Striveworks CEO Jim Rebesco on drone swarms, AI for warfighters, and the end of air superiority

Jim Rebesco is the co-founder and CEO of Striveworks, an AI ops company deployed across the US Army, Navy, and combatant commands. In April, Striveworks announced a $70M enterprise agreement with the Pentagon, giving nearly a million defense personnel access to its platform. Jim was previously a partner at Virtu Financial, a pioneering high-frequency trading firm. In this episode of Summation, Jim and Auren discuss: why the real question in AI warfare is who pushes the buttonthe case that aut...

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  That's cool that you think you're a smart guy. Turns out there's a lot of smart people in the world and most of them were not working with you.

• 0:06Auren Hoffman?

  My guest today is Jim Rebesco. Jim is the co-founder and CEO of Strive Works, an AI ops company deployed across the US Army, the Navy, combat commands, et cetera.

• 0:16Speaker 00

  There's a lot of smart people doing a lot of the same thing. And like you constantly like, hey, you built a cool algorithm? Cool. I built a cool algorithm. And like now we've like, we're in this kind of like a little bit of arms race.

• 0:29Auren Hoffman?

  Hello, fellow data nerds. My guest today is Jim Rebesco. Jim is the co-founder and CEO of Strive Works, an AI ops company deployed across the US Army, the Navy, combat commands, et cetera. In April, they announced a 70 million enterprise agreement with the Pentagon, giving almost a million defense personnel access to their platform. Jim was previously a partner at Virtu Financial, a pioneering high-frequency trading firm. Jim, welcome to Summation.

• 0:59Speaker 00

  Thanks for having me, Orin.

• 1:00Auren Hoffman?

  And full disclosure, Flex Capital, where I am a general partner, is a shareholder and investor, a very happy one in Strive Works. So I just want to let everybody know I'm a huge fan of what you're doing.

• 1:11Speaker 00

  And for the founders out there, we're super happy to have them as part of the team too. So there's your endorsement in both directions.

• 1:18Auren Hoffman?

  That's the clip. I'm going to replay that clip like 100 times as well. So now a lot of people think autonomous weapons are like the stuff of the future. What is operational now?

• 1:30Speaker 00

  Yeah, I mean, that's the $64,000 other question, right? And you know, there's that like that old Tao saying, you know, like those who know do not speak and those who speak do not know. And like, there's a lot of that going on in AI right now. So, you know, I think, you know, kind of with with the big caveat upfront, which is that like, you know, there is a kind of a security element to this. And might we won't be able to hit specifics. I think like, You know, where I'd start that question is, is like, you gotta, you gotta think about how like, you know, autonomy has existed on this continuum for ages, right? You know, like, well, I don't know if like, you consider a heat-seeking missile or a homing torpedo an autonomous system or not. Maybe, maybe not a little bit.

• 2:09Auren Hoffman?

  I kinda do, right? You, you shoot it off and it's, it might not hit the intended target because something else might come in. And you've seen the movies where they like, put out some other thing as a defense or like, it comes back and it hits the person who shot it or something, right?

• 2:24Speaker 00

  I think like, what makes it like, an interesting question, why people are like, maybe a little bit more focused on it today, right? Is that all those things exist in this paradigm, like people pressed a button and then the autonomy happened, right? And, and now we know, you know, whether it's, you know, OpenAI or Claude or whatever you like, there's stuff way left of that where people are chatting with agents and talking and doing stuff. And so like, kinda like the question that people are really centralizing is, is like, who pushes the button and when? And that's kind of, I think, like the core, what people are really like asking when they're asked, I'm like, hey, what's, what's autonomous today? Is that fair?

• 2:56Auren Hoffman?

  I know in like, the Ukraine war theater right now, a lot of, a lot of what's going on is about jamming the drones. And then, of course, they have like, these long tethers that go in, these essentially these like, 10 kilometer wires, so they can't be jammed or, you know, these other things. But of course, the, the main reason they wanna jam it is because they wanna jam the operator. If they were autonomous, then you wouldn't necessarily, jamming it wouldn't even potentially work. Or I don't know, maybe it would, or, you know, maybe spoofing the, the GPS would or would not work. But if it was purely using a camera to making an autonomous decision, okay, is this an enemy tank or not?

• 3:35Auren Hoffman?

  And that was like, its only goal was to take down an enemy tank, then you think like, okay, then like, this whole theater would change. And maybe that's already happening.

• 3:43Speaker 00

  Yeah, yeah, I think you're seeing that.

• 3:45Auren Hoffman?

  There's a continuum as a technologist and a continuum as like, I guess, an ethicist, like, how do you, how do you think about all these things?

• 3:51Speaker 00

  The metaphor of the button, right? I think is like, this kinda like, core question of like, you know, what, what I think like, in defense, people call it command and control, right? And I think one of the funny things about this debate right now is, is that like, you know, we're a little bit like unlearning this kind of mental paradigm that happened during the, the global war on terror, like the GWOT, right? So it's like, if you zoom out, like the sweep of like total history, you know, way left of that, you know, like your ability to kind of exercise command and control on a battlefield as a commander was like super limited by bandwidth. And that might've even been like, how fast can somebody on a horseback or on foot run? Like, how many bits per second can like a bunch of flags wave or whatever? Then in GWOT, you had this like unique confluence of technology where you had all these, you

• 4:35Speaker 00

  know, camera feeds and drones and radios and whatnot. And like a commander could like very plausibly like see and potentially even talk to individual soldiers on a battlefield in real time. So what's the point of that? The point of that is, is that as we think about this exponentiation of drones and just stuff on the battlefield, you know, people are kind of like, hey, wait a minute. Like, we're not like, are we going to be able to have like independent control of all these assets? And the answer is no, you know, but the, the metaphor or the visual I want to give you though, or I put to you a little bit is like, if you think about the sweep of history, like only for this really weird 20 year period in the middle, did we ever really have this ability to like plausibly say that like people can communicate and

• 5:20Speaker 00

  command every little asset on a battlefield? Like before that you couldn't and after that you won't. But I mean, I'm curious how you, how you think about that, if that resonates with you.

• 5:28Auren Hoffman?

  Yeah, no, I think that makes sense too. Well, I mean, one, one thing I've been thinking about a little bit is like the, the mine analogy, whether it's a land mine or a sea mine, the problem with mines is like, they could go off like many, many years later. And so, okay, maybe you mine a battlefield or you mine a ceiling and then, okay, well, okay, how do you like clear it? But I guess with a autonomous drone, you could potentially put it in a battlefield and then it could like self-destruct after a year or something like that. Or so, you know, obviously it could, it could, it could power itself up from solar for that year and then it could, it could decide at the end of the year, just blow itself up or, or I don't know, something like put itself to sleep or something like that. So it might be better than a mine, even though it kind of has that same kind of capability.

• 6:13Auren Hoffman?

  You could just leave it in a place and then it could, anytime it saw an enemy, it could just attack it.

• 6:18Speaker 00

  People tend to think about, you know, like AI and things is like, oh, like what happens when it's autonomous and it can go wrong?

