Helsing Scales AI Strike Drones to 3,000 a Month, Builds Fighter Jet

Helsing 将 AI 攻击无人机产能提升至每月 3,000 架，并着手制造战斗机 由于无人机和人工智能让坦克和有人驾驶战斗机在现代战场上无法生存，民主国家只能通过大规模制造自主武器来威慑侵略。 Torsten Reil 解释了乌克兰战争为何标志着坦克时代的终结，并认为未来的冲突将由在严密干扰环境中运行的低成本、大规模制造的自主无人机和人工智能系统来决定。他详细介绍了 Helsing 的起源，获得了 Daniel Ek 的支持以打造欧洲主权国防科技，赢得德国台风战机电子战项目，并开发了能在空战中表现优于人类战斗机飞行员的 AI 飞行员 Centaur。Reil 还概述了 Helsing 围绕批量生产的 HX-2 攻击无人机和分布式“韧性工厂”的工业战略，同时阐述了其严格的管理哲学，该哲学以极高的人才密度、季度绩效校准以及对实干而非公开评论的不妥协偏好为核心。 乌克兰战争表明，传统的装甲和火炮在无人机网络面前日益过时，推动军事战略走向“自主集群”——即能够在严重干扰环境中无 GPS 运行的大规模 AI 引导机器群。 (0:00) Helsing 开发了 Centaur，这是一款以 25,000 倍实时速度训练的强化学习 AI 飞行员，在波罗的海上空的实战场景中完全控制了一架萨博鹰狮战机，这促使该公司收购了一家巴伐利亚飞机制造商来建造自己的自主喷气式飞机 CA1 Europa。 (25:35) 为了满足消耗战的需求，Helsing 重新设计了其 HX-2 攻击无人机（射程 100 公里，载荷 2 至 4 公斤），将产能从每月数十架扩大到 1,000 架的基准和 3,000 架的爆发产能，并得到了 10 亿欧元的德国框架合同的支持。 (42:10) 尽管 Helsing 于 2021 年 3 月成立时欧洲国防科技被视为 ESG 的异类，但在联合创始人 Torsten Reil 随口提到这一数字后，Spotify 创始人 Daniel Ek 通过 Prima Materia 承诺投资 1,000 万欧元（注：此处原文为 1 亿欧元），从而巩固了创始人们永不出售公司的契约。 (11:59)

Torsten Reil, co-founder of Helsing, argues that the combination of autonomous software and mass-produced materiel is fundamentally altering warfare, rendering traditional armored assets obsolete and establishing deterrence as democracies' primary defense imperative. Autonomous Mass and Obsolescence of Legacy Platforms (0:02) Reil contends that modern battlefield dynamics in Ukraine demonstrate the end of the tank era, as cheap strike drones guided by intelligence, surveillance, and reconnaissance (ISR) systems rapidly hunt down both armored vehicles and artillery positioned 40 to 50 kilometers behind front lines. In his assessment, crewed heavy armor has turned soldiers into high-priority targets whose life expectancy is measured in minutes. To counter traditional forces, Reil advocates for "autonomous mass." He calculates that a stockpile of 100,000 loitering strike drones capable of operating without GPS and resilient to electronic jamming could neutralize an invading force of 3,000 tanks within hours. Because machines and algorithms increasingly outpace human reaction times and render contested zones uninhabitable for biological soldiers, warfare is pivoting toward an industrial contest of uncrewed autonomous platforms across ground and air domains. Software-Compensated Hardware and Distributed Manufacturing (42:10) Rather than building exquisite, costly defense hardware that requires long manufacturing cycles, Helsing adopts a paradigm where software compensates for simplified physical engineering. Looser manufacturing tolerances and cheaper baseline sensors are offset by dynamic AI calibration in flight, enabling scalable production. Helsing's fixed-wing strike drones—engineered with 2 to 4 kilogram payloads and 100-kilometer ranges—are designed to reach outputs of 3,000 units per month, with Reil noting that procurement velocity rather than factory capacity remains the critical bottleneck. To protect against concentrated strikes that threaten centralized defense plants and static military airbases, Helsing champions distributed sovereign operations. The company establishes localized "resilience factories" inside allied host nations and designs its autonomous platforms—such as the CA1 Europa fighter—to launch from dispersed civilian road networks and autobahns rather than fragile, pristine runways. Challenging Incumbents via Programs of Record (23:24) Founded in March 2021 by Reil, Gundbert Scherf, and Niklas Köhler, Helsing bucked conventional startup models by targeting established government "programs of record" directly as a prime contractor. Backed early by an €8 million seed round and a subsequent €100 million commitment from Spotify founder Daniel Ek, the company possessed the capital reserves required to absorb lengthy procurement cycles and qualification hurdles that eliminate typical venture-backed entrants. Within two years and four months of founding, Helsing partnered with Saab to win a German program of record to upgrade the Eurofighter's electronic warfare suite using deep-learning AI to identify software-defined enemy radar signatures. Helsing later vertically integrated by acquiring an aircraft manufacturing facility near Munich with its own runway to develop the uncrewed CA1 Europa jet, validating its Centaur AI pilot in live beyond-visual-range trials over the Baltic Sea using a human safety-pilot protocol to conform to German aviation rules. Deterrence Doctrine and Operational Realism (35:36) Reil defines Helsing's foundational objective as ethical deterrence: fielding capabilities so formidable that aggressors calculate any invasion attempt as prohibitively costly. Acknowledging that predatory impulses cannot be permanently engineered out of human nature, he maintains that peace requires democratic societies to appear "very prickly." Success is achieved only when the weapons developed are never fired. While external journalistic scrutiny over lethal autonomous systems faded after the 2022 invasion of Ukraine, Reil reports that Helsing enforces internal ethical guidelines, red lines, and regular workshops. He also rejects common European founder complaints regarding regulatory friction and early exits, arguing that Europe's primary limitation is an ambition deficit, and notes that Helsing's internal culture is anchored on the principles of "whatever it takes" and "the truth is in the field." Reil's strategic model depends on governments overcoming procurement inertia by committing to large-scale programs of record, while the broader defense sector faces unresolved ethical and operational questions surrounding lethal autonomous decision-making as live scrutiny shifts toward industrial output.

[0:00] Torsten Reil: the era of tanks is probably over [13:43] Torsten Reil: we will never sell the company [16:17] Torsten Reil: The mission is to protect our democracies [35:36] Torsten Reil: the AI is just smarter than humans for this use case [50:34] Torsten Reil: We can build them really fast. [51:34] David Senra: The single most powerful pattern I have noticed is that successful people find value in unexpected places. And they do this by thinking about business from first principles instead of formulas. [52:46] Torsten Reil: When you actually look at it in detail... it isn't the most important thing, and I would argue that hiring and finding the right individuals number one, and then secondly curating that group of individuals and also moving people out, is the most important thing in a company. [52:46] Torsten Reil: We listen to the managers explain why they give their reports a certain rating. And then the manager is held accountable to that rating. So we ask them, you know, 'Why did you give that person that kind of rating?' And then we might say, 'Well, that doesn't sound like a three, that sounds like a two,' or 'doesn't sound like a three, it sounds like a four.' And that allows us to make sure that the calibration that people have of what is good performance, what does good look like, is the same across the company. And it also means that it's hard to hide. [59:04] Torsten Reil: shed its skin almost like every nine months [63:19] Torsten Reil: the power law in talent is extreme [65:27] Torsten Reil: We want to rewrite our values. We want to throw the old ones away. We know now what Helsing is about and we want to like do it completely differently. [65:27] Torsten Reil: The truth is in the field [80:54] Torsten Reil: entirely worthless

[0:02] Let's start with what we were just [0:03] talking about before we started [0:04] recording. Uh your what your opinion is [0:07] of the future of tanks and battleships. [0:10] >> I think there's a um big change [0:11] happening and you can already see it in [0:13] Ukraine obviously where um you know [0:15] tanks are barely being used anymore. The [0:17] same is true for artillery systems. If [0:19] you um get yourself into a tank, your [0:21] survival rate is is very short. I mean, [0:23] apparently the um uh the time that a [0:26] Russian soldier survives now at the [0:27] front is about 20 to 30 minutes. You can [0:30] argue about whether that's, you know, [0:31] correct, but it's probably roughly right [0:32] in terms of ballpark. And it probably [0:34] gets even worse if you're in a tank. If [0:36] you put yourself into the tank, you are [0:37] one of the high priority targets for uh [0:40] anyone on the other side. You know, [0:41] those targets can be found pretty easily [0:42] now with with drones. They can be [0:44] destroyed very quickly with small drones [0:46] or with large drones if they're a little [0:48] bit further away. And so, I think the [0:50] era of tanks is probably over. And for [0:51] artillery systems as well, if artillery [0:53] systems try to hide 40 to 50 kilometers [0:56] away from the front uh to to shoot, you [0:58] can still find them with drones and you [1:00] can attack them with drones. And again, [1:01] they're high priority targets. So, I [1:03] think we're seeing a big change right [1:05] now in uh in in warfare. And those [1:08] lessons have been pretty apparent, I [1:10] think, in Ukraine over the past, I would [1:11] say, 18 months or so. And I think people [1:12] are starting to understand them now. [1:14] >> What's killing the Russian soldiers so [1:16] quickly? Is it all drones? [1:17] >> It's drones. Yeah. It's it's the ability [1:19] first of all to see them with ISR drones [1:22] and then secondly to strike them with [1:24] either FPV drones or larger strike [1:26] drones. [1:26] >> Wait, what's the difference between [1:27] those two drones? [1:28] >> So FPVs are basically um smaller drones, [1:31] quadcopters that um most of the time are [1:33] controlled manually like a remote [1:35] controlled plane and strike drones often [1:38] travel over longer distances um and like [1:40] the HX2 which is our drone um quite [1:43] often the AI controlled. you had a a [1:45] tweet that was interesting where and [1:47] maybe this can tie into the dinner that [1:50] we had a few years ago where we had this [1:51] fierce argument and I left thinking that [1:53] we hated each other and then now we've [1:55] become friends but you said something [1:57] like a 100,000 of your drones could [2:00] prevent a land war in Europe. What does [2:03] that mean and why why do you believe [2:04] that? [2:04] >> Yeah, I said this I think maybe a year [2:05] ago or so and I I got into a bit of [2:07] trouble for this because apparently it [2:08] was deemed to be provocative but um I I [2:10] still think that's the case. I think the [2:13] deterrence that you get out of having a [2:14] very large number of strike drones at [2:16] your border, I think is is pretty uh is [2:19] pretty strong. And the reason is that if [2:21] you turn up with tanks on the other [2:22] side, let's say you turn up with 3,000 [2:24] tanks, 100,000 strike drones that are [2:26] able to navigate without GPS that are [2:29] able to find their targets, obviously [2:30] specifically focusing on tanks, will [2:33] destroy those tanks, I think, in a [2:34] matter of hours. Uh it's very hard to [2:36] defend against. If you were on the other [2:38] side, you wanted to invade a country [2:40] that had a 100,000 of your drones, how [2:42] do you do that? [2:43] >> I think it's very difficult. So, so, so [2:46] using tanks would be extremely difficult [2:48] because again, that'd be targets [2:49] immediately. You'd probably want to not [2:52] use those kinds of vehicles. You [2:54] probably want to use autonomous systems, [2:55] smaller autonomous systems. And this is [2:57] basically where everything is going. So, [2:59] you have autonomy now um, you know, with [3:01] drones. You have autonomy coming up um, [3:04] now with with autonomous fighter jets. [3:06] And you also have it coming up now with [3:08] um with land systems and particular [3:10] robots. [3:10] >> So what do you mean by autonomous drones [3:12] though? So there it's all run by AI. [3:14] >> It will be. Yeah. At the moment you [3:15] still have a lot especially in Ukraine a [3:17] lot of drones are still remote [3:18] controlled. So you have um you know [3:20] thousands of very skilled drone [3:22] operators. That won't scale though in [3:24] particular won't scale to the west and [3:26] won't scale to NATO. If you if you [3:27] really want scaled drone usage they have [3:30] to fly autonomously. You want to have a [3:32] human uh surveying them, looking at what [3:34] they're doing and being on or in the [3:36] loop, but the drones themselves need to [3:38] be AI controlled and autonomous. It's [3:39] the only way to get the true mass and [3:41] it's the only way also to make it work [3:42] pragmatically because we we won't be [3:44] able to train thousands and thousands of [3:46] drone operators. We need AI to control [3:48] them. [3:48] >> Is that a solved problem yet or you [3:50] still haven't figured out? [3:50] >> It's reasonably solved. I would say I [3:52] think it's the so the need so far has [3:54] been on first of all making the drones [3:56] work for terminal guidance which is [3:58] basically the last 2 km or so and then [4:00] to to create the software to coordinate [4:02] them all and now it's going to go [4:04] towards almost full autonomy again [4:07] always with a human in the loop. I want [4:08] to go back to the conversation we had a [4:10] few years ago when we we were having uh [4:12] dinner at Daniel X's house about [4:15] essentially your your proposition to me [4:17] was, you know, you're just not going to [4:19] be able to have humans on the [4:20] battlefield anymore because it's just [4:21] going to be about who can produce more [4:23] material. I think we discussed this book [4:25] called Freedom's Forge that I've read [4:26] like twice and essentially it's about [4:28] the United States converting their, you [4:31] know, industrial and commercial uh [4:33] industrial power from making products [4:35] that people would buy to making war [4:37] materials. and just mass-producing more [4:40] than the opposition had. Yeah. And that, [4:42] you know, helped, you know, tilt the war [4:44] in uh the balance in their fa in their [4:46] favor. [4:46] >> I thought you had a very similar [4:48] perspective that that's what we would [4:50] have to do. Like the the the the winners [4:52] of the future wars are just going to [4:53] have to create mass-produce more drones, [4:55] more weapons, more everything. [4:57] >> I think it's going to be a battle of [4:58] material. It's machines versus machines. [5:00] The reason for this is that the uh [5:03] autonomous mass as as we call it, so the [5:05] number of drones um that can fly in a [5:07] coordinated way is getting better and [5:09] better and it's going to be very hard [5:10] for humans to survive in that [5:11] battlefield. And again, that's already [5:14] the case in Ukraine. You can't really [5:16] enter the battlefield as a as a as a [5:18] human uh and survive for very long. Uh [5:20] whether you're in a vehicle or you're on [5:22] your own, it's very hard to survive. [5:23] That's going to get worse and worse [5:24] because again, at the moment, the drones [5:27] attacking you are usually remote [5:28] controlled. Once they're going to be [5:29] autonomous, it's going to be even more [5:30] dangerous. Those drones usually don't [5:33] care about jamming anymore. Um, for [5:35] FPVs, it's because they're connected [5:36] through fiber optic cables, uh, [5:38] fiberglass. Um, for larger drones, it's [5:42] because they are already autonomous and [5:43] they don't have