and different user modeling, they're able to then sell the work over time, or they say they will be able to. Is it a transitionary period where you do co-pilot to then be able to sell the work, or do you think we'll see a shift?

A I think there's so much holistically in a company that gets wrapped up around a business model, right? And if you have a co-pilot business or your product is wrapped up around that way, your distribution model's wrapped up that way, sales education's wrapped up around it. And that's why I think business model, which encapsulates product and distribution and then markets, if you can be orthogonal to that, to an incumbent strategy there, It's an, and obviously you have to be correct. It is so much being orthogonal to this idea that you would have a UX, not for a worker. You'd have a UX for a manager. The idea that you'd sell work and outcomes, not software, right? Like the idea that you're Salesforce, right? Selling a cloud subscription managed by us versus Siebel selling an on-premise license. You install, you manage, you get passion for like that's orthogonal, right? Um, there's so much wrapped up in, in, in the arc fundamental business As in company architectures and all the systems internally and all their operatings around that, that it's really hard to just change, right? That's why it ends up being asymmetric competition. It's not to say they can't and they won't. It's a lot more painful of work to be done of, of organizational change to enact. And so I think any, a startup's opportunity is to build their strength and, and, and build from scratch that strength in a way tha…

AI assessment note: “all their operatings around that, that it's really hard to just change”