• 6:26Speaker 00

  Which is very true and very, very important. And there's also like to your point, like there's a type one and a type two error, which is like, you know, does this give me the ability to actually like make these more safe, to render them safe, to exit, to exercise kind of more control over them in an asynchronous fashion as well? And, and to your point, you know, I don't like, you know, God knows, like, there's so many, like, old battlefields in the world, there's still mines on them, and it's terrible.

• 6:51Auren Hoffman?

  Yeah, and I mean, you could even say, okay, if there is kids, don't do that. So you can, you can set it in a safe way that, you know, is only doing a certain thing. And of course, like, the enemy may take advantage of it, but you can still, if they know what your codes were. But you can be, you know, potentially, and you could potentially even like, potentially shoot updates to these things. So these things go lay dormant in the battlefields for a very long time, I imagine. And then you update and you might lose, you know, you might lose the ability to update some of them or something over time. I don't know how it works.

• 7:23Speaker 00

  Like many things in technology gets complicated, you know? But, but at the end of the day, I think like the big story there being that like, this actually gives you the ability to control, to control things at the scale that modern battlefield is going to demand. It's a force multiplier is like the important through line on all this.

• 7:38Auren Hoffman?

  I thought your GWOT analogy was pretty interesting. And in some ways, to me, GWOT was like both two things, both two diametrically opposed things happening at once. One was like, it did have a lot more command and control, as you were saying, but also they pushed a lot of decision out to the edge. And you saw these like very mobile kind of special forces teams, sometimes, you know, six to 24 people who could operate in extremely, extremely good ways. Like, how, how does this kind of coalesce? Does it, does it also allow those extremes or just maybe like that middle person is less important?

• 8:17Speaker 00

  It's an interesting question, you know, like, I'll, I'll give you maybe like the contrarian take here, which is that I think like a lot of these technologies are actually going to favor larger nation state players, you know, like if, if kind of like your point on GWOT was like, You know, you saw this kind of like metaphorical aircraft carrier versus insurgent and like the initial tape that these technologies like can make individuals themselves like more lethal and more capable of directing a tunnel lethality. You know, I think that's, I think that's fair, but the kind of counterpoint I'll give you would be like, kind of what you're seeing now in intelligence, you know, with like ubiquitous, like mass technical surveillance. Like, you know, it used to be like you could do very traditional human intelligence, you know, like do spy stuff,

• 9:01Speaker 00

  talk to a guy, you know, steal a blueprint. I don't know. Like, that's not my world. I don't, I don't pretend in that way, you know, but now if you're a nation state and you have the resources of a nation state, it's very plausible and very scalable to hold, you know, every foreign national, everybody, you know, in some level of kind of custody and mass technical surveillance on them that makes those traditional things harder. And I think-

• 9:25Auren Hoffman?

  I would imagine that even, even in a country as big as China, it would be extremely difficult for a spy to operate in a, in a place like that where, you know, maybe 20, 30 years ago, it'd be a lot easier.

• 9:40Speaker 00

  I guess, and then what I'd put to you is that in a world where, you know, we're in a world where you've got this like massive amounts of kind of attributable mass that are entered into the battlefield, you know, there's going to be kind of, you know, kind of exponential or like network effects to the people who can like not just have, you know, like, it's great that a human being can operate one drone that can carry a payload or operate three drones and that matters. But what's going to like really start to tell is like, how do you mass a hundred, a thousand, 10,000 to a point where you can actually overcome, you know, kind of a layered defense system too. And I think like the complexity in coordinating and managing that sort of stuff is actually going to favor scale players more than it is the, you know, the insurgent for the terrorist.

• 10:25Auren Hoffman?

  How do you think of like air superiority? Cause the Ukraine kind of Russia conflict is kind of interesting where like, it seems like Russia has air superiority at, I don't know, 10,000 feet and above, and Ukraine has air superiority at 5,000 feet and below.

• 10:41Auren Hoffman?

  Or something. And so it's like, it's like a weird thing where before it at least the GWAT either had it or you don't like now it's like, and same thing like Israel Hezbollah. Seems like Israel has it at the high level where Hezbollah may have it at the lower like lower level. So how does one kind of think about these things?

• 10:58Speaker 00

  When you go back five or 10 years and you had like, you know, kind of planners who are starting to see this and starting to talk about, and that's why they were like, you know, it's not about your superiority. It's about like, you know, access and denial, you know, it's not like, can I control the entire airspace? It's like, can I create a little tube to move what I want to through the tube and then the tube closes behind me or whatever. The kind of like analogy then like in my head, because I'm a technologist and not an aviator, right? Is that like the evolution you saw in like cybersecurity where there used to be this kind of like castle moat framework, you know, where it's like, I've got a firewall. I've got a VPN, everything in network is safe and trusted, everything out is not. And that led to like huge fricking problems when that got breached.

• 11:38Speaker 00

  I think you're going to see the same thing here where like it's like the concept of like air superiority, like can I kind of draw a curtain? And like stuff on that side is yours, stuff on this side is mine is already obsolete. And now it's just more of a question of like, how can I

• 11:54Speaker 00

  generate ephemeral access to kind of do what I need to do and then, you know, presumably it'll go away once I'm done. So, so yeah, like it, it again, like the, the point of it is, is that like, when you move into that, now you can, you're continuing to push autonomy down the stack. Because whether it's like a 155 artillery shell or a drone or an F 35 or whatever else, like need to manage all the kind of network stuff to let that thing fly and then close it up behind it when it's, when it's done flying through. And it's just not stuff that human beings are going to be able to do.

• 12:25Auren Hoffman?

  Now you sell into the Department of War, you sell into the intelligence agencies. These are. At least historically have been not maybe the best buyers of technology, haven't moved the fastest. How are they getting like, since you started, how is it getting better?

• 12:44Speaker 00

  There's a ton of good news. And I mean, especially kind of in in the AI space where, you know, eight years ago, you know, your kind of first conversation was maybe explaining what AI is and explaining why it's relevant. Everybody's already there today. But I think like, the biggest thing that's changed is is that like, you've got such a better, you know, understanding of and appreciation for like incentives and business models on both sides of the table. Like, it's not that long ago, you know, where you had folks saying, like, you know, I think, I think, like, you know, software should be bought on a on a labor hour basis with a fixed profit margin. And you're like, hey, like, that is fundamental, like, regardless of your opinion about it, like, it's just so fundamentally incongruous

• 13:27Speaker 00

  to how software is actually built, you know, like, I take a giant pile of money from wonderful investors, you know, you you use it, and then hopefully, you have a product with zero marginal cost that you can distribute afterward. Like, that's the business model of software.

• 13:41Auren Hoffman?

  And you're going to hole at first because it's you have to you have to create it and go in the hole. And it's not ideally, it's not so customizable to one client that you can't reach. So because then then it's a bad model to go in the hole.