to worry about GPS [5:44] because they have GPS free navigation. [5:46] They can detect targets and they can [5:47] engage once you've allowed them to do it [5:50] uh, autonomously. [5:51] >> Wait, explain how you get around the [5:52] jamming. I don't understand that. [5:53] >> So, the problem with jamming is it's [5:55] twofold. First of all, if you if you [5:57] want to control your drone from afar, um [5:59] you need a video signal back and usually [6:01] um the the battlefield is jammed. So, [6:03] you don't get the video signal back. So, [6:05] that means that the drone needs to fly [6:06] autonomously or it needs to be connected [6:08] with a cable. And that's that's what's [6:09] happening right now. And the second [6:10] problem is GPS jamming. Pretty much the [6:13] whole front line is GPS jammed. So, you [6:15] can't use your GPS receiver. So, you [6:16] need to find a different way to navigate [6:18] and know where you are. You either do [6:20] that um with a human trying to recognize [6:22] where they are just almost like a like a [6:24] pilot of a plane does or you do it with [6:26] AI that tries to do the same thing. And [6:28] and those algorithms now are very [6:30] robust. They work uh in in daylight at [6:33] night uh you know in the snow um even [6:35] with kind of deformed terrain which [6:37] obviously happens quite a lot with with [6:38] shell impacts. So that technology [6:40] already exists. So these drones are [6:42] getting better and better and [snorts] [6:43] at the same time if you want to have [6:45] humans on the battlefield I think it's [6:46] going to be very difficult and I think [6:48] the trend is going to be and I think [6:50] people are realizing this is that it's [6:51] going to be unethical to send humans on [6:53] a battlefield where you know they're not [6:55] going to survive but at the same time [6:57] you still have to retake territory. You [7:00] have to protect um and defend territory. [7:03] So you need you need robots on the [7:05] ground. If you extrapolate from where we [7:06] are right now I think that's where we're [7:07] going to go. I want to go into the way [7:09] you're running housing right now, but I [7:11] think to start is like why you decided [7:13] to found the company to begin with. We [7:16] can give like a real quick overview. You [7:18] came from gaming, right? You sold a [7:21] company for half a billion dollars. You [7:22] were angel investing kind of looking for [7:24] your next thing. Did you know for a fact [7:26] that you were going to found another [7:27] company after your first sale or No, [7:30] >> after I left um Natural Motion and Zinga [7:33] um I I I didn't think I was going to [7:34] start another company. I was [7:35] >> You did or did not? [7:36] >> I did not. Yeah, I was pretty exhausted [7:38] to be honest. Uh I'd been doing it for [7:40] 15 years and I was kind of very low on [7:43] uh kind of in my in my fuel tank. That [7:45] didn't take very long to recover about 3 [7:47] months maybe or so and I felt kind of [7:49] the energy was back. But then it took [7:51] another maybe 2 or 3 years or so uh [7:53] before I thought actually I want to [7:54] start something again. And [snorts] [7:56] honestly I wouldn't have done it had I [7:57] not met my two co-founders Gbert and [7:59] Nicholas. Especially with what we're [8:01] doing in defense, it's so complex. Um I [8:03] mean being a founder is already lonely [8:05] uh being a single um founder but in [8:07] defense that the companies are so [8:09] complex. You have to think about so many [8:11] things at the same time. You have to [8:12] obviously deal with technology um you [8:14] have to deal with um politics um with [8:17] export law and a whole bunch of other [8:19] things that it's very difficult to to do [8:21] it as a single founder. So when I met my [8:24] two co-founders um we all independently [8:27] had thought that there needed to be a [8:29] new defense company in Europe. How did [8:31] that conversation even start? [8:32] >> Back then, I invested in Nicholas's [8:34] company uh in in Munich um which was [8:36] working on solving AI problems for some [8:39] of the legacy primes in Europe. And I [8:42] wanted to find out more about defense [8:44] because I'd invested but I didn't know [8:45] that much about it. And someone um [8:48] recommended I I meet up with Gundad um [8:50] who at the time was at at McKenzie [8:52] responsible for aerospace and defense. [8:54] and and we um met up in a cafe in Berlin [8:56] and hit it off and within 30 minutes we [8:59] thought we should start this company [9:00] together went to see Nicholas who had [9:02] the same feeling and then like within a [9:05] few days or so we decided um to to start [9:07] Helsing and and this was very high [9:10] conviction and there was never a doubt [9:11] either and it's kind of still the [9:13] feeling of the company now the company [9:15] moves I think still very fast and a lot [9:17] of it is based on quickly settling on [9:20] founder conviction um you know in terms [9:22] of what we do um that we need need to go [9:24] into certain areas um certain [9:26] territories etc. [9:27] >> Hold on. I don't want to move on from [9:28] the conviction yet because this the year [9:30] you're having these conversations what [9:31] year is this? [9:32] >> Uh that was end of 20 to 2020 beginning [9:35] of 21. [9:36] >> Okay. I can't think of a more unpopular [9:38] company to start at that time than a [9:40] European defense tech company. [9:42] >> Yeah, it was very unpopular. Um we we we [9:44] founded the company in March 2021 and uh [9:48] it wasn't uh popular in our social [9:50] social circles. It caused all kinds of [9:52] problems with banks. Although this has [9:53] got better now obviously but back then [9:55] it was quite unpopular and and it was [9:57] hard to raise money like we didn't [9:58] >> wait cause problem with banks. [10:00] >> There's a risk and and some people have [10:02] been debbanked uh who've been close to [10:04] our company um around the time and uh I [10:08] I wasn't um but I know um another person [10:10] close to the company um had difficult [10:12] getting credit difficulty getting credit [10:13] because we were involved in in defense [10:16] so and and this is still the case now. I [10:18] mean it's it's it hasn't completely [10:19] disappeared. I think ESG requirements [10:21] are still there. And back then um the [10:23] defense of a country of a democratic [10:25] society was grouped for banks um in the [10:28] same bucket as pornography and gambling [10:31] and smoking because it was seen as [10:33] weapons and weapons are always bad. And [10:35] I I empathize with the kind of gut [10:38] feeling on this. It's just that when you [10:39] think it through obviously it doesn't [10:40] make sense. And this is also why we [10:42] started the company. It felt like if we [10:44] don't start the company then no one else [10:46] will. We were speci specifically worried [10:49] about um Europe's reaction to um Crimea [10:52] in in 2014. Uh the invasion and the [10:55] taking of Crimea where I think we all of [10:58] us didn't really want to think that this [11:01] uh you know was the beginning of [11:02] something bigger. We all didn't want it [11:04] to be true. But if you looked at um you [11:06] know what the ambitions were, it was [11:08] pretty clear that something else was [11:10] going to happen. And we were concerned [11:11] that you know Europe was going to be [11:13] very dependent on other territories you [11:15] know for its defense and on other [11:17] countries. And you know the [snorts] US [11:19] had companies like Andreal and Palanteer [11:22] that were very forward-looking defense [11:23] companies with a big focus on AI cheaper [11:26] hardware and abundant hardware um and [11:28] the idea of mass manufacturing but [11:30] Europe had nothing. And so we felt we [11:32] have to we have to do this and we have [11:33] to prove that we can build something [11:35] like this although with a slightly [11:36] different flavor uh in Europe and that's [11:38] why we started it very difficult to [11:41] raise money in uh 21 for defense like no [11:44] VC wanted to talk to us and we didn't [11:46] even try in the end so we just raised [11:48] money from uh individuals family offices [11:50] and then we were lucky that Daniel um [11:52] you know put in hund00 million euros of [11:54] his own money very early on and that [11:56] allowed us to to scale the company [11:57] faster than otherwise would have been [11:59] >> this I always read about you guys and [12:00] they're like Yeah. And they got a $100 [12:02] million investment from Prima Materia. [12:04] Like that's his own money. [12:06] >> Yeah. [12:06] >> That's a high conviction bet really [12:09] really early. This is before the [12:11] invasion of uh Ukraine. [12:13] >> Yeah. [12:14] >> There was a seed round first, right? [12:16] Didn't you guys raise raised like what 8 [12:17] or 10 million or something like that? [12:18] >> Yeah. I think it was 88 8 million euros [12:20] in uh in March 21. [12:22] >> And he was in that seed round. [12:23] >> Yeah, he was in that seed round too. [12:24] >> And then you guys had this conversation [12:26] where he asked, "Well, how much do you [12:27] need?" [12:28] >> Yeah. Like 3 months later. So So [12:29] >> yeah. And your response was [12:31] >> ideally 100 million euros. [12:32] >> How'd you come up with that number? [12:33] >> I didn't really do a um specific budget [12:36] for it to be honest, [laughter] [12:38] >> but it felt about right. [12:39] >> It felt about right. [12:40] >> And um but that's true for almost all [12:42] funding rounds, isn't it? Like it's it's [12:43] it's it's you you kind of eyeball it [12:46] because so many things um you know [12:48] change so quickly in a company. It's [12:50] it's very hard to kind of find an exact [12:52] number. But Dan Daniel had the same [12:54] level of conviction that um Gbert [12:56] Nicholas and I had um and and I think is [12:58] a similar conviction person. Doesn't [13:00] take very long and then the conviction [13:02] is very deep and then it makes [13:03] everything easier. It made all of the [13:05] conversations easier. We didn't have to [13:06] really kind of struggle to to come to a [13:08] deal that just happened really quickly. [13:10] >> You consider him more like a another [13:12] co-founder than an investor? [13:14] >> Yeah, I think he's closer to a [13:15] co-founder than an investor because he's [13:17] been with the company from the [13:18] beginning. I knew Daniel before. He knew [13:20] about the idea um from the beginning. It [13:22] was a big commitment for him um to to be [13:25] also be publicly seen uh you know to be [13:27] investing in this kind of company and [13:29] obviously then what happened only 11 [13:32] months after we started the company the [13:33] Ukraine war happened and public [13:35] sentiment shifted and it is now easier [13:37] and now a lot of people are investing in [13:39] defense you know both in the US as well [13:41] as in Europe but at the time it was [13:42] extremely unpopular [13:43] >> but I think this is one of the things [13:44] I'd most admire about you this this idea [13:46] of founder conviction and I love that [13:48] you're you're like I've heard you say [13:49] this before it's like the limiting [13:51] factor is not capital. It's like [13:52] actually ambition. Yeah. [13:54] >> And one thing that I've talked to a [13:56] bunch of people that have been on this [13:57] podcast is like I try to encourage them [13:59] stop selling your goddamn companies. [14:01] Yeah. [14:01] >> It's like% [14:02] >> Yeah. Your whole thing is like it's stop [14:05] selling your company for 100 million and [14:06] then buying a house in Italy and [14:08] becoming an investor. Like we need more [14:09] people. We have enough people in [ __ ] [14:10] finance, dude. You need more people [14:12] doing [ __ ] [14:12] >> Yeah. And so I completely agree. I have [14:14] this theory that the reason this happens [14:16] so often in Europe is partly because [14:18] Europe is such a nice continent to [14:20] retire in because you can buy a house in [14:22] Tuskany and you can fly there from [14:24] Scandinavia. You're there in like [14:25] whatever 2 hours or so and and you can [14:27] live an amazing life and once a few of [14:30] your friends start doing that then it's [14:31] kind of socially accepted but it [14:34] shouldn't be socially acceptable that [14:35] that you retire as a founder that's been [14:37] successful and can start another [14:38] company. [14:39] >> My friends that sell their companies I [14:40] say this jokingly but it's a lot of [14:42] truth is set in justest. They're like, [14:43] "Got acquired." Like, "Sorry to hear [14:44] that." [laughter] [14:46] >> Yeah. Exactly. [14:46] >> Entrepreneurship is the best game in the [14:47] world. You took yourself out of the game [14:48] and on the [ __ ] sidelines to what? [14:50] Write checks. [14:51] >> No, I agree. I mean, it's easy for me to [14:52] say cuz I sold my previous company as [14:54] well. Um, and it may be too early, [14:56] although I've been doing it for a while [14:58] and as I said, I I I for a while thought [15:00] I wasn't going to start another company [15:02] because building companies is exhausting [15:03] and, you know, sometimes you need to [15:04] recharge your batteries. But once you've [15:07] done it, I think people are also afraid [15:08] that they might not find the right [15:10] mission or even that, you know, they [15:12] might find that the second company isn't [15:13] as successful as the first one. And that [15:15] is a risk. You know, I thought about [15:16] that too. What what if the whole thing [15:18] just doesn't work at all. But, you know, [15:20] I think that's a risk you have to take. [15:21] >> I just talked to Scott Woo from [15:22] Cognition and you know, they're throwing [15:25] all kinds of crazy money at him. You [15:26] know, this is this huge like a million [15:29] people are trying to acquire him as a [15:31] person or the entire company. And his [15:33] whole thing is like, "Yeah, but this is [15:34] like the biggest opportunity of my [15:35] life." He's like, "I couldn't live with [15:36] myself if I didn't at least try." Like, [15:38] "I'm fine failing." And he then he said, [15:40] "I'd be salty as hell." Meaning, you [15:41] know, he's super competitive, but at [15:43] least I [15:44] >> try to like be removed from the game [15:46] completely as some kind of piece on [15:48] somebody else's chessboard. Just like [15:50] intolerable to somebody like that. [15:51] >> Yeah. No, no, totally. And for us with [15:53] with Helsing, it's the same thing. We we [15:54] made a decision probably 4 and a half [15:56] years ago, just a few months into the [15:58] company, that we're going to go full [16:00] throttle and we either kind of [16:01] completely crash the thing or be super [16:03] successful. But we didn't want to do [16:05] anything in between. And that's what [16:06] we're doing right now is that we try [16:08] really to keep the the the foot on the [16:10] gas really aggressively because we want [16:12] to build a company that really matters [16:13] and and we we will never sell the [16:15] company. [16:17] I have a funny story about this and then [16:19] we're going to get back to housing, but [16:20] um Adam, the founder of Apploven, was [16:21] sitting in the chair that