Invest Like the Best with Patrick O'Shaughnessy - Anduril: Building the Future of Defense - [Business Breakdowns, EP. 59]

Episode Date: May 26, 2022

Today, we are running a special episode of Business Breakdowns. With geopolitics playing an increasingly important role in society again, this episode with Anduril’s CEO offers an inside look at the state of the defense industry and how it is changing. If you enjoy this episode, subscribe to Business Breakdowns on your preferred podcast player, where you’ll find past episodes on Block, Goldman Sachs, AutoZone and many others.   Today, we are breaking down Anduril. Anduril builds high tech defense systems for the US Department of Defense and its allies. Crucially, it does so with speed that emanates from Silicon Valley. Founded in 2017 by Palmer Luckey, who previously built and sold Oculus to Facebook, Anduril has achieved the rare feat of challenging the established order in the defense industry.   To break down Anduril, I’m joined by the company’s CEO and co-founder, Brian Schimpf. We discuss the history of the defense industry, how Anduril’s business is counter positioned against the legacy cost-plus model, and what Brian has learned about selling to the DoD. Please enjoy this breakdown of Anduril.   For the full show notes, transcript, and links to mentioned content, check out the episode page here.   -----   This episode is brought to you by Tegus. Tegus is the new digital hub for market intelligence. The Tegus platform empowers Investors and Corporate Development teams to invest smarter by pairing best-in-class technology with the highest quality user-generated content and data. Find out why a majority of the top firms are using Tegus on a daily basis. If you're ready to go deeper on any company and you appreciate the value of primary research, head to tegus.co/breakdowns for a free trial.   -----   This episode is brought to you by Daloopa. Daloopa streamlines a major pain point for investors. By capturing all of a company's KPIs and adjusted financials into their database - Daloopa makes it easy to quickly update your models for what matters. Daloopa uses AI to find every KPI disclosed - from charts, to text, and even from footnotes of investor presentations. Daloopa updates these KPIs and data points in your existing Excel models in one click, regardless of your source or format. Test Daloopa for free at daloopa.com/Patrick.   -----   Business Breakdowns is a property of Colossus, LLC. For more episodes of Business Breakdowns, visit joincolossus.com/episodes.   Stay up to date on all our podcasts by signing up to Colossus Weekly, our quick dive every Sunday highlighting the top business and investing concepts from our podcasts and the best of what we read that week. Sign up here.   Follow us on Twitter: @JoinColossus | @patrick_oshag | @jspujji | @zbfuss   Show Notes [00:02:52] - [First question] - The history of defense technology and the technological and competitive landscape when he set out to build Anduril [00:08:22] - What the early experience was like when approaching the government and finding an early adopter [00:12:44] - Necessity being the mother of invention when it came to developing drones [00:16:37] - What it’s like to develop hardware and software products at the same time  [00:24:44] - The state of military technology and military conflict today writ large [00:31:10] - Are we heading to a future where warfare is mostly machine against machine?  [00:33:34] - Comparing the ghost drone system to predator drones [00:38:40] - Guiding principles as a firm and deciding on their product roadmap [00:43:25] - An overview of their product lineup and what they’ve built so far  [00:51:56] - Most difficult decisions he’s had to make through Anduril’s history  [00:53:51] - How he overcame Anduril’s lowest points and biggest challenges  [00:58:38] - Thoughts on effectively compounding hardware innovation  [01:02:23] - A moment he’s most proud of and regrets most in Anduril’s history [01:04:20] - Lessons learned from observing Palantir and SpaceX  [01:08:37] - The kindest thing anyone has ever done for him

This episode is sponsored by Teegis. Teegis is the new digital hub for market intelligence. The Teaguez platform empowers investors and corporate development teams to invest smarter by pairing best in class technology with the highest quality user-generated content and data. TIGAS content is powered by many of the world's leading institutional investors, where their expert calls are recorded, transcribed, and uploaded to the shared platform, leading to the highest quality content and datasets. Tecis also recently acquired BAMSEC, which will allow users to see,