• 13:55Speaker 00

  So yes, I mean, if I plot a little graph over time of like, the number of conversations I've had with folks on the government side who, you know, they're like, hey, explain to me how, explain to me how your business model operates. Explain to me, you know, like, what could we do in terms of buying to be more congruent with how you want to operate? Like, and it's such a good, it's such a good framework, because like, it's not like people are saying like, oh, you know, like, you need to, you know, you need to give contracts to a certain segment to startups to this to that or whatever. Like, that's entitlement. That's not relevant. Like, what's it's interesting, you have to earn your business. But like, what's interesting here is they're like, I want to know, like, what buying models align

• 14:36Speaker 00

  to your business models and vice versa, because maybe we're not happy with the business models that have won in this space previously.

• 14:42Auren Hoffman?

  You guys are now like pretty far along, but if you're giving advice to people who are just kind of selling into the IC or selling into the Department of War, what advice would you give them to somehow hack the system to get their product moving?

• 15:00Speaker 00

  First is, is like you got to understand that like the people who use your thing, the people who buy your thing, and the people who create the requirement to buy your thing are three different people. They're all very important and you have to be engaged with all three of those people.

• 15:15Auren Hoffman?

  Typically, when you're selling software, ideally, all those three are the same. Every, only one time, there's like a procurement person or a legal person you have to kind of engage with, which makes sense. But usually, there's like a, kind of like a, the decision maker is the key user of the product.

• 15:33Speaker 00

  Exactly. And like at a, at a minimum, usually like the user is unified with the budget. Like you're asking someone to spend their budget to use your product. And, you know, most often, the user and the budget are actually not unified in the government space, you know, which is, which is a problem.

• 15:52Auren Hoffman?

  One of the things in Department of War, which I think is different than the IC, is that people are constantly rotating. So you engage with someone today, like a person, you're kind of getting them up to speed. And like, I don't know, a year later, they're like rotating to new thing, you got to get, engage to. Whereas IC, they may have a little bit more staying power where they are, they're also still rotating. Like, how do you deal with that rotation of person? You, you built this like strong relationship with someone, it took them a long time to educate them, and they also have to trust you, et cetera. And then they're just like, I don't know, they, they moved to Germany or something.

• 16:27Speaker 00

  It's a great question to ask. If you figure it out, let me know.

• 16:33Auren Hoffman?

  It's just hard, right? I mean, it does make, it makes that whole decision harder. And also hard for that person to like make, because they're, they're like making a decision, they're not even going to be there to see it through too, which is also a little weird and makes sense. Like, it's just, it's just harder. Though I guess in some ways, it allows them to think beyond themselves, which is also kind of a, there's a feature in. A bug in it, I guess.

• 16:54Speaker 00

  Yeah, yeah, I mean, I think I think that's right. And I think maybe like the biggest bug is is just that like not all time scales like run at the same clock. You know, like it's a cliche, but it's a good one. Now, you know, like famously, you know, like when when the Navy had to develop nuclear reactors for its ships, you know, they put a guy Admiral Rick over in charge of that. And they kept him there for years, like completely out of that two year typical rotational cycle, because there's an understanding that like. This is gonna take a continuity of leadership and a continuity of effort to kind of hit this moonshot. And so I think maybe that's like, I think the takeaway there is like not. You got you have to slow down everybody's clocks, but you do have to have a pretty sober idea of like, hey, if we're gonna really transform.

• 17:35Speaker 00

  This big thing, that's gonna maybe require a little bit of a different model and just kind of like roll in two people every year. Or see if you could person every two years.

• 17:43Auren Hoffman?

  When you think of these kind of like. Archetypes, you really got. Stuff done in the military over the years, Rick over General Groves, who who built the Pentagon and then built the Manhattan Project like. You know, it which is like just insane that he's able to kind of like do all this stuff like. I know you're kind of a student of history, like what were those people? Because they still had to work, there was still a big bureaucracy. Maybe it wasn't as big as today. Back then, there's still big bureaucracy, all these stakeholders, like, how were they able to do it? Or how was the nation able to empower those people to do it?

• 18:18Speaker 00

  You know, one of the questions I've always had is like, how have we got here? You know, like where we have this kind of. You know, acquisition system that, you know, maybe doesn't move as fast as as we all like would like it to, right? It's not about the people, it's about the system. And, you know, and you compare it to the stuff you mentioned, right? And the reality is, is that I think like a lot of. The kind of current acquisition framework is actually like new new laws and new relatively new laws and new policies that have kind of come into place. You know. Since the 80s.

• 18:49Auren Hoffman?

  Like to prevent fraud, prevent overspending, like those types of things. Yeah, okay.

• 18:54Speaker 00

  Exactly.

• 18:55Auren Hoffman?

  They were like, well, intentions.

• 18:56Speaker 00

  Yeah. Yeah, yeah, it's like, again, like type one and type two errors, right? Like I would submit to you that's like. It's extremely hard to like sell your buddy a $400 screwdriver, you know, in the DOD now, like that kind of fraud is. Like I would think virtually impossible or very difficult at least, but. You have the type two error now of like, it's extremely, extremely difficult to move at the pace of technology in a lot of places.

• 19:24Auren Hoffman?

  I think you may have told me this, or I forgot who told me this, but like, someone was saying that they were talking to procurement officer and they said that it costs them $200,000 to procure anything. So it's just like really hard to actually procure things of low cost because like, okay, well, why would you spend $200,000? It has to be at least $2 million to even make sense to to kind of spend it on that.

• 19:49Speaker 00

  For the margin. Yeah, that's right. That's right. Yeah.

• 19:52Auren Hoffman?

  That's crazy, by the way, like that it costs that much to just procure anything, right? Just with all the paperwork, which kind of makes sense. I mean, if you have to go through like eight different people and triplicate this and have a lawyer review this and, you know, I can see how those costs like get in there really quickly.

• 20:06Speaker 00

  Well, that's it. It's like, and so you get this kind of like self reinforcing, like kind of do loop, you know, where the government is like rightfully concerned like, hey, we want more flexibility in buying decisions. We want more competition. We don't want vendor lock. Well, the way to do that, in my opinion, is is like, you, you know, you can't just say like, I go faster. You can't just say like, I ignore, ignore this stuff if you're buying in these like billion dollar chunks. You just, you just can't like a billion dollars is a lot of money. You got to actually like, do your own work to spend it, right? You know, the point is, is like, but you know, if you're buying 10, something that's $10 out of petty cash, would you need the same level of like

• 20:45Speaker 00

  stringentness and rigor? No. And I'll tell you that like the guy who's selling the candy bar is probably doing it for a lot lower margins than the person who's like selling you the color printer that you buy once every 10 years.

• 20:56Auren Hoffman?

  Yeah. And you hear these kind of crazy stories of like soldiers have to like, you know, go to the local Dick's Sporting Goods to buy themselves the sunglasses they need or something like that.

• 21:07Speaker 00

  Yeah. Yeah, that's right.

• 21:09Auren Hoffman?

  No, I mean, okay, there's like, it's hard to even know whose fault it is, right? It's like, okay, we've got like the system, we've got the congressional rules, we've just like the bureaucracy, we have these like large primes that are like boxing people out and stuff like that. Is it kind of like we're all at fun and it's like the public? It's like, it's just like a collective. It doesn't seem like there's like one boogeyman.

• 21:33Speaker 00

  You know, if it's everybody's fault, it's nobody's fault.

• 21:36Auren Hoffman?