you're in now [16:23] and he tells this hilarious story. He [16:24] takes the company public. It's like 28 [16:26] billion or something like that. Every [16:27] day for like 200 days, the stock just [16:29] goes down and down and down and he goes [16:30] it gets down to like three billion. He's [16:32] like he's like this is either like a [16:34] multiple hundred billion dollar company [16:35] or it's nothing. So then he winds up [16:37] borrowing like he goes into debt by like [16:39] six billion and does stock buybacks and [16:41] then winds up making like 60 billion on [16:43] that trade. It's like I'm not running a [16:44] [ __ ] $3 billion company. It's like [16:46] this is either a big company or nothing [16:48] and I'm not playing this little game. [16:49] >> I think that's the only way to do it. [16:51] >> I want to tell you about the presenting [16:52] sponsor of this podcast, RAMP. I have [16:54] been reading a lot about SpaceX lately. [16:57] SpaceX is one of the most valuable [16:58] businesses in the world and one of the [17:00] main themes in the history of SpaceX is [17:02] constantly attacking and questioning [17:04] your cost. Ramp helps many of the most [17:06] innovative businesses in the world do [17:08] exactly that. The median company running [17:10] on ramp cuts their expenses by 5%. And [17:13] one thing SpaceX has demonstrated is [17:15] that a religious dedication to [17:17] controlling costs can help actually [17:19] increase revenue because you can pursue [17:21] opportunities you couldn't otherwise. [17:23] And we see that in the RAMP data too. [17:26] The median company running on ramp also [17:28] grows their revenue by 16%. So when [17:31] you're running your business on RAMP and [17:32] your competitors are not, you have a [17:34] massive competitive advantage that [17:36] compounds over time. Ramp is the only [17:38] platform designed to make your finance [17:40] team faster and happier. Many of the top [17:43] founders and CEOs I know run their [17:44] business on ramp. I run my business on [17:46] ramp and you should too. Go to ramp.com [17:49] to learn how they can help your business [17:50] save time, save money, and grow revenue. [17:53] That is ramp.com. [17:56] Deal is how the best founders turn the [17:58] world into their talent pool. I've been [18:00] studying how history's greatest founders [18:01] operate for a decade. And one thing they [18:03] all have in common is they understand [18:04] that recruiting and hiring the very best [18:06] talent is your most [music] important [18:08] priority. A players recognize other A [18:10] players which is why top companies like [18:12] Ramp, Shopify, 11 Labs, Uber and Door [18:15] Dash all use Deal. Many of the top [18:17] founders I know have personally invested [18:19] [music] in Deal after using their [18:21] product. And what they discovered is [18:22] that Deal is the best company in [music] [18:23] the world at building infrastructure for [18:25] global hiring. Deal will help your [18:26] business hire, pay, and manage any [18:28] worker [music] anywhere in the world so [18:30] you can retain the best talent anywhere [18:32] and spend the rest of your time focusing [18:34] on what you do best, delivering value to [18:36] your [music] customers. The founder of [18:38] 11 Labs has a great description of the [18:40] value Deal can give your company. He [18:42] [music] said, "We built 11 Labs to break [18:44] down language and communication [18:45] barriers. With deal enabling us to hire [18:48] and support exceptional talent anywhere, [18:50] we can accelerate [music] our innovation [18:52] and bring more voices, stories, and [18:54] ideas to every corner of the world. Deal [18:57] is trusted by over 40,000 [music] [18:59] businesses. Learn how they can help your [19:00] business today by going to [19:02] deal.com/enra. [19:04] that is deal.com/enra. [19:08] >> Okay. So the identification of what was [19:10] missing in Europe, right? Which leads to [19:12] this mission that we need a European [19:14] defense tech company. Then the mission [19:17] comes before the conviction or the [19:18] conviction comes before the mission. [19:20] >> The mission comes first. Um and and the [19:21] mission is to protect our democracies. [19:23] And I always say it's um it's it's [19:26] always a real privilege. So I always [19:27] feel like we have um we talk to our new [19:29] starters every month and we always say [19:31] to them, you know, it is a real [19:32] privilege to work in a company. Um at [19:34] least we feel that's where the mission [19:35] is so easy that we protect our [19:37] democracies. That's what we're here for. [19:39] Um and and lots of other companies, you [19:41] know, have to spend time and go on an [19:43] offsite retreat to come up with some [19:44] kind of mission paragraph that everyone [19:46] agrees on eventually. And we don't have [19:47] that. It's it is to protect our [19:49] democracies. And we can point at where [19:50] we're doing this right now with, you [19:52] know, what we're doing with our deployed [19:53] drones, for example, in in in Ukraine. [19:56] and it makes everything easier. Um we [19:58] obviously never expected the Ukraine war [20:00] to happen that quickly but now that we [20:01] are deployed it's you know it's very [20:03] concrete. Um this mission is a is is a [20:05] real thing. And then the conviction you [20:07] know comes that they need to be these [20:09] kinds of companies and and that they [20:11] need to be built in Europe. And one [20:14] other reason why we are building Helsing [20:16] is also we we wanted to build a big [20:18] company in Europe. We honestly looked [20:21] at, [snorts] you know, Spotify as a as a [20:24] as a beacon, as a as an example of, you [20:27] know, a large company you can build that [20:28] is that has, you know, a global [20:30] ambitions, but then also actually global [20:32] impact and is up there with other big [20:34] companies from the US and and other [20:35] countries. And we wanted to build [20:38] something like that too. We wanted to [20:39] show that you can do this and you can do [20:40] it in what is a heavily regulated area [20:43] like defense because you know a lot of [20:46] people find excuses why they can't build [20:48] a big company. Uh it's very often [20:51] regulation, bureaucracy, not enough [20:52] funding. Um and none of that is true. [20:55] They're all excuses. [20:56] >> You have a great line about this where I [20:58] love what you said the excuse is like [21:00] people are way too concerned about the [21:01] regulatory environment, the tax [21:02] environment. It's like what's missing is [21:03] actually you need aggressive founders [21:05] with large ambitions. It's just [21:07] aggression that's missing. Exactly. It's [21:09] it's it's um strategic ambitions. You [21:11] have to think about what you want to be [21:12] and you and it needs to be big and then [21:14] you need to be truly aggressive and a [21:16] lot of founders I think in Europe have [21:17] have lost that aggression but it's easy [21:20] to find it again. You you just need to [21:22] you just need role models like we use [21:24] Spotify as a role model and um you know [21:27] once you have them that you know that's [21:29] what good looks like and everything [21:30] below that is is is failure. It is [21:33] underestimated how much of a difference [21:36] aggression and ambition makes. [21:37] >> Say more about that. [21:38] >> I I think it's 95% of a company success [21:40] is that everything else like I said is [21:43] excuses. So regulation doesn't matter. [21:45] Yes, sure it takes longer to uh start a [21:47] G&BH in in Germany. Who cares? Yes, you [21:49] have to go to a notary. Who cares? I [21:52] mean like you know we spend a couple of [21:53] days more starting the company. It's not [21:55] a problem. um the the regulation that we [21:58] are subject to and and other defense [21:59] companies are as well um with regards to [22:02] the certification qualification of our [22:04] systems, export law and a whole bunch of [22:05] other things um dealing with [22:07] confidentiality um classified [22:09] information is like 100 times more than [22:11] what a startup has to deal with in terms [22:13] of bureaucracy and you can still deal [22:14] with it. It still doesn't slow us down. [22:16] You just have to set the company up [22:17] properly. [22:18] >> You're probably one of the most [22:20] disagreeable people that I know. Have [22:22] you always been that way? [22:23] >> Yeah, you said that to me before. [22:26] >> [laughter] [22:27] >> and I've been trying to reflect on it. [22:29] Um, I don't think I am. [22:30] >> You just demonstrated here. Yeah, we [22:32] have more regulation. Who gives a [ __ ] [22:34] Stop making excuses. Like there is it's [22:36] a it's a superpower for founders. I'm [22:37] not obviously know it's not a pjorative [22:39] when I say [22:40] >> I just I just don't like when people [22:41] make excuses and and and these things [22:43] happen very quickly. You just need to [22:44] have a bubble of people coming together [22:46] and all reinforcing each other with [22:48] these kinds of excuses and then they [22:50] have a you know content life and then we [22:52] um sadly have a continent where we don't [22:53] have breakout companies because people [22:55] just lack the aggression. But again I I [22:58] am very optimistic. There are quick [23:00] fixes. You just have to have a few very [23:02] aggressive companies that point out that [23:04] everyone else is basically mailing it [23:06] in. And I think a lot of people are. [23:07] >> Me and Daniel talked about this this [23:08] this idea of like how many prime movers [23:10] are there in the world? you know, the [23:11] people that are inventing the the [23:12] technology that we're all using, [23:14] creating the companies, you know, power [23:15] laws, rule, everything around us. And [23:16] it's just like it's not that many. And [23:18] if you just have a handful, like you [23:20] literally can change the trajectory of [23:22] an entire country, entire world, [23:23] whatever the case may be. When you [23:24] raised the 100 million though, where did [23:27] you see like your entrance into doing [23:29] this? Was the idea like, hey, we're [23:31] going to what you just said, it's like [23:33] cheap, simple hardware and then all the [23:35] innovation comes in software. Is that [23:37] the way you thought about it? Actually [23:38] the biggest insight came and I think it [23:40] came from from Gonbad when we um when we [23:42] said we're we're just going to go for [23:44] programs of record which as a back then [23:47] as a as a small company um was almost [23:49] impossible and was extremely risky. [23:51] >> Explain what a program of record program [23:53] of record is a is a program that is an [23:55] official government program that's in [23:56] the budget that is usually quite large [23:58] and where you ideally want to be the [24:00] prime. So you have the direct [24:01] relationship with the government rather [24:02] than working as a subcontractor through [24:05] another company and that's extremely [24:07] hard to break into. And the reason most [24:08] companies didn't do it back then was [24:10] because you know you got funding and you [24:13] needed to generate some revenues and [24:15] some proof points you know reasonably [24:16] quickly and that's almost impossible to [24:18] do with a program of record because it [24:19] takes so long um to get one uh and to [24:21] qualify for one. But because we had the [24:23] 100 million, we thought, okay, well, now [24:25] we can take those risks. And if the [24:27] first two or three don't work that we go [24:29] for, we still have enough money left. [24:30] And we felt it's the only way to build a [24:32] defense company reasonably quickly. And [24:34] so we we um uh uh applied for a big [24:37] program of record for the Euro Fighter [24:39] electronic warfare upgrade. [24:40] >> What is that? [24:41] >> In in Germany. So this is electronic [24:43] warfare capabilities are um essentially [24:45] essentially around detecting radars um [24:47] detecting you [snorts] know the types of [24:49] radars in this case anti-aircraft radars [24:52] that you know try to take you down uh [24:53] detecting their intent etc and the [24:56] situation had become you know more [24:57] difficult um for for Europe because a [24:59] lot of those radars had become software [25:02] defined which means that they basically [25:04] are shapeshifting and so the old [25:05] approaches didn't really work anymore [25:07] and we came in with an AI based approach [25:09] deep learning based approach that kind [25:11] of recover that ability ability to um to [25:12] interpret those radars and it was a big [25:15] technology jump and we managed to win uh [25:18] that program of record together with our [25:19] partner SAP against you know a lot of [25:21] other consortia and we were the dark [25:23] horse but we ended up winning it. How [25:25] big was the probability? I don't know [25:26] but there was a high probability that [25:28] we'd fail but we went for it because we [25:29] thought we can afford to because we have [25:31] enough money in the bank and this this [25:34] happened after 2 years and 4 months of [25:36] founding the company. I think we became [25:38] the fastest to go from founding to [25:40] program of record. [25:41] >> How much of like your previous work at [25:43] like Oxford stuff you were doing at [25:45] Natural Motion did you bring into [25:48] housing? [25:48] >> A lot of it was the um I think team [25:52] building and fundraising and maybe [25:54] prioritization um pattern recognition [25:56] that I had from natural motion where um [25:59] I when I sold the company and where I [26:01] left the company I was reasonably [26:03] pleased with myself. I thought, you [26:04] know, this all went quite well in the [26:06] end. But it was only in the in the kind [26:08] of two or three years afterwards where I [26:09] started to reflect on some of the [26:11] things, you know, I'd done and how I [26:12] built the company, especially when I [26:14] started to angel invest, I realized I [26:15] actually screwed up a lot of things at [26:17] Natural Motion that I should have done [26:18] better. I I didn't deal with performance [26:20] management the way I should have. I [26:22] definitely didn't take hiring um uh [26:26] serious enough, just the kind of general [26:28] hiring of great engineers. And I also [26:30] didn't think enough about how to, you [26:32] know, think about your your leadership [26:34] team, how you grow your leadership team, [26:36] how you bring the right executives in at [26:38] the right time. And it kind of worked at [26:40] Natural Motion, but we could probably [26:41] could have done a lot of things better. [26:43] And and so a lot of that I think pattern [26:46] recognition uh we were able to use at [26:48] Natural Motion and particularly in the [26:50] first few years, but it's also true now [26:52] that that Helsing um is now at a point [26:54] where it's kind of uncharted territory [26:56] for me as well. The company is a lot [26:57] bigger. I think we're about 1,300 [26:59] people, much bigger than Natural Motion [27:00] ever was. And so we we have to now kind [27:03] of think through things either from [27:05] first principles or by, you know, [27:07] working with people who have already [27:08] built