  Okay, so, and you guys are obviously AI company, you're using all these different AI models. There's the kind of like frontier models that are out there. There are like these amazing open source models that are kind of that are out there as well. Some of those, you may have to, some of those could even run locally on different things. Like, how are you thinking about how you're using this kind of like coalition of models to, to help your customers?

• 22:05Speaker 00

  The analogy I'll give you, or maybe like the, the kind of visual metaphor would be like, I think it's a lot like chips, right? Like where there's like, there's this ravenous demand for the frontier chips, whatever they are now, like two nanometer process or three nanometer process. But like your blender, you know, or your car radios, back in, back in car radios existed. Um, you know, like those have microprocessors too, and they don't need, they don't need like frontier as chips. They don't need two nanometer chips.

• 22:32Auren Hoffman?

  Even, even your phone, I mean, obviously has tons of chips in there, some of frontier chips, but they have a lot of chips, even your phone, to do lots of different types of things, right? You know, so.

• 22:41Speaker 00

  The model ecosystem is going to end up looking a lot like that too, where there's always going to be use cases where you're going to want that hyper frontier model and then there's going to be other stuff. I mean, you know, you're seeing it already where you're on people like, hey, does this need to be a model or can this just be like a linear regression, you know? And like, there's plenty of cases where like a linear regression is the right answer.

• 23:01Auren Hoffman?

  Only have a billion parameters or, you know, whatever, you know?

• 23:05Speaker 00

  Incidentally, like both with some of our, our users and a lot of other folks in the space, like you're even seeing that kind of like managed and obfuscated, if you will, behind the user interface where, you know, if, if you think you're, you're talking to me through like a chat interface and you're like, hey, give me a. Deep analysis of the history of semiconductor lit though, I might need to spin up the big gun model for that. And then if you're like, I've got salmon, carrots and kale in my fridge, give me a recipe. Like, I don't need to spend 50 billion tokens to answer that query for you, even though it's coming in through the same interface. And I'm going to save you money and time and I'm going to save myself a lot of inference costs by just routing that to a lighter weight model. So, you know, I think at the end of the day, we've talked about this, it's dry works a lot.

• 23:50Speaker 00

  Like the, the phrase we used back in the day was like disappearing ML ops, you know? And I think that's more true now than ever in the sense that users, humans are just going to have a lot of like very kind of like standard, uh, familiar interfaces and all the like, in the same way that like routing a packet through the internet is this hyper complex process. And I'll just, I won't give any more details because otherwise like somebody from like Akamai will like hear this and get really, really spun up that Jim doesn't really understand that process. But the point being is like, you get, you got browser, you can type a URL in there and like everything else is abstracted away from the end user. Like that's where AI is going very, very clearly, like on a complexity, none of it touches a user.

• 24:32Auren Hoffman?

  There's many different reasons why you'd want to have different models, right? One would be cost, as you said, one might be, um, like privacy things like you, you don't want to send this out to the server for various reasons, um, either because like, you know, you know, it could get intercepted or other things like that. Another might be like latency, like the closer it is, the better. And, and then, and then of course, if you may have it on like a, you know, on a drone. So you may want something like super local on the drone. Of course it's gotta be, you know, you may only have like one powerful chip on that drone and you, or less, you know, maybe a few less powerful chips and then you have to kind of make it all work.

• 25:16Speaker 00

  One of the fun things about doing this after doing high frequency trading for a while is that I feel like I got a little bit of a, a sneak preview in terms of like how this is going to play out based on how finance electronified. And in more of the aspects of that is just exactly this where like when people were arguing like, Is it edge? Is it cloud? You know, it's like, just like you see in HFT, it's both. You know, there's your point. There's some things where, where you have to hyper, hyper, hyper optimize for latency. And you're going to push compute as far to the edge as possible. All the times you have to optimize for bandwidth, you know, and, and, and all those sorts of things. So.

• 25:53Auren Hoffman?

  Well, let's talk about high frequency trade, because I do think there's a lot of analogies here. Like, first of all, like, I think everyone kind of like knows about it, but I don't think very many people understand like, how the heck does it actually work? Like, how does one make money in high frequency trading?

• 26:08Speaker 00

  The core of like what high frequency trading is, first of all, is like, it's a public utility. And, and what I mean by that is, is like, you're providing people with a service, which is liquidity, and you're charging them a fee for it.

• 26:20Auren Hoffman?

  So you would say like what Citadel does is, is a high frequency trading, they're market making essentially.

• 26:25Speaker 00

  When I talk about like HFT and what we did at Ver to, the core of that is market making. And, and market making is like, you're providing liquidity, like someone, someone else wants to trade a certain amount now, and you're providing that to them as a service. And the way you're making money typically is that, that spread or that difference between the bid, what people are willing to buy at, and the offer, what people are willing to sell at, you know, for like, hey, people buy at nine, sell at 10, that's kind of the core of how you generate a return for that.

• 26:54Auren Hoffman?

  When you're buying something, you have to, in some way, you have to be smart enough to know that you can sell it for at least a little bit more than you bought it for very quickly. And you're not like holding on to this inventory. If somebody wants to sell, you have to somehow know you could just buy it really quickly and sell it and make that difference too, right?

• 27:14Speaker 00

  It shifts the point of certainty, though, from like, the way you kind of framed it was like, I know that I can make a profit. And the way I'd actually like flip it would be like, you know that you probably won't lose money. Like, if you look at the statistics of how like electronic market makers perform, like their average profit is actually like a fraction, a tiny fraction of the bid offer spread. And like what that's aggregated from is like. A lot of trades where you basically buy, you do not make a profit like the terminology like you scratch or like I buy it at 10 now and 10 seconds later I have to sell it back at 10. A lot of those trades mixed in with a very small number of trades where you make a profit.

• 27:56Auren Hoffman?

  And every once in a while, I assume you have losses too. And then hopefully at the end of the day, but ideally at the, I mean, I know nowadays like you people, it's 24 hour trading, but ideally, if you think of it as a nine to five kind of trading apparatus or whatever, ideally at the end of the day, you're not holding any inventory, right?

• 28:14Speaker 00

  Generally speaking, no, it's like, like, you know, it's probably a little bit more of like, you're not holding risk, you know, or you're offset there. Like is if you're, if you're holding, you know, the S and P 500 ETF and the 500 component stocks, do you have any risk there? A super, super tiny amount. If you're holding the ETF against the future or you have any risk there, maybe a little bit more. You know, if you're holding Apple stock hedged against Google, are you holding risk more than the first two scenarios? And you can kind of think of that layer.

• 28:44Auren Hoffman?

  I mean, obviously these things like you could blow up or whatever. I mean, obviously, if you're doing it, it also depends if you're how much debt you have when you're doing this or how much you're just using your own cash and stuff, right?

• 28:56Speaker 00

  Well, I mean, yeah, and you know, and I think like, it's kind of obvious as you think about this, the structure of these sorts of things. Like if we're saying, hey, I'm doing 100 trades, 90 of them neither make nor lose money. And then if seven of them make money and three of them lose money, like net plus four on 100, that's a heck of a good day. Like if you have that kind of mental construct, you can start to understand how like the, the leverage and the trading capital you need to make that happen is not revealed. Like you need, you need cap to make that efficient. Yeah.

• 29:27Auren Hoffman?

  I know a lot of folks who are like trying to provide different types of liquidity and you know, the more maybe Susquehanna or somebody are a little bit more famous for doing that. And of course you can, and you have all these different exchanges, the Kausi's, Polymarket, Coinbase. Now Facebook may have an exchange of all these different types of things. Like if you were kind of designing. Something today with like prediction markets in mind, or I guess crypto would be kind of similar where you have all these different marketplaces. How would you be thinking about these problems?

• 29:57Speaker 00

  I left HFT like kind of around the time that the big electronic crypto markets, the coin bases and whatnot of the world started, you know, kicking off and you know, like one of the things I was hearing kind of from folks that were still in the space was that, um, how, how wide the, the spreads were on these crypto markets.

• 30:17Auren Hoffman?

  It's like the Korean market might be different than the Japanese market or something and.

• 30:21Speaker 00

  So before that, just when we go back to thinking about like the stock markets and the future markets, like I gotta put a plug in, which is just like, it is, if you think about it, it is shocking how tight and how efficient those markets are that like you can, you know, you like, we can, we could log into our phones and trade, you know, a hundred shares of the S and P 500, pretty much blind, pretty much knowing that regardless of where we traded it, we would get like, to like a thousandth of a penny, like the fair price of the underlying 500 stocks. Like that's a freaking miracle.

• 30:55Auren Hoffman?

  And that's because there are all these market makers essentially who are making it efficient, right?

• 31:00Speaker 00

  Crypto like has gotten a lot tighter and a lot more efficient, but definitely wasn't like that back in the day. And like with these prediction markets, I just, you know, I, I would imagine that right now, the, the ability for, you know, kind of mechanical arbitrage is, like those kind of those earlier ones.

• 31:17Auren Hoffman?

  Yeah, I imagine though, it's like, in some ways, if there's a ticker and it's very, very clear what you're trading, it's a lot easier to, to understand the arbitrage opportunity. In a prediction market, if you have like something on Calci about, I don't know, the Lakers and you have another one on polymarket about the Lakers, like they actually could be like slightly different. It might look the same, but have a slightly different, and then it has slightly different risk things that, so you have to kind of like understand the underlying ticker and, and it might not make sense to understand it if there's only $200,000 being traded on that particular name or something.

• 31:53Speaker 00

  What a great use case for, for an AI agent to read those two contracts and understand the economic.

• 31:59Auren Hoffman?

  Yeah, really, really dive in and try to understand the very, very weird nuances of it as well. Personally, I was doing like some, a lot of trading in like, in like. January, February, March of this year, I'd say by like by March, everything had like tightened. And it seemed like a lot of the ARPS had, at least for me, I'm sure some was much smarter than me, but like at least doing like the simplest things that you could do in January, February, you could make some consistent money. And I, I thought like you'd have to be like a real high frequency kind of thing to, to make some money, uh, starting like in March timeframe.

• 32:34Speaker 00

  Know that progression from like, I can manually click the ARB to like, I need an FPGA.

• 32:41Auren Hoffman?

  Obviously you, you started your career in high frequency trading was super successful. Then you moved to this kind of like ML ops for national security. How, what, what are the similarities? What did you learn from the first one that is applicable to the second?

• 32:57Speaker 00

  There were two big similarities that mattered. You know, the first one was, was that like, the world was always changing, you know, the data scientist and you would say the data was non-stationary. But like what that really meant to was that like every single trading day, you know, you were going to be presented with new data, data that maybe you'd never seen before. And your ability to like detect that, react and respond mattered a lot more than the kind of baseline quality of the algorithm, you know? And then the second aspect being that like, you know what? Like, that's cool that you think you're a smart guy. Turns out there's a lot of smart people in the world and most of them were not working with you. And so like a kind of very adversarial environment where there's a lot of smart people doing a lot of the same thing. And like you constantly like, hey,

• 33:39Speaker 00

  you built a cool algorithm? Cool. I built a cool algorithm. And like now we've like, we're in this kind of like a little bit arms race. Those two dynamics were very, very crisp in the kind of finance world. And I think, you know, again, I don't want to make a facile or superficial metaphor, but I think it's obvious to see how those two dynamics are, are central to what's going on in national security too.

• 34:02Auren Hoffman?

  In some ways, if you think of like the high frequency trading world, like there's always these super smart people on the other side, many, many of them. And they're constantly updating their systems on a regular basis, could be daily, could be hourly, to respond. And they could start to see where they're winning and losing and how things are trending, and they could try to be updating. And then, of course, you have to see what they're doing and respond. And some ways, if you think of like Russia, Ukraine, I mean, some ways it's different because like, okay, you don't have as many players at once in the system as you might have, but they're like constantly updating and then you have to constantly be responding. And I would say if like I was the army, like it would be very hard for me to like pick

• 34:46Auren Hoffman?

  a vendor because I wouldn't know, you know, we're not necessarily like actively in a war right now, we're constantly not, how many, you know, and it's like, okay, we really have to respond in a war, you may have to respond on a daily basis, whereas in, you know, in, in here, we're like thinking of it in a five to 10 year lifetime, like how does one work with the DOW to kind of show them, okay, we're fast moving and we can actually help you move really quickly?

• 35:13Speaker 00

  I'll disagree with a little bit of your premise, just in that, you know, between, you know, the global war on terror, the conflict in Eastern Arab, you know, what's going on in the Middle East right now, Venezuela, probably forgetting a couple here, like you actually like have a lot of opportunities to demonstrate in the real world what's what works, what doesn't.

• 35:37Auren Hoffman?

  I know that some of your talent when you started came from Virtu, and so these are people you worked with before, the high frequency trading world. How do you think about bringing these like amazing engineers into the company, convincing them that this is a valuable thing, getting them like, how do you think about a whole kind of ecosystem?

• 36:00Speaker 00

  Yeah, I think like the first part of it is, is just like it's, it's always like super humbling because like the folks that I get to work with here at Strive are like Best people I've ever had the opportunity to work with in my career and like the fact that like this team's come together is just like it's it's what motivates me and excites me every day and you know and if if you asked like if the question is is like how that team they come together like part of the headline is is like heck of I know but I'm not going to ask too many questions and risk spoiling it but but you know beyond that there's that old joke you know that we were promised flying cars we got 140 characters I think like one of the most exciting things about what's going on in kind of Silicon Valley in kind

• 36:42Speaker 00

  of startups writ large is is that you know maybe if we look back a little while ago it would have felt like most of the energy is was around like solving the problems of 24 year old Stanford graduates like how do I get food delivery and how to go to my car valet and how do I get my laundry done and like now when you look at where venture capital money is going to is going to defense is going to quantum it's going to building power and infrastructure like venture capital is funding more like fusion reactors I think than they are dog walking apps in the last couple of years and I think that's a frickin awesome thing and so kind of within that lens like insofar as that like you know in our own in

• 37:25Speaker 00

  our own way like we're trying like strive is trying to contribute to something that's a lot bigger than ourselves and hopefully like a lot more lasting that's an important that you know that's an important thing like you know none of us the you know we've got the time we've got on that we've got on this earth and like we're not really going to be able to negotiate more on that on that particular topic and like spend it on big problems spend it on things that matter and that's a huge part of that story.