big companies like like Daniel. [27:10] And then the other part is we were lucky [27:12] that we had as three founders very three [27:15] different skill sets. You know, I had [27:16] already started a company and and sold [27:18] it. I think I knew how to um fund raise. [27:21] Gun was very close to um to the various [27:24] forces to the military. um also [27:26] understood the the processes political [27:28] processes really well and Nicholas's [27:30] background was on the AI and engineering [27:32] side um which is what we really needed [27:34] to to first first of all win win win [27:36] this [27:37] >> and this is what I'm that that was the [27:38] reason I asked that question it's like [27:39] well how the hell do you go from [27:40] founding to program a record faster than [27:42] anybody else like how did you even know [27:43] you could solve this problem of [27:45] >> it was just the feeling I think and and [27:47] uh particular Nicholas and Gumbad um you [27:50] know pushed that Gmbad I think had a [27:52] strong conviction that this was a really [27:54] worthwhile program of record for to have [27:56] and Nicholas had an intuition as to how [27:58] we could do it differently to anyone [27:59] else. [28:00] >> Where did Nicholas get that intuition? [28:02] >> Uh I don't know. Uh I think it's just a [28:04] I mean he'd already been in this space [28:06] um with his with his previous company. [28:08] Um so he already had a [28:10] >> what was his previous company doing? uh [28:11] it was solving AI problems for um for [28:14] for legacy primes and so he already had [28:16] a feeling for this kind of thing and [28:17] then if you combine that with with [28:18] GMAT's insight as to what programs of [28:20] record worked um it was beca honestly [28:23] became the blueprint how we approached [28:25] pretty much everything now um we we try [28:27] to find programs of record that are [28:29] large where we have a feeling that we [28:32] can contribute something to it that's [28:33] differentiated that's basically our lens [28:35] >> can you give me an example of that [28:37] >> well another example is is um the [28:39] autonomous fighter jet that we are [28:40] building right now. It's called CA1 [28:42] Europa and we went into it and invested [28:45] a lot of money into it um knowing that [28:47] Europe didn't have anything like it and [28:49] that we felt um this is going to be a [28:51] big category and that we could [28:53] contribute something to it and the [28:54] reason for that was we started a almost [28:56] like a science experiment. We have we [28:58] have a great um chief scientist Anton B [29:01] um who joined us from Meta where he ran [29:03] u um co-headed Facebook AI research and [29:06] we set ourselves a task two years ago. [29:07] So to to create a superhuman uh fighter [29:11] pilot uh for beyond visual range combat [29:13] so the kind of combat that's [29:14] >> hold on you got to slow this down. What [29:16] [laughter] [29:16] superhuman say that thing again? [29:18] >> It's a superhuman fighter pilot. Um so [29:20] >> superhuman fighter pilot but there's no [29:22] pilot in there. [29:22] >> No it's it's an AI. So it's it's a [29:24] neuronet network. [29:25] >> Okay. And um it's for beyond visual [29:28] range. So the most advanced type of air [29:30] combat. Um it's essentially chess in the [29:32] sky. You have you know the planes um [29:35] usually it's two versus two quite far [29:37] away from each other. These are fighter [29:38] jets and and they use radar. Um they use [29:41] different types of radar. They switch [29:42] the radar on off. They use missiles. [29:44] Obviously you have to go offensive, [29:47] defensive. You have to try to bleed off [29:49] energy of missiles that are following [29:51] you. You have to have a feeling for when [29:52] you can re-engage. It's it's extremely [29:55] complex. It's almost looks like ballet. [29:56] Uh it's quite beautiful to watch [29:58] actually when you look at it at the [30:00] tactical views. And we've been told by [30:03] pilots, active pilots, that this is the [30:05] kind of last bastion of of air combat. [30:08] Um you know, nothing [snorts] will ever [30:10] replace the intuition that a human has [30:12] in the sky to um read the situation to [30:16] [snorts] have a feeling for, you know, [30:18] how to engage, what types of um you [30:20] know, weapons to use. And that [snorts] [30:21] turns out to be [ __ ] um you know it [30:23] it actually the AI is much much better. [30:26] Um and the reason is that it's able to [30:28] explore the search space um much more [30:30] efficiently than than um you know a [30:33] human can because a human learns it from [30:34] a textbook and you know in in flight [30:36] school but um but they can't expose a [30:39] search base and so we use the same [30:41] approaches basically as um you know deep [30:44] mind used for alpha go and alpha star [30:46] using um reinforcement learning and uh [30:48] we have a extremely high performance [30:50] simulator that runs um 20 25,000 times [30:54] realtime speed And that allows us to [30:57] explore the search space with [30:58] reinforcement learning much much more [31:00] efficiently. And that now means that we [31:02] can train uh a neuronet network to be [31:06] better than a active fighter pilot like [31:09] we we regularly fly against Euro fighter [31:11] pilots for example or or Rafael pilots [31:13] or Grian pilots. [31:14] >> So wait you were running tests with your [31:17] autonomous fighter jet against a [31:21] humanpowered [31:22] >> pilot. So we used to do it in simulation [31:24] only. First it was only our simulation [31:27] and we won. Then it was uh you know an [31:29] official simulator but from a of a of a [31:32] uh you know of a big prime. And then um [31:36] and then last year we flew um a real [31:38] fighter jet and we flew the um the the [31:40] SAP Grien over the Baltic Sea. And when [31:43] I say we, our AI did and took full [31:45] control of the plane um in air combat [31:47] against uh another Grien which was human [31:50] controlled. And so we flew three um air [31:53] combat um scenarios um above the Baltic [31:56] Sea and and proved at that point and [31:58] this by the way was also the point where [31:59] we realized actually there's something [32:00] bigger happening where we this actually [32:03] works and it works in the real world and [32:05] and so we developed the system we called [32:07] it sent to uh and it's now able to work [32:10] in swarms as well by the way not just um [32:13] two vis swarms of fighter jets now [32:16] swarms of drones now we've moved on to [32:18] >> swarms of fighter jets and it scales [32:20] really [32:21] And um it's extremely powerful and that [32:25] gave us a conviction that crude manned [32:27] you know fighter jets were going to be [32:29] very difficult to deploy in the future [32:30] because you'll you'll end up facing [32:33] these swarms of autonomous fighter jets [32:34] with superhuman capabilities and it's [32:36] not science fiction. Um what's like one [32:39] observation of what's happening at the [32:40] moment not just in defense AI but AI in [32:43] general is that things happen faster [32:44] than you think. you know back in the old [32:47] days like 10 years ago or so like [32:48] everything was always five years away [32:50] that was always kind of the number that [32:52] people used it's about five years away [32:53] whether it's kind of particular medical [32:55] breakthrough or maybe robotics [32:57] breakthrough that's now [snorts] all [32:59] shrunk the 5 years usually ended up [33:01] being 7 to 10 years by the way same was [33:02] true autonomous driving it's now the [33:05] exact opposite when you think something [33:06] is 5 years away it's actually usually [33:08] now 2 years away and so something has [33:10] changed and we felt that with Sentar as [33:12] well it's like we're thinking like we [33:13] couldn't believe how well this worked [33:15] and and and still to this day I always [33:17] feel it feels a bit like science fiction [33:19] but that gave us the conviction that we [33:21] needed to be in this space and to be in [33:23] this space because it's so slow usually [33:25] to work with primes like we could have [33:26] partnered with a prime and ask do you [33:29] want to do an autonomous fighter jet we [33:30] provide the AI you provide um you know [33:32] the the jet it would have taken too long [33:34] so so we decided to acquire an aircraft [33:38] manufacturer um outside Munich but our [33:40] outside Munich with its own runway [33:42] hangers and full aircraft manufacturing [33:44] capab capabilities and build our own [33:46] fighter jet which is now called the CA1 [33:49] and then put scent tower on it as the [33:52] core technology component and so that's [33:54] going to fly for the first time um next [33:56] year and um and you know we want to show [34:00] that you can it's really leaprock [34:02] everyone in this category [34:03] >> we were having this conversation me you [34:05] and Michael Dell were talking about this [34:06] at Dell's house a few months ago you [34:09] told us about this acquisition and then [34:11] Michael's like can we come see it and I [34:13] was like I'll come with you But I don't [34:14] want you to shoot down Michael's plane [34:16] as we're coming in. But you told us [34:18] about this funny regulation where even [34:21] though it's going to be an autonomous [34:23] jet, fighter jet, don't you still have [34:24] to have a guy in it? [34:27] >> So eventually, no. Um but [34:29] >> right now, yes. [34:29] >> Yeah. I mean, we we so we we don't have [34:32] the um I think the advantage of um the [34:34] Mojave Desert, for example, where you [34:36] can fly those planes and be completely [34:38] autonomous because if it crashes, it [34:39] doesn't really matter. It you know, it [34:41] crashes into sand. if you you know [34:43] crashed where we fly you crash over [34:46] built up Bavaria uh which is not ideal. [34:48] So we have a human um that's optionally [34:51] in the plane uh for um for the first [34:53] prototypes eventually um that's going to [34:56] change obviously and and the the pilot [34:58] is a safety pilot the same way you'd [35:00] have um safety drivers for you know the [35:03] early rollouts of Whimo and and other [35:05] companies. [35:05] >> The fact that you have this regulation [35:07] at the moment and you need a person that [35:09] that has to change the design of what it [35:10] looks like. Correct. No, it's we found a [35:12] way to accommodate it without changing [35:13] the design. It's just it's the same it's [35:15] the same system. And by the way, it's [35:17] another example of where you can um you [35:19] know beat on the uh EU or Europe or in [35:23] this case Germany for having [35:24] regulations. Um but you just find a way [35:27] to deal with it and this is the way to [35:29] deal with it. You [35:30] >> just have a pilot sitting in there like [35:31] this. Is [laughter] [35:32] >> he in the back? Where is he? [35:34] >> They're in the front. [35:35] >> Okay. So you what you just described [35:37] that's very interesting. You said it's [35:38] like chest in the sky and then you said [35:39] it was ballet. So, the way you see the [35:42] short-term future is the same way it's [35:44] essentially going to be immoral. I want [35:45] to get into like the the morality, the [35:47] way you look at the morality of like [35:48] what you're building, the way it would [35:50] be immoral to send a human into the [35:51] battlefield because you're going to get [35:52] sworn by drones. It's going to be [35:54] immoral to send a pilot into the sky. Is [35:55] that the same? [35:56] >> Yeah, it's the same thing. I mean, there [35:58] are two reasons. A a it's more powerful [36:00] to use autonomous systems that they will [36:02] just be superior because the AI is just [36:05] smarter than humans for this for this [36:06] use case. And [snorts] secondly, as a [36:09] result, it will be immoral to send [36:10] someone out because you have to assume [36:11] that the other side will eventually have [36:13] those systems as well. [36:14] >> Can you explain like how you view like [36:16] the morality of what you're building and [36:18] how you may view the other people's [36:21] interpretation of you know essentially [36:23] banning to creating new weapons or [36:25] creating new defense like you've thought [36:26] a lot about this. I kind of just want [36:27] like a what's in your mind like a [36:30] complete overview of like how you think [36:31] about the morality of building the [36:32] company you're building. The underlying [36:34] goal has to be that you minimize the [36:38] lives lost in general and ideally you [36:40] get it down to zero. And the way to get [36:42] lives lost down to zero is to not have [36:45] any kind of conflict at all. And that's [36:47] why deterrence is so important. And this [36:50] is also where I mean this is a very old [36:52] um obviously argument um but it still [36:54] comes up all the time is that people [36:56] don't understand that you have to be [36:58] able to deter in order to not be a [37:00] target that looks valuable or or [37:03] achievable for the other side. And we [37:05] set up the company to deter um we never [37:08] set it up honestly to to to have the [37:10] systems deployed um at scale um in [37:12] Ukraine which is what we have right now. [37:15] will do it obviously because we'll have [37:16] to do it and um they you know it needs [37:18] to make a contribution but the idea in [37:21] general is to avoid conflict altogether [37:23] and and you do that by being able to [37:25] deter by having a a a credible posture [37:29] that makes the cost calculus on the [37:31] other side such that they don't want to [37:33] attack because it'll be too devastating [37:35] for them. And that is the first thing. [37:39] If you have to go into a conflict, you [37:40] want to do it in such a way that you [37:42] minimize the lives lost on on your side. [37:44] And again, this is where especially [37:46] right now, and it's going to get worse [37:48] in the future, you need to use autonomy [37:50] because you cannot send humans into [37:52] [snorts] tanks um you know, in in in [37:55] fighter jets um or even as infantry onto [37:57] the onto the field. That's the [37:59] underlying theme. And then there's a [38:01] another category where you need to think [38:03] about the the the way you deploy your [38:07] software and and what as a society um [38:09] you know we're happy with with the [38:11] deployment of your hardware and software [38:13] in terms of autonomy. So are we okay for [38:16] a machine to make a decision over life [38:18] and death if it is more accurate than a [38:20] human or [snorts] do we say as a society [38:23] um no a machine should never make that [38:25] decision and these are legitimate [38:27] conversations. is no easy answer because [38:30] people have different values on this and [38:32] the thing that worries me is that the [38:34] these discussions happened a lot um when [38:36] we started the company before the [38:38] Ukrainian war no one in particular [38:40] journalists weren't interested in asking [38:42] us um you know about the capabilities of [38:44] why we were doing things everyone always [38:46] asked about killer robots and autonomy [38:48] and our red lines and do we have a human [38:51] in the loop which we