• 37:48Auren Hoffman?

  15 years ago the best engineers were either going to either their goal was to help people click on more ads that was like one goal or the goal was to in trading like let's make more money for financial services and not to say those are like bad for society I'm a big fan of both of them but there's probably like a lot more important things that the best engineers can be working on that could really propel society forward and it does seem like over the last let's say five years or so that there's been this like massive shift. And I would say like a lot of the engineers I know who are going to click on ads or going to

• 38:28Auren Hoffman?

  to financial services today are actually like the second tier engineers. They're no longer the top tier engineers.

• 38:35Speaker 00

  Again, it's like one of those things where I don't know if it's true or just it it needs to be true because it tells such a clean story. But you know, people always plot like, where do MBAs go? And they've always kind of like use this as like, hey, it's like, be careful. It's this like horrifically lagging indicator like MBAs flooding like flooded dot com companies in like 1999 and got blown up. They flooded banks in 2008 and got blown up and they flooded crypto in like 2018 and got blown up. And I, I, I, I hope that engineering talent is a leading and not a lagging indicator in that because I think I'm seeing the same dynamic as you.

• 39:12Auren Hoffman?

  My friends who are running like these financial services companies, obviously they always paid more. And so they could get, you know, top here down. That's part of the reason of top tier talent went to them. But now they are like, they're having a real like they're like massively up their salaries to get even like the same tier. They are, they're losing people all the time to these like more tech related companies, even people going for lower salaries. But obviously they get stock, which could end up, you know, becoming a higher comp. But also just like they, they enjoy their work better, like they're more, they were more fulfilled, et cetera. And then the, you know, the other problem with these, um, a lot of the financial services is you do have,

• 39:55Auren Hoffman?

  you, you, you've got a lot of these kind of like more exclusionary hiring practices where you, you, you can't go to a potential competitor. And, and so you just have like a diffuse learning that takes a lot more time to like invade the industry. And it's hard to know how your career will be impacted from that as well. Whereas in tech, you usually can just like, if, if for whatever reason you choose the wrong place, you can just jump after a year into, you know, some other place.

• 40:22Speaker 00

  Ben Thompson, like, I think like he made this point. I'm just kind of give credit where credit is due. But you know, he said like, that was always like one of the kind of like backstops, if you will, of the tech industry, which is like, if you started a startup and it didn't go anywhere, this or that or the other thing, you know, it's like, You were accumulating enough skills and enough network that like you could, you know, your work, if your, if your floor was like a soft landing at a fang, like that's a pretty gentle floor, you know?

• 40:47Auren Hoffman?

  If there was one defense prime that you, we could just kill and make the world a better place, like what would it be?

• 40:53Speaker 00

  You're gonna give me a hard time for this answer, but like it's not the primes, it's the system, right? Like I'm a, this is, this is the traitor in me coming back out again, which is like, if you want to understand the system, understand the incentives, right? And like, you know, a lot of ink has been spilled about, you know, the, the last supper and like, William Perry's consolidation of the defense primes and like the impact that that's had on that. I'm not gonna like rehash all that, but like, what I will say is, is that

• 41:25Speaker 00

  the, the insistence on big, big, big buying decisions with really long sales cycles is like fundamentally in like, is, is contrary to, to a startup's business model where like the number one number, you know, this like, you know, you wake, you, you're running a startup, like what's the number one number in your head in the morning? How much cash do I have? You know, and like that is a shot clock that you're on. And so I think like the,

• 41:54Speaker 00

  like, let me put it this way. If you look, I, I know what they are today, but like, um, the market cap of Palantir is enormous. You know, Andrew just raised around at around 60 billion, which means that like on a market cap size, they're about as big as like a Northrop Grumman at this point, or you know, like, but nobody would consider them a prime. Like, why is that? You know, and I think, I think a big part of that is, is like the business models. And this is where I think the distinction is really, really important. So where do you invest your capital? Do you invest your capital? Before you're given a contract or after you're given a contract?

• 42:32Auren Hoffman?

  And yeah.

• 42:34Speaker 00

  And it's crystal clear to me, I think that like, you know, the, the kind of Silicon Valley venture sort of world, I mean, it's almost tautological to say like, it's defined and like, we're giving you the capital to invest it to build a product and then you generate revenue from there. And so I guess kind of like what I put back to you, I'm like, what I think is like, super terrible is this mindset where like, the first thing that happens is a contract's awarded and then companies start spending their money and their capital. It shifts all the risk onto the buyer and that's not useful at all. And so like, you know, you want to fix defense primes, fix the way you acquire, force companies to invest their own capital, their own risk, build products and then buy the products that work.

• 43:17Auren Hoffman?

  And by the way, it also means that these products almost by default are going to be way more expensive because that if by buying it before these products exist, it essentially means you want something custom. So you want something like very, very custom to you, which, you know, and, and you're making some prediction eight years down the line about what is actually custom to you. And then so it's like, this is custom for this division of the Navy. And then of course, like this other division in the army wants something like, you know, might be slightly different. And then all of a sudden, like you have to like scrap it all and do this thing again, which doesn't like totally make sense.

• 43:49Speaker 00

  If you need to buy a, an ICBM or an aircraft carrier, like I can buy the argument that that's pretty custom, right? Like that, that makes sense. But if you need to buy like, like travel booking software, you know, like it's called DTS, like defense travel system. Like, I don't understand why you need a, like, why don't you just use kayak? You know? I mean, like, there's, it's, it's not, you know, you need to be custom.

• 44:12Auren Hoffman?

  Yeah. Or like, you know, it's like, um, maybe you need a pickup truck and then it's like, okay, well, like, okay, this Ford pickup truck like works pretty well. Okay. Maybe you need like different armor plates on it afterwards. So there's some sort of finishing that you do post buying a pickup truck or you need like a slightly different comm system inside of the pickup truck or, but it's like, it's a pickup truck and, you know, we all need pickup trucks to do certain things, right? And you could probably just buy the one that's like already existing that's out there that like a lot of people and yeah, yeah, maybe you need different tires or you need different types of things. Okay. Like great. Then figure out on the margin that you might need something different for the mission. And by the way, even within the army.

• 44:51Auren Hoffman?

  Like there might be like 50 different missions for a pickup truck. It's okay, great. We had to swap the tires or we have to put the armor over here or we have to change the fuselage or we have to do some other thing, but like those are all things that we can do like on top of the thing that we bought.

• 45:05Speaker 00

  I'll just yes and you where like, you know, not only is there that kind of like, hey, what's, what's kind of common and can be built off of, but then it's also like, you know, let's not, let's not let people wriggle off the hook and say, hey, look, you know, this, this pickup truck is 80% of what you need. You know, put me on contract and then I'll do the other 20%. Just like, no, like if you know what I need and you know that this is a good solution for it. Give me the, give me the pickup truck, the best pickup truck you think you can from me and then I'll make a decision and I'll be, you know, and it'll be quick.