were always happy [38:52] to answer these weren't annoying [38:54] questions they at the core of what we [38:55] think about now though no one asks these [38:59] questions anymore and I'm not quite sure [39:01] why but no journalist ever asks those [39:04] questions you know how are we thinking [39:05] about the ethical side we do a lot of [39:07] this ourselves we have ethics workshops [39:09] we have a lot of deployment guides for [39:11] our technologies um we have to think [39:13] very carefully about you know things [39:15] like proliferation and a whole bunch of [39:17] other things but we have to do it out of [39:19] an inner discipline and um journalists [39:22] don't ask the questions anymore and the [39:23] public doesn't either [39:24] >> I don't even understand the counter [39:26] arguments that you receive that's like [39:27] no we should not have any new [39:29] technology. It's like well humans are [39:30] violent by nature. We've had war since [39:33] we've existed. Like [39:34] >> yeah but but it is an interesting topic [39:36] because I like we get this push back as [39:39] well obviously at Helsing and we get it [39:41] we get push back from people who say war [39:43] is always bad. Um you know and of course [39:45] at at at some level that makes sense but [39:48] again if you have a situation where you [39:49] are attacked um if you have a situation [39:51] where there is an adversary that is [39:54] building up um you know capabilities you [39:56] need to be able to deter in order to [39:59] avoid war and then you need to be able [40:01] to fight in order to then not lose lives [40:04] um you know are necessary lives on your [40:05] side and like I said we get a lot of [40:07] push back and and it is often from [40:09] well-meaning people so it's not that [40:13] like we hate to people who who have a [40:15] problem with our mission. Um I disagree [40:17] with the way they look at it. I think if [40:19] you think it through from first [40:20] principles, you need to have this kind [40:21] of deterrence, but we don't kind of [40:23] dislike people. It's like very often [40:25] people who just want to have peace. It's [40:27] just that we want to have peace too. We [40:29] just think you have to achieve it by not [40:32] letting a war happen and not being [40:33] inviting to a war. [40:35] >> But I think there's something [40:36] interesting. It goes back to like your [40:38] high conviction, the fact that you're [40:39] extremely disagreeable. you're very [40:41] comfortable taking like a counter [40:43] position like you started this company [40:44] when everybody else said you know it was [40:46] very unpopular. Uh something that I [40:48] asked Daniel about and he said he's like [40:50] he's a German building a defense company [40:52] in a country whose entire postwar [40:54] identity was never again. [40:55] >> Yeah. [40:55] >> Like you're also willing to cut across [40:58] your cultural like the culture [41:00] expectations too. [41:01] >> Yeah. Yeah. I think that's still the [41:03] case in Germany and you know we're set [41:05] up as a European company. We have big [41:07] offices in in in London and Paris and [41:09] and a bunch of other countries as well. [41:10] But um but our biggest is is in Munich [41:14] um in in Germany and and German society [41:16] is a pacifist society and it it's taking [41:20] a lot of [sighs] I think facing reality [41:23] right now which is what's been happening [41:24] in particular over the last couple of [41:26] years or so where where I think Germans [41:28] realize it hasn't been the end of [41:30] conflict. um there are now conflicts and [41:33] they're coming closer and they might be [41:35] coming even closer again and it's very [41:36] uncomfortable for for a country that [41:38] that thought those times are over but [41:40] it's also the country now that has one [41:42] of the most thriving new defense [41:44] ecosystems um I would say in the world [41:46] now so there is now a rethinking of some [41:49] of those attitudes and it's also the [41:51] country that has um in absolute terms [41:53] probably now the biggest growing defense [41:54] budget going from 52 billion euros last [41:56] year to originally I think 153 billion [42:00] in 2020 29 I think it's gone up again um [42:02] to over 180 billion and and this is a [42:05] large increase if you think about that [42:06] there won't be more soldiers most likely [42:09] so most of the money is going to go on [42:10] to new equipment spend [42:11] >> okay so we go back to materials your [42:13] whole thing is like the winner is going [42:14] to be able to produce more materials [42:15] than than their adversary so I I still [42:18] don't think I completely understand [42:21] the way you think about achieving that [42:23] goal is the hardware that you're [42:25] >> manufacturing has to be as simple as [42:27] possible and then you're offsetting all [42:28] the innovation in software is Is that [42:30] still the basic thesis? [42:32] >> The idea is that you build hardware that [42:33] is good enough um and build it um in [42:36] such high numbers that you eventually [42:38] have economies of scale but also the [42:41] ability to overcome any um you know [42:44] adversary just through pure mass. And [42:46] the way you achieve that is that you you [42:48] need to be able to massproduce. You need [42:49] to design to massproduce. But then [42:51] crucially you need software to um to [42:53] allow those systems to first of all be [42:56] smart and and be able to to coordinate. [42:59] Um but you ideally also absorb a lot of [43:02] the complexity of the of the hardware [43:04] that usually you know will be these [43:06] exquisite systems that um legacy primes [43:08] will make into your software. So you can [43:10] make the hardware simpler. That means [43:11] that your sensors are simpler and you [43:14] regain a lot of the accuracy through [43:16] very good AI algorithms. It might also [43:18] be that your um your manufacturing um [43:22] tolerances are a little bit bigger and [43:24] you regain a lot of the control and [43:26] accuracy again through software that [43:28] might recalibrate the system as it goes [43:31] and and and this is what we do and that [43:33] works extremely well. But you have to [43:35] think it through as as one unified [43:37] system and once you do that you can go [43:39] into mass manufacturing. So the the HX2 [43:42] is now a mass manufactured strike drone [43:44] and like we had to go through like all [43:47] of those learnings in in in July and [43:49] August last year where um we had to go [43:51] from dozens of HX2s per month to a [43:53] baseline of 1,000 and it can now search [43:55] to in in the small factory that we have [43:57] to 3,000 a month and it's a completely [43:59] different kettle of fish to to producing [44:01] just a few dozen. You have to think [44:03] about your supply chain completely [44:04] differently. um obviously the way you [44:06] design um the drone, quality control, um [44:10] tracking every single step. And so you [44:12] have to learn this. But once you've done [44:14] this and you're able to produce in very [44:16] large volumes, um you you just have much [44:18] more powerful systems because you can [44:20] now overcome any adversary or target by, [44:23] you know, having saturation attacks, for [44:25] example, by sending more and more of [44:26] these drones, by having them highly [44:27] coordinated, and if necessary, just [44:29] produce more and more of these. And [44:31] drones are just one example. The same is [44:32] true for fighter jets. want to build you [44:34] know many hundreds of these fighter [44:36] jets. Um same is true for um you know [44:39] for other systems maritime systems. It [44:41] is going to be about being able to build [44:43] these things in large numbers to design [44:45] them so that they can be built in such [44:47] numbers and to have the industrial [44:48] capacity and the supply chain to produce [44:51] a lot. [44:51] >> Hold on though. If we're talking about [44:52] war materials you [snorts] can never [44:54] predict the the potential new adversary [44:57] you might have in the future. That means [44:58] you guys have to be completely [44:59] vertically integrated. Like if I'm a [45:01] country like or a continent, let's say [45:02] we're European, it means I can't [45:04] outsource some of this production to [45:06] China or to America or to South America [45:09] or anywhere else. Correct. [45:11] >> I wouldn't say any I mean with China, [45:12] yes, that's true. Um um there's still [45:15] allies that you can work with obviously [45:16] and that that we all work with. Um but [45:18] there will be some areas and it depends [45:20] on the s sensitivities of particular [45:22] countries where you want to be fully [45:24] vertically integrated in that um country [45:26] and you want to be um having your entire [45:28] supply chain in that country as well. [45:30] There are other areas where you you work [45:32] with allies and um I think that will [45:34] continue and probably even grow and then [45:36] there are other areas where you say um [45:38] no it's we're actually there's so many [45:40] different suppliers it doesn't really [45:42] matter we are kind of free to choose but [45:44] I feel like you would go to more towards [45:47] complete control like I'm not talking [45:49] about what you're doing now but in the [45:50] future you're going to be running this [45:51] company for the rest of your life you [45:52] just said I'm not selling right this is [45:54] my last company this is my grand [45:55] ambition [45:56] >> so like would you not want [45:58] >> to work towards an area where like you [46:00] have complete control over everything. [46:02] >> Yes. But um you also want to work with [46:04] allies like there there are many [46:05] countries um that you want to work [46:07] together with and where you are stronger [46:09] if you're stronger if you work together. [46:10] There's some areas where um you know [46:13] specific countries have sensitivities [46:15] around particular data. For example, uh [46:17] the UK for example is very um specific [46:19] about what they're doing in maritime and [46:21] understandably so where you want to be [46:23] like really just national. But then [46:25] there are other areas where where you [46:27] work with allies and you need to be able [46:29] to dial it up and down. So what we're [46:31] doing, we have this concept of a [46:32] resilience factory where if a country [46:35] wants to, you know, buy and build uh our [46:38] hardware, they can do it on their own [46:39] territory. We put a factory there and [46:42] they can also um buy build up their own [46:44] supply chain to the point where they [46:46] feel comfortable with and then buy from [46:48] >> Explain how that partnership would work. [46:50] >> So [46:50] >> so your technology manufactured in their [46:52] factories on their soil. So, we can run [46:54] the factory, but we make sure it's on [46:56] their soil. Um, and we find a way to [46:58] work together with them to also make [47:00] sure that the supply chain is as as [47:03] national, as domestic as it needs to be. [47:05] And again, the sensitivities in [47:07] different countries are different. Some [47:09] countries don't care at all. They just [47:10] buy it from us and we can produce [47:11] wherever. Others really care that it's [47:13] built um, you know, on their soil so [47:15] that they have more control over it. [47:17] >> Why do you call them resilience [47:18] factories? because we designed them to [47:21] be not this kind of one big factory um [47:24] that is a a very easy target for attacks [47:27] um but actually they can be distributed [47:29] all over a continent or even else or [47:31] even be underground as is the case with [47:32] us so that they are harder to hit. [47:34] >> You studied the crop family? [47:36] >> No, I know of them but I don't know in [47:38] detail. [47:38] >> There's a great book I'm going to send [47:39] you. It's like 700 pages. I should do an [47:41] episode on it for my other podcast [47:42] founders, but this they ran into this [47:44] problem during World War II where the [47:46] Allies just found their centralized [47:48] factories and just bombed the [ __ ] out [47:49] of them. [47:50] >> Yeah. I mean, by the way, this this [47:51] problem is completely underestimated [47:52] right now. Coming back to um to um uh [47:55] fighter jets again. So, we have these [47:57] exquisite fighter jets like the Euro [47:58] Fighter Rafal and um you know, a bunch [48:00] of others, but they all need airfields [48:02] with with the exception of the Gripen. [48:04] They they they they tend to use um neat [48:07] air bases um to take off and land. those [48:09] air bases are obvious targets obviously [48:12] for for an attack and and so if you look [48:14] at something like the um like the [48:15] [clears throat] CA1 um our autonomous [48:17] fighter jets it's designed in such a way [48:19] that they can take off from any kind of [48:22] makeshift um air strip which would be a [48:24] road an autoban a motorway a highway etc [48:27] >> essentially adapting it to where you [48:29] can't bomb you're not going to bomb all [48:31] the road [48:32] >> you can't bomb all of it exactly okay [48:34] >> yeah and and I think there is still I I [48:36] think certain governments and militaries [48:38] are still in denial what actually is [48:39] going to happen in our conflict. Um, you [48:42] know, in many cases we won't be able to [48:44] take off with our crew fighter jets [48:45] anymore. [48:46] >> So, how do you go about trying to [48:48] explain that to them? Like, how do you [48:49] just make it obvious? Do you just do [48:50] through product demonstrations? [48:52] >> Yeah, I mean it's like with everything [48:53] else you um you you do product [48:55] demonstrations, you show what the [48:56] capabilities are. We've done this a lot [48:58] with drones and then and and to honestly [49:01] everyone's credit, people do get it. I [49:03] mean, it's not like people complain [49:04] about procurement is so slow and and [49:06] governments are so slow. Yes. [snorts] [49:08] But then when they do get it, it goes [49:10] very fast. Like the the the large um [49:12] strike room contracts that happened [49:13] happened really quickly um with [49:15] political support with um support of the [49:17] procurement agency and and they went [49:19] through all in record time and now if [49:22] anything I think that the companies you [49:23] know are really under pressure to [49:24] deliver all of this stuff really fast. [49:26] So it's like we we we all complained [49:29] about it in the past but it's a bit of a [49:30] legacy problem. I think um procurement [49:33] has got faster and and the realization [49:34] that these things have to happen is [49:36] happening faster too and and you know [49:38] there will be big um autonomous programs [49:40] in Europe as well now. Two years ago [49:42] there was nothing even one year ago [49:43] there was nothing but it's now [49:44] happening. [49:45] >> Something that just came to mind while [49:46] you were speaking earlier is you're [49:48] talking about how the you know in the AI [49:50] age everything you got going to happen [49:52] faster. If you think it's in five years, [49:54] it's actually two maybe two two years [49:56] maybe might be two months. Might even [49:57] be, you