• 45:35Auren Hoffman?

  Or I can use another vendor who can take it from 80 to 95 and then the other vendor can take it from 95 to 100 or something, right? Like it doesn't have to be like the same vendor. Companies do this all the time. Like they buy this thing and then they go do sit, you know, or else they go somewhere else. It's like, or you buy a piece of software from one vendor and then you use another vendor to help integrate it or, you know, it's like, it's like totally reasonable to it. It shouldn't like have to be like the same group that does everything.

• 46:02Speaker 00

  You know, everybody, everybody runs Oracle or SAP and in ERP, but there's like, it's your point. There's these nested layers of customization implementation that wrap around them. Yeah.

• 46:11Auren Hoffman?

  Now you guys are based in Austin. Obviously you have presence in many other places, including DC, but you're based in Austin. What is it? How do you feel about that? It's kind of like the ecosystem that you're kind of developing. You think of Austin as, I don't know, maybe a second tier place after like the Bay area or something.

• 46:29Speaker 00

  The way I frame it would be like the way that, you know, like the Intel guys, how they decided to get out of memory was like, hey, if, if we fired ourselves, what would be the first decision that the people who came in after us did? And they're like, we'd get out of memory.

• 46:42Speaker 00

  Kind of like with the same lens, like if we didn't have any history binding us, like where would we be headquartered? And I think like for us, Austin is 100% like the right, the right place. It's such a great tech ecosystem. It's got, I think, like a much more manageable standard cost of living than like the Bay area. But you've got like a critical mass of talent. You've got like a real talent density. And, um, and frankly, you know, for, for folks who have to spend a lot of their time flying from point A to point B, like you can get anywhere in the continental United States in like under four hours, which is a huge, huge win.

• 47:18Auren Hoffman?

  Two kind of personal questions we like to ask all of our guests. How are you personally using AI like your own hands on keyboard? How are you personally using it?

• 47:26Speaker 00

  Previously, like my house, we were running on a, like a home automation system that was like all kind of heavy software based. It's all a bunch of like JSON configuration stuff.

• 47:36Auren Hoffman?

  I'm sorry, what is this like the lights, when the lights should go on and that type of stuff? When the pool should clean itself?

• 47:42Speaker 00

  Rules based logic, like, hey, you know, if it's after, you know, like you get an Amazon package. If it's after sundown or whatever, like turn the lights on when a car is coming in the driveway. So, you know, somebody can see where they're going, like all that stuff, um, you know, thermostats, you know, the, the whole, the whole lot. But it's great because like in particular, like, you know, my family, like wife, two kids, like they need to have interfaces for this thing too. And like just building out the UI.

• 48:09Auren Hoffman?

  Yeah.

• 48:09Speaker 00

  Managing the rules and making that stuff super intuitive. It's been, it's been fun. I'm still in the middle of like porting it all over, but that's kind of been my, my fun project with it.

• 48:18Auren Hoffman?

  In the UI, you can log into when you're not at your home too. So if you need to do something or whatever, or do you have to like VPN into this box somewhere or?

• 48:28Speaker 00

  You gotta log in and all that, but you know, I think like to, to maybe tie it back to like the creepy parts of it. Like where I want to go with this too, is like just building out, you know, building out rules like, hey, when, you know, when all the cell phones that are usually attached to the home wifi network drop off, like a scene, nobody's home and like turn the lights off.

• 48:45Auren Hoffman?

  By the way, you're like the, the 10th person who I know who's like playing with this thing. It's so cool. Like someone needs to build something more like, I don't know, like, like we shouldn't all have to like recreate this. I'm sure like other smart people are all doing this kind of thing. Ideally, it was like a thing you could just buy that would do this all. Well, for whatever reason, like everyone like it's they they want so much to be in this like weird ecosystem that like no one's willing to do this kind of like relatively simple thing that we all need.

• 49:12Speaker 00

  The 64,000 other question is, is that a feature or bug? Like, is it? Is it? Yes, this is all common and there actually could be like an app for it. Or is it like, because you're talking about like? Your house and the way you live and all that stuff. Is this like the kind of like X ring of like neurotic personalization where like if it's 1% wrong between Orin and Jim? One of us or probably both of us are gonna hate it.

• 49:36Auren Hoffman?

  Yeah, that might be true. Though, I mean, it does take time to put it all together. So then it's just like, it is, there's still a function of a trade off there. That that go in and what are you like? And you're just like doing this in? Clog code or codex at your house and stuff? Okay.

• 49:51Speaker 00

  And then don't, you know, don't ask like, hey, how long do you actually think over the lifespan of you over your lifespan, you're gonna spend turning lights off that you forgot in like? Are you gonna spend more or less time doing this?

• 50:02Auren Hoffman?

  Yeah, I found like just all these quote unquote smart things that the home are really, they're very brittle. Like they connect to like weird Wi Fi's often like go out. It's like it's sometimes I just like, I just like wanna switch. Do it myself or, you know, or something like that. And, you know, it's like, these things are always like less. Good and and, you know? I don't know. Maybe they're like also made in China and spying on me too. Like, I don't even know how all these things all fit together.

• 50:30Speaker 00

  You the the complexity exponentiates on this stuff and it becomes unmanageable, you know, where it's like? If if you're talking about, you know, raising and lowering blinds and stuff, like, well, why do we need AI for that? You know, it's like, I'll just create like some pretty binary rules. And, uh, you know, then you rapidly get into situations where it's like, oh, I want the the shades to go up and down. Uh, at sunrise and sunset or like an hour before an hour after. Okay, fine. I want the shades to go up and down. If we're gonna watch a movie, I want them to go down so we can watch the movie. And like, you start like burying this stuff together and where you end up with is like this spaghetti monstrosity, you know, and that's where you're just like, dude, like? I need to abstract this out to a tent and then kind of clean it up from there. And that's where we're we're at. becomes really, really helpful.

• 51:15Auren Hoffman?

  Why is it so hard? Like, I'm sure you have cameras all over your house and stuff like that for security, whatever. Like, why is it so hard? Do you get all these like false positive alerts on your camera? Like, I just want to kind of simple. Okay, like, if this is like me, or my kids, or my spouse, or like, you know, people I know, my friendly neighbor, or whatever, like the friendly delivery guy who delivers UPS package, like, I don't need alert. But if it's like some sort of stranger, or some sort of thing. And by the way, I also don't need a alert if it's like a rabbit just walking down the lawn either, right? So like, why is it so hard to like, take things to the next level?

• 51:55Speaker 00

  Part like with the question like, why are like, why am I such an AI optimist? And it's like, I think very provably, like, yeah, I got a PhD doing computational neural stuff a long time ago. So I like always loved the brain and like fascinated me and like very provably, like, there are some things that our brains do so much better than any kind of artificial system we've constructed, you know, and I think it, I think for things like vision, those are just purely mechanical computations. Like, it gives me so much optimism. There's a whole lot more, more room to run in what we do with this stuff. But I mean, yeah, I mean, it was in the seventies, like, first time, like, somebody connected a computer to a camera, you know, this really smart guy, like a

• 52:39Speaker 00

  Pathfinder and AI is that Marvin Minsky, very famously, he like assigned computer vision is like a dissertation project to like one core grad student. He's like, you know, it's like, oh, this is easy. You'll figure it out in three years.