know, faster than than even you [50:00] suspect. It's so AI is making your [50:02] products better than and better faster [50:06] than you expected. [50:07] >> Can that also Can AI also help [50:09] manufacture and increase your [50:12] manufacturing faster too? Yeah. Because [50:13] if you're saying we were going to [50:14] produce a couple thousand a month, I [50:16] think isn't China doing like 100 million [50:18] drones a year or something like that? [50:20] >> Not really. I mean it's and Ukraine is [50:22] building I think 2 million or so a year. [50:24] Um but those are different types of [50:25] drones. They're FPVs. They're [50:27] quadcopters and again they have their [50:28] role and and they work but they're much [50:30] shorter range and they've got smaller [50:31] payloads or explosives. Strike drones [50:33] are much bigger. I mean they're they're [50:34] like this and um with [50:35] >> how big is it your strike zone? [50:37] >> So so the payload for example the [50:39] explosive is 2 to 4 kg um or even [50:41] larger. And [snorts] the uh range that [50:43] they can fly is is around 100 [50:45] kilometers. And so it's it's quite a lot [50:47] more than than FPVs and they're used in [50:49] different ways as well. Those numbers [50:51] right now like up to 3,000 a month um [50:53] are really quite high for this kind of [50:55] category and you can scale it up even [50:57] further. At the moment honestly the um [50:59] the bottleneck isn't manufacturing [51:01] speed. We we can we can build them [51:03] really fast. There will be a point [51:05] though in the future where you want to [51:06] be even more efficient and faster and [51:07] then AI will play a role in [51:09] manufacturing as well. At the moment [51:10] it's too early. It's because we're not [51:11] limited at the moment. [51:12] >> So wait, what is the bottleneck then? [51:14] >> Uh honestly the bottleneck right now is [51:16] the big programs need to come. So the [51:18] first ones are coming. Um we just closed [51:19] a 1 billion framework contract in in [51:21] Germany but they now need to come [51:23] everywhere and then we can crank up [51:25] manufacturing accordingly. [51:26] >> What do you mean by everywhere [51:27] >> in in the various countries because like [51:29] >> in all the western countries? [51:31] >> Yeah. [51:31] >> Okay. [51:32] >> Yeah. There'll be some in Asia as well. [51:34] >> I found one of my all-time favorite [51:36] quotes when I was reading the book 0ero [51:37] to1. The quote says, "The single most [51:39] powerful pattern I have noticed is that [51:41] [music] successful people find value in [51:43] unexpected places. And they do this by [51:46] thinking about business [music] from [51:47] first principles instead of formulas." [51:49] That is exactly what Apploven has done [51:52] with their advertising platform. [51:53] Apploving connects you with over a [51:55] billion potential new customers [music] [51:57] inside mobile games. Apploven allows you [52:00] to capture undivided attention. Apploven [52:02] ads are full screen video ads that are [52:05] watched for an average of 35 seconds. [52:07] [music] That is retention that blows [52:09] other ad platforms out of the water. And [52:11] you can launch [music] on Apploven in [52:13] minutes. You set the goal and AppLin [52:16] achieves it. There's no complex [music] [52:17] setup, no expertise needed, and Apploven [52:21] scales quickly. They can put your ads in [52:24] front of over a billion potential [52:26] customers. Other businesses have seen [52:28] immediate results, have scaled to [52:30] hundreds of thousands of dollars of [52:32] spend per day, and increased their [52:34] revenue by millions. So, you want to get [52:36] started quickly before all of your [52:37] competitors are on Apploving and you can [52:40] do that by going to apploven.com. That's [52:42] apploven.com. [52:46] >> All right. I want to get into how you [52:47] think about running the company. You [52:48] mentioned earlier the mistake that you [52:50] made with your previous company was not [52:51] putting enough emphasis on talent and [52:54] recruiting. What did you mean by that? [52:55] >> When you build a company, you obviously [52:56] try to hire good people and and every [52:58] founder will always say, you know, the [53:00] most important thing is talent and their [53:02] people and they spend so much time on [53:04] recruitment. But when you actually look [53:06] at it in in detail and I include us [53:08] there at a natural motion previously as [53:10] well um it is I wouldn't say it's [53:12] halfass but it isn't the most important [53:14] thing and I would argue that hiring and [53:17] and finding the right individuals number [53:19] one and then secondly curating that [53:21] group of individuals and also moving [53:22] people out is the most important thing [53:24] in a company and it [snorts] is the [53:26] reason why a company has great culture [53:28] and it's also the reason why a company [53:30] is high performant and it feels like [53:32] it's lighfooted uh and and Helsing feels [53:35] that way. Um I I think it it feels like [53:36] it's a very fast, highly performing [53:39] company that that can adapt really [53:41] quickly and [snorts] it is almost [53:43] entirely because the talent density is [53:45] so high. So if if that's your kind of [53:48] thesis and it was our thesis then you [53:51] have to think how to solve it and and [53:52] and the two main factors. The first one [53:54] is you need to hire well. So you need to [53:57] build a machine that high as well and [53:58] and and be able to monitor how it does [54:00] it. And the second one, it needs to be [54:02] able to curate as well. So you need to [54:05] have a machine, a system, a set of [54:08] processes that ideally are very easy [54:10] that um allow people to to also manage [54:13] people out when they don't work out. And [54:15] what a lot of people don't I think [54:17] appreciate is that these two are are [54:19] linked because it becomes much easier to [54:22] move people out that are not a good fit [54:24] for the company. If you know your hiring [54:27] pipeline is so good that that position [54:29] will be filled almost immediately. Then [54:31] it becomes kind of a low opportunity [54:32] cost thing. What happens instead in a [54:34] lot of startups and and probably also [54:36] larger companies is that that managers [54:38] are hesitant to move people out of a [54:40] company because a it's always [54:42] unpleasant. So you need to hold them [54:44] accountable but B because they maybe [54:47] rightly know that it will take 9 months [54:49] to fill that position again. And very [54:51] often the argument and and many founders [54:53] will be familiar with this. Very often [54:55] the argument of a manager will be yeah [54:57] but if I fire this person now then we [54:59] won't be able to hit our deadline [55:00] because at least they're going to get [55:02] the job done somehow. Maybe a bit slower [55:03] than they should. Maybe they create [55:04] cultural problems. They have an attitude [55:07] issue but at least eventually they'll [55:09] get the job done. And that's not good [55:10] enough. you need to always have enough [55:12] pressure on the pipeline to be able to [55:14] fill those positions quickly so that it [55:16] becomes easier. And so what we did with [55:18] Helsing at the beginning cuz we knew [55:20] this was uh you know going to be [55:22] important. The first contract we ever [55:24] signed even before we um founded the [55:26] company was with a third party [55:27] recruitment agency because we didn't [55:29] have anything internal. We always wanted [55:30] to start with pressure on the [55:32] recruitment pipeline. and uh Goomba [55:34] Nicholas and I then spent easily the [55:36] first year just reviewing the KPIs [55:39] easily every week um to see how our [55:42] recruitment worked and always injected [55:43] ourselves u if if it needed to be [55:46] readjusted because the recruitment [55:47] pipeline was so important building then [55:49] the internal recruitment team was really [55:51] important and then building processes [55:53] for how you manage um performance was [55:55] important as well and the most important [55:57] thing is a quarterly performance uh [55:59] calibration review and what we do there [56:01] is um we don't actually review the [56:04] performance of all the individuals. We [56:08] listen to the managers explain why they [56:11] give their reports a certain rating. And [56:14] then the manager is held accountable to [56:16] that rating. So we ask them, you know, [56:18] why did you give that person that kind [56:20] of rating? And then we might say, well, [56:22] that doesn't sound like a three, that [56:23] sounds like a two or doesn't sound like [56:25] a three, it sounds like a four. Um, and [56:27] that allows us to make sure that the [56:29] calibration that people have of what is [56:31] good performance, what does good look [56:33] like is the same across the company. And [56:35] [snorts] it also means that it's hard to [56:37] hide. Uh, if someone is an [56:39] underperformer um, and you you give them [56:41] a low rating, the expectation is that [56:43] you do something about it. You either [56:45] make sure that they become better and [56:46] that's your task as a manager or if [56:48] that's not possible, then you find ways [56:50] to part ways. And this kind of [56:53] accountability um I think is really [56:55] useful for a company to do that because [56:57] it allows you to eventually scale up and [56:59] have a general view of you know what is [57:00] the benchmark that we're looking for. [57:02] And it also gives us a feeling for which [57:03] managers try to protect the teams and [57:05] which don't. [57:06] >> I got to pull something up. Do you [57:08] remember um we were together we were [57:10] racing cars in uh in Sweden and Luca [57:13] Ferrari was there. [57:14] >> Yeah. [57:15] >> Did you spend any time with him? [57:16] >> I don't I didn't know. [57:17] >> You need to or I'm not telling you what [57:19] to do. You should. How about this? So, [57:21] we had this crazy lunch and Luca is a [57:23] very I like him a lot. He's but he's a [57:25] very strange cat in the sense that like [57:26] I consider his the way he runs company [57:28] almost like Galopagos island of [57:31] entrepreneurship because he didn't like [57:32] look to other entrepreneurs. He doesn't [57:34] really know much about entrepreneur [57:35] history when I talked to him. He doesn't [57:37] really pay attention to anything. He's [57:38] just like essentially like try to figure [57:40] out how to run a company based on what [57:42] he wanted to do. And so, he's got a lot [57:43] of very unique ideas. But I I remember [57:46] we had these notes and we were just [57:47] going back and forth on WhatsApp. I was [57:48] like, "Dude, I'm just going to type all [57:49] the notes of the crazy [ __ ] that you [57:51] said during this lunch because you have [57:52] some very interesting ideas on talent [57:56] and specifically HR." And so he had this [57:59] idea. These are some notes from this. [58:01] You should spend some time with him. He [58:02] goes, he talks about hiring. He says, [58:03] "We make it a science." He's talking [58:04] about what he's doing at Bennett Spoons. [58:06] He has a team of 50 in HR, but they're [58:08] all engineers. [58:10] >> I've never heard of this before. And so [58:12] what they do is they'll compare one, [58:14] two, and threeear performance to signals [58:16] from their CV. So we're he's saying, [58:18] "Hey, you know, we think quality X is [58:21] really important, but we're just [58:22] guessing." So what we're going to do is [58:23] say, "We hired this person because of [58:25] quality X. Well, what's their [58:26] performance based on all the people we [58:27] hire from Quality X? You're 12 months [58:29] out, 24 months out, 36 months out." And [58:32] he says they actually have developed [58:33] like this algorithm of a 100 different [58:35] signals of what they figure like is [58:38] important for talent. And he's like, [58:40] "We've completely we treat it as an [58:42] engineering process." And then he says, [58:43] "We c remember he has all these [58:45] different companies." And he's like, "We [58:46] centralize hiring and firing." And he [58:49] says, "Talent managers are in charge of [58:51] both uh entries, so people coming into [58:54] your company, which you just described, [58:55] and then access, which you also just [58:56] described." And he's like, "People," his [58:58] whole thing was like, "You should have [58:59] brilliant people in HR." That most [59:01] companies treat HR as just like% like B, [59:04] C, D players. He's like, "No, you should [59:06] have the best people in HR cuz it's so [59:07] important." [59:08] >> Totally agree. And I think this kind of [59:09] investment taking it so seriously. It's [59:11] just it compounds over time because you [59:14] just get the best people in. It carries [59:15] a different culture and and it becomes [59:17] something that uh it will be it's [59:20] central um to being um successful [59:22] eventually because especially if you can [59:24] keep that muscle over time when you grow [59:25] as a company. It's easy to lose it over [59:28] time and we always say kind of a company [59:30] like Helsing has to shed its skin almost [59:32] like every nine months and you have to [59:33] rethink how you do your processes again. [59:35] You have to rethink about how you grow, [59:37] what kind of leadership you need. But if [59:38] you can keep that kind of investment in [59:40] in a making sure you have a great intake [59:42] of people up, it's it's super valuable. [59:44] >> Are one of the co-founders having to [59:46] approve every hire? How do you guys how [59:48] do you do this? [59:48] >> We don't have to approve every high. I [59:50] think that's too difficult now. I think [59:51] we we do get involved in hiring. [59:53] >> I'll I'm going to argue with you since [59:54] you like to argue. Yeah. [59:55] >> You have how many employees? [59:57] >> 1,300. [59:57] >> Okay. It's not too difficult. Elon [59:59] approved the first personally approved [60:00] the first 3,000 hires of SpaceX. That's [60:03] in one of the books, the biographies of [60:04] >> apparently. Yeah. If you know that's [60:05] great. I I we don't do it that way. we [60:07] have um very good leadership that is [60:09] able to make those decisions. Um it is [60:11] true that you know sometimes we get [60:13] pulled in you know for a second opinion [60:15] and it's also true that we get pulled in [60:17] um to interview people and sometimes to [60:19] convince them to come into the company [60:21] and then for certain roles uh we do a [60:23] lot of the hiring ourselves like we [60:24] spend a lot of time the founders and [60:26] executive team on hiring. I want to hear [60:27] more about this [60:28] >> key positions. Um we we will sometimes [60:31] do end to end uh end to end the hiring [60:34] um as as founders. Um we we'll do the [60:36] search, we'll work together with search [60:38] agencies, we work together with our [60:40] network because they're so important [60:42] that we want to get it right and we know [60:43] that if we get the right person, it'll [60:45] be transformational for the company. So [60:47] Daniel also gets involved in our hiring. [60:49] um he uh just um you know really played [60:53] a very big role in hiring one of our [60:54] most senior executives recently which [60:56] got really well. So we all chip in. It's [60:58] so important to get the right people in. [61:00] It is again it's the most important. [61:02] >> But if somebody's in front of you, how [61:03] do you actually know if they're good or [61:04] not? [61:04] >> So everyone I think has different [61:05] methods for for me is I I try to talk to [61:08] them and I try to get a feeling for the [61:09] person without a standard set of [61:11] questions. I just try to find whether [61:14] out whether a person is effortless. [61:16] That's the probably most important [61:17] person. characteristic I look for. [61:20] >> Wait, what does effortless mean? [61:22] >> You It's like one of those things you [61:23] can you can tell if they if they don't [61:24] try to impress if they if you get a [61:27] feeling they have deep knowledge of what [61:29] they're talking about and at the same [61:30] time they're still curious and have high [61:32] agency. That's kind of the combination I [61:35] look for. And and com that combined to [61:36] me is effortlessness and it doesn't I [61:39] don't think it takes that long to to to [61:41] find it out. It takes in a conversation [61:43] maybe 10 minutes or so to get a good [61:46] feeling for whether a person is has this [61:49] this kind of effortlessness or not. That [61:51] that that's what I look for. But I don't [61:52] interview um software engineers for [61:54] example. I that wouldn't apply there. [61:56] >> But I mean you're essentially I would [61:59] imagine involved at the most key [62:00] positions that you're trying to prove [62:02] for. What can you what do you think is [62:04] happening in these 10 minutes? [62:05] >> Honestly, I haven't thought about it. Um [62:07] I don't know. I just I just try to get a [62:09] feeling for the person. I don't reflect [62:11] on these things that much to be honest. [62:12] I I just I I I know it works for me and [62:15] the only thing that matters for me is [62:16] whether it whether the outcome is right. [62:18] So I I do look obviously at the [62:20] statistics cuz if I kind of came in and [62:22] I just kind of viped the whole thing and [62:24] everyone and it kind of ended up being [62:26] crap after 6 months obviously that [62:28] wouldn't work. But that's not the case. [62:30] We um you know like all the senior [62:33] positions obviously interviewed by all [62:34] the founders and so we always have a [62:36] kind of second and third opinion on it. [62:38] But we we tend to agree on people. I [62:40] think we we look for similar similar [62:42] things. Sometimes you have to go into [62:44] longer interviews and people do longer [62:45] interviews with us as well. For [snorts] [62:47] me, it works better if I do short [62:48] interviews and I just get a feeling for [62:50] the person. [62:51] >> Have there been examples where your [62:53] initial intuition about a person has [62:56] been proved wrong in in housing in [62:58] particular? [62:58] >> Not at Helsing um but uh three times at [63:02] Natural Motion and I still know the [63:03] names and it still really annoys me why [63:07] I didn't read it. early enough in in all [63:09] three cases, they were so good at faking [63:13] the effortlessness and and so good at [63:15] coming up with um fake facts that I [63:18] didn't read through it. [63:19] >> I heard you say something on uh Harry's [63:21] podcast. Did you actually think there's [63:22] a lot more exceptional people out there [63:24] than other people believe, especially if [63:26] you have a couple thousand employees? [63:29] >> People's [63:30] character and and and and also [63:32] performance um is is not set in stone. [63:34] So quite often it's also a matter of in [63:36] larger companies sometimes it might [63:38] actually even be harder for people who [63:40] have a 10x potential to to become a 10x [63:43] person. So it's a combination of things. [63:44] It's probably harder to keep just hiring [63:47] 10 out of 10 people but also harder to [63:49] develop them. But it's still worth it. I [63:51] mean the the the power law in in talent [63:53] is is extreme. They're very few 10 out [63:55] of 10 people but they make I mean it's a [63:58] cliche but like we see this every day. [64:00] They make a disproportionate difference [64:02] to the company. So it is worth trying to [64:04] find these people and it is worth [64:06] finding a way to um give them a home in [64:09] in a company where they can really [64:10] thrive. We try to do this very very [64:13] systematically at Helsing and we we make [64:16] for I think hopefully for everyone we we [64:18] we create a a great environment where [64:20] they're surrounded by amazing people but [64:23] um you know for the 10 out of 10 we we [64:25] we go the extra mile. [64:26] >> How do you do it systematically? don't [64:28] want to go into too much detail how we [64:29] do it but we are we create we we create [64:32] an environment where they can do their [64:34] best work because and and we make [64:36] exceptions that we wouldn't always make [64:38] for everyone because they are such big [64:40] contributors to the to the company and [64:42] many of them now have contributed so so [64:45] much already. It's also a feeling of [64:46] gratefulness to be honest um to have a [64:49] bunch of those key employees that have [64:51] um made a big difference and you can [64:54] point at the successes of the company [64:55] and you always know there I mean [64:57] obviously it's all about talent because [64:59] people make all of this stuff. Maybe in [65:00] the future it's more AI, but at the [65:02] moment it's it's still mainly people. [65:04] And you can point at those people that [65:06] made a difference to the company. And so [65:08] we I think we all feel a lot of [65:09] gratefulness as well um to those top [65:11] individuals, to everyone in the company [65:13] obviously, but in particular to the ones [65:14] that you know have have turned a project [65:17] from really difficult to um you know [65:20] from red to green. That that happens [65:21] very often and it's quite often very [65:24] small teams or specific individuals that [65:25] make a big difference. Based on the [65:27] conversations we had, the way I would [65:28] summarize your perspective on this [65:29] because you talk about this a lot. It's [65:31] like talent density above everything [65:32] else. [65:33] >> Yeah. I think it's the number one thing [65:35] it's it's I I feel that culture comes [65:37] from talent density. It it's we had a [65:40] situation I don't know maybe it was 18 [65:42] or 24 months ago or so where we we had a [65:45] set of values um at Helsing that we came [65:48] up with at some point I think in a [65:50] workshop or so and they were on our [65:51] website and they were fine I think. And [65:54] then um we had one of our gettogethers [65:56] where the whole company got together and [65:58] a bunch of people came and said we want [66:01] to rewrite our values. We want to throw [66:02] the old ones away. We know now what [66:05] Helsing is about and we want to like do [66:07] it completely differently. And and then [66:09] so at this get together there was this [66:11] workshop and I have to say I was [66:12] dreading it cuz I didn't want to do [66:13] another value session cuz I never like [66:16] them. And it was effortless like the the [66:19] the team came up with their own values [66:22] based on their experiences and it wasn't [66:24] one of those exercises where you where [66:26] you kind of all contributed ideas. It [66:28] was just kind of a mind melt and it was [66:30] clear this is what it was going to be. [66:32] and um and and the top value was uh [66:36] whatever it takes. And then everyone [66:38] gave examples for what why that was. [66:39] Some had just come back um you know from [66:41] the front line in Ukraine. Some had not [66:43] slept for 48 hours um and and done a [66:47] whole bunch of things that are just um [66:49] amazing and impressive. And so that that [66:52] that was one that came out of it. [66:54] Another one that came out of it was the [66:55] truth is in the field. like it doesn't [66:57] matter how much you claim that your [66:59] system works um you know in in tests and [67:02] government tests etc. We found out it [67:05] has to work in reality and in combat and [67:07] and so the truth is in the field came [67:09] out of that and I was so impressed by [67:11] the team and I I was kind of thinking [67:12] like how do you how did that part of the [67:15] culture come about and it as far as I [67:18] can tell it came about because a bunch [67:19] of really good people got together [67:22] really cared about the stuff had [67:23] initiative and thought about it and then [67:25] they came up with it. It it it was just [67:27] that I think just getting really good [67:29] people together that have good chemistry [67:32] and when they when they sometimes maybe [67:34] have the sparks flying they can find a [67:36] way to work together again. That's what [67:38] creates culture. [67:38] >> So you have this idea to start this [67:40] European defense company. You have ideas [67:43] for the products you want to build. You [67:44] raise money. So then you go from the [67:46] idea to now your products are actually [67:48] being used in Ukraine for the first [67:50] time. Do you remember the first time [67:52] that you found out that your product was [67:54] being used there? where were you and [67:56] then what that felt like. [67:57] >> We knew from our team in the field and [68:00] um I do and everyone thinks differently [68:01] about these things. I think um and I [68:04] respect I think most attitudes towards [68:07] it. I I personally found it very [68:09] difficult. I think it became a more [68:11] serious company and like some of the [68:14] kind of light-heartedness uh you know [68:17] was gone a little bit because if you if [68:19] you build a video games company or a SAS [68:21] company you can be you can be quite [68:24] lighthearted about things and and [68:25] unserious. Yes, it's a commercial [68:27] business, but um you know, you you can [68:29] be it fits culturally if you are [68:32] deployed in what fundamentally is a sad [68:34] thing um this kind of conflict uh and [68:36] you contribute to it because you you [68:38] have to um because you can think it [68:40] through from first principles but in my [68:42] opinion it's still sad that it happens [68:44] um that makes it a serious you know [68:47] changes the culture and we are um I [68:49] think very aware of it and um that [68:52] doesn't mean that we don't want to [68:54] create the best systems of course we [68:56] Um, and doesn't mean that we are not [68:58] super ambitious. We are, but it is [69:00] serious. It's very serious. We don't [69:01] take it lightly. [69:03] >> All right. So, you've used the word [69:04] deterrent multiple times throughout this [69:05] conversation. So, that means the best [69:08] possible outcome for housing is nothing [69:11] happens. [69:11] >> Yeah. You don't want a conflict. And um [69:14] you know in general obviously the the [69:16] the reason why you build capabilities is [69:19] in order to avoid conflict not to use [69:22] them in a fight [69:23] >> which this goes back to this dinner we [69:25] had like 2 years ago this argument we [69:27] were having uh about this the fact that [69:30] war has been you know constant [69:32] throughout human history. So like what [69:33] actually has to take place in your mind [69:36] for that outcome to actually happen [69:38] where it's like no there is no more [69:40] human conflict. I [69:42] >> mean that's a big question. Um [69:44] I do think it's in human nature to um [69:47] have at least parts of the population [69:49] that want to have conflict that enjoy [69:51] conflict that want to uh you know [69:53] increase their um sphere of influence. [69:56] Uh sometimes you know that is political [69:58] sometimes it's geographical sometimes [70:00] you just literally want to have more [70:01] land around you. I think that is partly [70:04] cultural, but it's also partly hardwired [70:07] into a part at least of the population [70:10] and and and will probably be um genetic [70:12] in a part of the population. In order [70:14] for that not to dominate everything [70:16] else, uh you need to make sure that the [70:17] rest of the population is is [70:19] sufficiently protected so that it's too [70:21] expensive to try and you know use force [70:24] to take what you want. That's basically [70:26] what it boils down to. And if you could [70:28] eliminate um you know the outliers let's [70:31] say who who really are that aggressive [70:33] and want to use force then you know I [70:34] think forever you'd have peace it's [70:36] unlikely you can eliminate it. I think [70:39] it's more likely that you have to [70:40] contain them and I think that's okay as [70:42] long as you see it early enough that you [70:44] have to contain them and you and you are [70:46] able to look very prickly to them so [70:48] that you they don't attack. We mentioned [70:51] um your previous company a few times and [70:54] you said something when you guys were [70:56] developing the autonomous fighter jet [70:58] that you used a very similar technique [70:59] that Demis used to do Alpha Bow. I just [71:04] finished reading this biography of [71:05] Dennis. It's interesting to me that he [71:07] got started his first introduction to AI [71:09] was through video games. Yours was too. [71:12] Some of the best entrepreneurs come from [71:13] gaming. What do you think it is about [71:15] that? [71:16] >> Yeah, and I think arguably Palaki did [71:18] too. I mean certainly very close to the [71:20] gaming industry. Um the gaming companies [71:22] are complex to build because you you [71:25] marry art with technology and in many [71:28] cases um especially you know console [71:30] games you know back in the day today [71:31] it's a bit different with some of the [71:33] most I would say sophisticated IT um [71:36] projects um that that you could think of [71:38] and at the same time they had to have [71:40] artistic value as well and you had to [71:42] accommodate you know people from all [71:44] walks of life like um you know very [71:47] focused software engineers with um [71:49] [snorts] you know artists who um you [71:51] know just try to create something [71:53] beautiful and that is not dissimilar to [71:55] what happens in in the defense industry. [71:57] Um you know maybe a little bit less [71:59] artistry although that exists as well [72:01] but we have very different people um [72:03] that work on very complex systems that [72:04] have very different cultural backgrounds [72:07] and still need [snorts] to work together [72:08] to produce something amazing in the end. [72:10] I think that's that's one part and maybe [72:12] it's also exposure to um defense through [72:14] games. I don't know a lot of a lot of [72:16] games are, you know, have a have a [72:18] military military um theme. Maybe that's [72:21] part of it as well. [72:22] >> Explain the founding of your first [72:23] company like what was happening right in [72:26] your life right before you started that [72:27] one. Now [72:28] >> I was studying in the UK and um I I was [72:31] working on [72:33] in biology. I was working on systems on [72:35] simulations that