• 52:50Auren Hoffman?

  Yeah, it's amazing how bad just like, yeah, basic, you know, photo, face recognition is on these systems and stuff like you would think like a lot of the stuff would be like mostly solved today and still like, pretty, pretty weak. All right, last question we ask all of our guests. What conventional wisdom or advice you think is generally bad advice?

• 53:11Speaker 00

  The problem is, is that, like, I listen to your podcasts, and so

• 53:17Speaker 00

  You know, I've, I've, I've worked, I've heard.

• 53:20Auren Hoffman?

  Right. And you can't say follow your passion because everyone says follow your passion, which is like.

• 53:25Speaker 00

  You, you literally took the words right out of my mouth, which is like, I know if I just said like, you know, you need feedback, like you need feedback. I'm like, orange is going to give me a hard time because everybody says that, so I won't. But we both knew that's where we were going to go with it. So I'll, I'll give you something different and, and I'd be curious your reaction to it because you are, I think without a doubt, one of the best networked people I know. And like, and best here doesn't mean biggest, it means best. And, and so like the, the kind of like, I think like the conventional wisdom I've heard a lot is like, you know, like the value of networking and like, you got a network and this and that and the other thing.

• 54:03Speaker 00

  And I think, I think the way I think about like relationships and things sometimes is like in the, in the atomic element, like you need to think about them transactionally. Like if I'm going to ask somebody for an hour of their time, how am I going to be sure that like, I leave them with something like a bit of knowledge, information, something that makes that hour worth what I'm asking them to spend with me. And so I think like, when people think about like, hey, you got to grow your network and that stuff like, that can be true. But it can only be true once you're at a point where you're like, I know that I can provide value to these people myself. And I'm not just like, I'm not just kind of being a time thief with them. And so I say like, so I guess kind of like the like, grow your network.

• 54:47Speaker 00

  I'd say like, I don't really agree with that up until the point where like, you have something to contribute yourself to that network. And when you have something to contribute, it's probably the right time to grow it. But you're, you're much better at this than I am. So.

• 54:59Auren Hoffman?

  I'd a hundred percent agree with that. And there's always this kind of like, there, there used to be this phrase like, it's not what you know, it's who you know. And I think that was true for a very brief time. It was true from probably like the seventies through the nineties or something. You know, if you think like that, there's this tension between like the what you knows and the who you knows. The who you knows are were often like putting two what you knows together and they were taking a vig on the transaction. And then sometimes the who you knows, because they had this monopoly of knowing the what you knows, they actually became what you knows themselves because they like had it. And whereas today, I think like knowledge has become, oh, sorry, learning about who has knowledge is more diffuse.

• 55:44Auren Hoffman?

  It's a lot easier to find the what you knows. So the who you knows are are less, still important, but less important. You see that in financial services, like in the eighties, like the the top coolest, you know, financial services job was the investment banker, which is essentially who you know. And by the 2000s, the top coolest job was the hedge fund person, which was a who you know. I'm sorry, what you know, right? And so it has like, these things do kind of change and stuff. Now with AI, it's very hard to know because of AI like, okay, maybe the what you knows become less important. And so I'm not really sure how that's going to change. Certainly, I think for the last 20 years, the what you knows were ascendant above the who you knows. The next 20 years, like we'll all learn together what what is, what's the most important thing.

• 56:29Speaker 00

  That's an interesting way of thinking about it. I hadn't thought about that, the the whats and the the whos, like the the nodes and the edges in this graph.

• 56:37Auren Hoffman?

  Yeah, and how you put it together and stuff like that. Like back in the day, if you wanted to find this like amazing expert about the oil in the Gulf of Mexico versus the Gulf of the Persian Gulf or something, right? You would, it would be like hard to find that person, right? It's like now, it's like, I don't know, like three clicks later, you can find like the exact person to talk to. And in fact, you may not even need to talk to that person because you might just be able to talk to Claude or OpenAI or whatever. And they they actually may be able to help you kind of do that. You don't need to like go to the library anymore to get the, so it's it's it's an interesting thing to try to figure out all this stuff.

• 57:15Speaker 00

  You might ask Claude who that person is and then be like, you know, I'm just gonna ask it the info, the question I was gonna ask that person and see what I get.

• 57:22Auren Hoffman?

  Totally. And by the way, before asking advice nowadays, at least for me, I just, I first go in a call and try to ask all the questions. And then if there's like something I really don't understand, I may go to somebody else and ask them for advice on it. And so often like people asking for advice before they even like do the first step, it's like, well, okay, well, wouldn't you have at least gone there and, and, and, and then try to learn before you go to this person?

• 57:47Speaker 00

  Both of us kind of coming from like the technical world, you know, it's like the most, you know, savage put down you can ever give someone is to like, when somebody asks a question, you're like, let me Google that for you, you know? And now it's kind of like, let, let me ask Claude for you.

• 58:00Auren Hoffman?

  Yeah, exactly. Sometimes people email me the question like, oh, can you ask me a little email me the question or something like that? And I'll just like, I'm like, I think, I think they probably could have just put this in Claude first. And I'll like, ask them like, hey, did you, you know? So it's kind of a funny kind of thing out there. But it's an involvement, which is, which is great. Sometimes people are, the reason why they're asking isn't to learn the answer. They're, they're asking because they want permission to go do something they've already decided. And so then you have to help people understand that. It's what so it depends on, okay, is it like, is it like your best friend or your spouse or something? Or is it like this actually random person on the side who, you know, who actually, you know, so all these things are like hard to, to wrap together.

• 58:42Speaker 00

  We could spend an hour just talking about that last point, which is like, what fraction of times I ask a question, am I asking, asking for honest information? Or am I asking for a post hoc justification of something I already decided on? And

• 58:53Auren Hoffman?

  Right, right. Do I look good in this shirt? Like, I don't know if I want the real answer, right? You know, I just want someone to be like, yes, you look great. Okay, awesome. Okay.

• 59:02Speaker 00

  Well, he's wearing the shirt. So I mean, like, you know, whatever my, you know, so whatever my posterior probability is on what the right answer is, like, I got awaited by the fact that like, well, clearly he likes enough to pop it on. So.

• 59:14Auren Hoffman?

  All right. This has been awesome. Thank you, Jim Murbescu, CEO of StriveWorks for joining us on Sumation. This has been incredible. I really appreciate, I knew it would be good, but this was really, really great. I really appreciate you being on the, on the podcast.

• 59:26Speaker 00

  Awesome. Farid, thanks for having me. Great talking. This was a lot of fun.

• 59:30Auren Hoffman?

  One more thing before we go. I heard a blog called Summation. It's the same as this podcast. The blog is about non-obvious idea sharing on business, talent, data, longevity, and random contrarian takes. If you like the conversations on this show, you'll probably like the blog. It's free. New content comes out twice a month. You could subscribe at orin.substack.com. That's orin.substack.com.