um simulated neuronet [72:37] networks. [72:38] >> In what year? [72:38] >> The year was 1999. [72:41] Wait, [laughter] [72:43] you were working on simulating neuronet [72:44] networks in 99? [72:45] >> Yeah, using genetic algorithms. I was [72:47] looking to optimize them um to to do [72:50] something useful. [72:50] >> Why were you interested that to begin [72:52] with? [72:52] >> I liked complex systems. So, I was I had [72:54] just started a PhD in complex systems in [72:56] biology and neuronet networks are a [72:58] complex system and and specifically the [72:59] ones I looked at were fully recurrent [73:01] neuronet networks. So, once that um you [73:04] know with the right weights could start [73:05] to create spontaneous cyclic activity. [73:08] So they start [73:09] >> is that not what you were just described [73:10] on the fighter jet? [73:11] >> Uh well it has some similarity because [73:13] they the so genetic algorithms have [73:15] fallen out of favor as a search um [73:18] algorithm and now it's almost entirely [73:20] um reinforcement learning but you could [73:22] theoretically also use um genetic [73:23] algorithms. So there there's definitely [73:25] some similarity. The network [73:26] architectures are different. Um back [73:28] then I I based my network architectures [73:30] on on biological what's called central [73:33] pattern generators. So CPGs, they are [73:35] the the neural structures that create um [73:38] rhythmic movement for um you know for [73:40] for for walking, for your heartbeat and [73:42] and and a bunch of other things. And so [73:44] I tried to recreate one of these for [73:47] walking um using genetic algorithms and [73:49] a neuronet network, a quite simple [73:50] neuronet network to optimize the [73:52] weights. And in order to see whether it [73:54] worked, I hooked it up to physically [73:56] simulated creatures um you know very [73:59] simple four-legged and then two-legged [74:01] creatures. uh physics engines had just [74:04] started back then with you know Math [74:06] Engine and Havoc and and other companies [74:08] and to my surprise it ended up working [74:11] quite well. So overnight um I remember [74:13] it's quite rare that you have these [74:15] Eureka moments but I had this thing set [74:17] up overnight and those computers were [74:19] very slow back then. Um, and then the [74:21] next morning when I looked at the [74:22] experiment, I literally had this [74:23] two-legged creature walk over the screen [74:25] fully physically simulated. You know, [74:27] did a bit more research um, you know, in [74:29] that area and then realized this was [74:30] going to work very well and and thought [74:32] I'd spin out a company to use these [74:34] things as virtual stuntmen in in movie [74:36] scenes. [74:37] >> So, if I go back to you at 99, did you [74:39] have any inclination that one day you'd [74:41] be a founder, an entrepreneur? [74:42] >> No. No, not at that point. [74:43] >> What What did that person in in 99 think [74:46] that they were going to do? I thought I [74:48] was going to be a scientist. I I I [74:50] wanted to um to stay in in science and [74:53] and be a biologist and and specifically [74:55] do um software and simulation work in [74:57] biology. That's what I wanted to do. [74:58] >> Okay. But then you have this Eureka [75:00] moment and your first thought is I'm [75:03] going to start a company. [75:04] >> It was my first thought. It it took a [75:06] little bit of time but um this was a [75:08] time of the com boom and everyone wants [75:10] to start [75:11] >> 99 was an important point. Okay. [75:13] >> Yeah. And it was still um you know like [75:15] it it looked like it could do that kind [75:17] of thing. Um, and it was inspiring to [75:18] see lots of other people start [75:19] companies. And so I started Natural [75:21] Motion and um, we first used it, uh, you [75:25] know, we produced a software called [75:26] Endorphin that was used in in big [75:28] Hollywood movies like Lord of the Rings, [75:30] Troy, um, Spider-Man, Harry Potter, and [75:33] that was fun and and great, but um, we [75:35] didn't make much money uh, cuz we only I [75:37] think we charged $50,000 per license and [75:40] they ended up kind of doing all of the [75:41] scenes with them. Um, and it it was [75:44] clear that the real value was going to [75:45] be in using that technology for for [75:48] games in in real time. And uh, Rockstar [75:51] Games were one of our customers for, you [75:53] know, for the old tool Endorphin. They [75:55] asked us, "Do you have a version that [75:56] runs in real time on the upcoming [75:58] consoles?" And and we said, "Kind of." [76:01] And and so then we started a long [76:03] collaboration with with Rockstar Games, [76:05] who who ended up using our technology in [76:07] GTA 4, GTA V, Red Dead Redemption, and [76:10] and a whole range of other games. And [76:13] then that company grew and it grew [76:15] faster than before, but it still didn't [76:17] have the um the kind of takeoff, the [76:20] kind of liftoff characteristics that I [76:22] kind of was hoping for as a founder. you [76:24] always kind of obviously hope for more [76:25] of a hockey stick thing and and it felt [76:28] like the kind of chess pieces were kind [76:29] of drawing in and it was going to be [76:31] hard to move. So, so we ended up saying [76:35] like we need to make games ourselves and [76:38] and what you would now call vertically [76:40] integrated. We didn't think about it [76:41] that way. We we felt like we were so [76:43] inspired and learned so much from [76:45] Rockstar Games that we thought how hard [76:47] can it be? It turns out to be quite hard [76:49] but it ended up working quite well. We [76:50] acquired a games team um did the first [76:53] games, the iPhone came out and then we [76:55] had a string of hit games on iPhone and [76:57] then a big breakthrough um which which [77:00] was in a completely different area. All [77:01] of our previous games have been in [77:03] characters cuz that was our technology. [77:05] And then there was one game that I [77:07] really wanted to make cuz I was really [77:08] into cars which was a drag racing game. [77:11] Um so a car racing game without any [77:13] steering and and that became by far our [77:15] biggest hit Scott CSR racing and um [77:17] became the biggest game of all time on [77:19] iPhone at the time and it that gave us [77:22] the hockey stick. Then it was lift off [77:23] for the company and then it was off and [77:25] it became a completely different [77:27] experience. I I feel there's some like a [77:30] theme connecting here and I still don't [77:31] have an understanding of this. So you'll [77:33] see this in all these biographies that I [77:35] read, right? So I'm reading one right [77:36] now Raymond Plank writes it when he's 90 [77:39] years old and he starts this company [77:41] called Apache Corporation with $250,000 [77:43] in 1954. By the time the book is [77:45] published, you know, 50 years later, [77:46] it's like a $30 billion company. And his [77:49] insight was he was a tax uh tax planner [77:53] and accountant. and some of uh the [77:56] businesses that he was working for were [77:57] these new American oil companies and he [77:59] realized there's these huge tax [78:01] incentives that were taking place in [78:02] America at post World War II. At the [78:04] same time, you had tax benefits if you [78:06] would invest in oil discoveries because [78:08] the United States was trying to [78:09] incentivize domestic oil production and [78:12] at the same time uh because there's so [78:14] much debt after World War II, they [78:16] increase the top marginal tax rate to [78:17] like 90%. And so the very wealthy people [78:19] were trying to find ways to offset. And [78:21] that was the the original idea. Like he [78:23] discovered Apache Corporation which [78:25] winds up being this massive oil and gas [78:26] company because of the work that he was [78:28] doing in tax planning and accounting for [78:32] relatively small businesses. I feel [78:34] there is some kind of connection between [78:36] your first company in your second that [78:37] I'm trying to get to. Am I just wrong on [78:39] this? [78:40] >> So I don't think it was that direct [78:41] necessarily. One theme that I've always [78:43] felt with everything I ended up doing is [78:45] I I think is that I ended up doing [78:48] something slightly different to maybe [78:49] what the mainstream was. And part of the [78:51] reason is I've never felt kind of part [78:52] of the group of the peer group. Like I I [78:55] never as a biologist and I did a lot of [78:57] biology and I loved it, but I never [78:58] really felt like I was like part of the [79:01] biology group. And that made it a little [79:03] bit easier to do different things. [79:04] That's why I ended up just doing [79:06] software and simulation rather than go [79:08] out into the field and and watch animals [79:09] and um dissect animals. I never liked [79:12] that stuff. I wanted to do something [79:13] slightly different stuff with gaming. I [79:15] also never felt part of the the gaming [79:18] group and culture the same way. And as a [79:20] result, I think we ended up doing quite [79:21] different games that I remember like the [79:23] pricing for our games was very different [79:25] to anyone else's. We [79:26] >> the pricing [79:27] >> Yeah. The first games that we made, we [79:29] priced as low as we possibly could [79:31] because we felt that um we we just [79:33] maximized revenues that way. And [79:34] everyone else in the game industry said, [79:36] "No, your game needs to have a value. [79:38] Like your your mobile game needs to be [79:40] at least $5 so that people feel it has a [79:42] value." And it never made sense to me [79:44] because all I wanted to do was maximize [79:45] revenues. And I felt like um there was [79:48] so much price elasticity that we could [79:51] just go all the way down and people [79:52] would spend uh overall we would make [79:54] much more money because we'd have much [79:56] more people. So we we sold our game for [79:57] 99 cents, the cheapest that was [79:59] possible. And then when freeto play [80:01] came, we were the first big three [80:03] free-to-play 3D game, and we made it [80:05] free and then charge money elsewhere. [80:07] And that would have been much harder had [80:09] we been part of the group. And and I [80:10] think a similar thing is also true for [80:12] defense. It would be much harder for us [80:13] to build Helsing if the backgrounds had [80:17] already been in defense companies. Just [80:18] much harder. It's easier to be an [80:20] outsider. That maybe is a parallel. But [80:23] then the other thing is like I always [80:24] try to do things that are [snorts] that [80:27] are different and sometimes cool. Like I [80:28] wanted to like I really wanted to do the [80:30] number one top grossing game in the [80:32] world. I remember sitting like standing [80:34] in the shower at at one point in in the [80:36] UK and thinking I would like to do the [80:38] number one top grossing game. It was [80:40] just a goal. Um and and and with Helsing [80:43] I think we all want to build a company [80:44] that has impact. So I think that's [80:47] similar to you just kind of want to have [80:49] those goals. You want [snorts] to have a [80:50] bit of a northstar and then um and then [80:52] go for it. [80:54] >> Okay. I love that you use the word [80:55] northstar. The one other I think [80:57] probably the last thing I want to talk [80:58] to you about today is I've heard you say [81:00] this essentially this criticism of like [81:02] this LinkedIn or ex commentary culture [81:06] that commentary is actually relatively [81:09] worthless or completely worthless. [81:11] What's your opinion on this and like can [81:12] you just expound on that? Yeah, I think [81:14] commentary is entirely worthless and um [81:17] it has almost has no positive [81:20] contribution. The only thing that it [81:21] does is it uh I think it drags some [81:25] people down. Uh so it has a net negative [81:27] benefit and it obviously is is [81:29] encouraged through um you know the way [81:31] the social platforms are um uh you know [81:33] incentivizing contributors. So, you're [81:36] incentivized [snorts] [81:36] if you comment and particularly if you [81:38] comment negatively um but you're not [81:40] incentivized, you know, to be a creator. [81:43] And I don't complain about this um [81:45] because it is what it is and you know, [81:47] they should do what they want. But I do [81:49] think that creators are still somewhat [81:51] undervalued um and commentators are [81:53] still overvalued because they get so [81:55] much airtime um and and there are many [81:57] of them now. I think even more than than [81:59] before. And it it contributed to some of [82:01] the European I think um perceived [82:04] malaise um you know people talking [82:07] themselves and each other down. I just [82:08] said this the other day again I was just [82:10] in a car with someone who who's actually [82:12] is a great entrepreneur but he himself [82:15] just kept on commenting on how bad [82:17] certain things are but there's no point [82:19] in commenting about it. It's it creates [82:21] zero value. You either do something [82:23] about it or you don't talk about it. Um [82:25] but if you don't do um either um then [82:29] yeah I don't think it's I I don't think [82:30] it's a good contribution. Just [82:31] commenting on negative stuff is not [82:32] useful. [82:33] >> What would your advice be to do then? [82:35] >> Just to focus on building your company. [82:36] I I always think that um you have a real [82:39] privilege as a founder that you can make [82:41] a difference. You know depending on on [82:43] which sector you're in you can make big [82:45] differences in the way a society works. [82:47] Like arguably all the social networks [82:49] have made a massive difference for [82:50] better or for worse. All the AI [82:52] companies are making a massive [82:53] difference. many health companies. I [82:55] think defense companies are making a big [82:56] difference too. Uh and if you are a [82:58] founder that um is very ambitious and [83:00] you're in one of those fields, you can [83:02] change society and not many categories [83:05] of people can do that. Um politicians [83:06] can as well. Certain types of other [83:08] people can as well. Sometimes even [83:10] investors, but um founders I think have [83:13] the biggest lever. And if you that's [83:15] also why you have so few founders I [83:17] think on Twitter and and uh and and [83:20] other social platforms because it's [83:21] already very satisfying to do the stuff [83:23] that you do. You don't then need to [83:24] comment on how bad everything is. You [83:26] just do your stuff and it has an impact. [83:28] >> That's the perfect place to end. Thanks [83:29] for making the time. [83:30] >> Cool. Thank you. [83:31] >> I hope you enjoyed this episode. Please [83:32] remember to subscribe wherever you're [83:34] listening and leave a review. And make [83:36] sure you listen to my other podcast, [83:37] Founders for almost a decade. I've [83:39] obsessively read over 400 biographies of [83:42] history's greatest entrepreneurs [83:43] searching for ideas that you can use in [83:45] your work. Most of the guests you hear [83:47] on this show first found me through [83:49] founders.