seamlessly toggle between financial data, management commentary, and expert interviews as they get up to speed on a company. Any customer who signs up for TIGIS before May 31st will receive a free BAMSEC license as part of their subscription. Find out why a majority of top firms are using TIGIS on a daily basis. Head to TIGIS.com slash Patrick for your free trial. This episode is sponsored by DeLUPA. DeLUPA streamlines a major pain point for investors by capturing all of a company's KPIs and adjusting

financials into their database, DeLUPA makes it easy to quickly update your models for what matters. So many investors I speak to complain about juggling multiple company earnings or rushing to ramp on a new investment. Delupa uses AI to find every KPI disclosed, from charts to text and even from footnotes of investor presentations. Delupa updates these KPI's and data points in your existing Excel models in one click, regardless of your source or format. Try Delupa for free at dilupa.com slash Patrick. This is business breakdowns. Business breakdowns is a series of conversations with investors and operators diving deep into a single

business. For each business, we explore its history, its business model, its competitive advantages, and what makes it tick? We believe every business has lessons and secrets that investors and operators can learn from, and we are here to bring them to you. To find more episodes of breakdowns, check out join colossus.com. All opinions expressed by hosts and podcast guests are solely their own opinions. Hosts and podcast may maintain positions with securities discussed in this podcast.

This podcast is for informational purposes only and should not be relied upon as a basis for investment decisions. Today we are running a special episode of business breakdowns. With geopolitics playing an increasingly important role in society again, this episode with Anderil's CEO offers an inside look at the state of the defense industry and how it is changing. If you enjoy this episode, subscribe to business breakdowns on your preferred podcast player, where you'll find past episodes on Block, Goldman Sachs, AutoZone, and many others. Today, we're breaking down Andrel. Andriel builds high-tech defense systems for the U.S. Department of Defense and its allies. Crucially, it does so with a speed that emanates from Silicon Valley. Founded in 2017 by Palmer Lucky, who previously built and sold Oculus to Facebook,

Anderil has achieved the rare feat of challenging the established order in the defense industry. To break down Anderle, I'm joined by the company's CEO and co-founder, Brian Schimph. We discussed the history of the defense industry, how Anderl's business is counterpositioned against the legacy cost plus model, and what Brian has learned about selling to DOD. Please enjoy this breakdown of Anderl. So Brian, ever since we met dinner, I don't know what it was two months ago now, I've been really looking forward to this conversation. I've got a weird fascination with military history.

My best friend growing up was a Special Forces officer. I've just always been around this world without being personally in it. So I'm really excited to try to suck as much information out of you today as I possibly can. And given that Anderl is kind of like Tesla, the first new car company that made it in 100 years or something, something similar could be said of Anderil in the world of defense. I thought an interesting place to begin would be to have you give us a bit of a history lesson of that industry, as you saw it maybe when you and your co-founders started Andrel. People have heard names like Lockheed Martin and Northrop Grumman,

but give us from your perspective what the history looked like from the seat of trying to start a new company in this very intimidating space. Might actually rewind a little bit back to 1950s era. I talk about a lot of the projects there that were incredibly successful. So defense was this sion of being able to move fast, do really innovative things where all the new technology came from. you look at a lot of the past successes, even relatively trivial things like Pentagon, largest office building in the world, built in 13 months, built anything in 13 months these days.

You look at ICBMs and the intercontinental ballistic missiles, a lot of the first rockets. Those were done within the DoD in a very short time sphere, just a couple of years, going from not existing at all through to having these incredibly capable systems going very, very quickly. We built something like 50 different airplanes and configurations and things like that over just a couple decades period. Skunkworks was legendary in everyone's mind around building the most advanced technology, pushing the limits, and they did this all with pencil and paper. So there was this period where this was just an incredibly innovative industry, where the best and brightest went to work. It was really generating a lot of novel ideas, the new R&D. It was just moving so quickly,

such an impressive pace, building such impressive technology. Throughout the Cold War, that was really the case. Things started to slow down late seven, or early 80s, where DoDD went to a model of much more cost controls, very predictable schedules, trying to drive a little more discipline into the organization, which was hard to argue with, but the result has been just a lot more bureaucracy, a lot more focus on process over what is the most innovative. And then you fast forward through to the 90s, and there was this dinner where they called it the Last Supper in DoD where there was, I think at the time, something like 50, defense companies that were considered significant players in the space. I think it was the

Secretary of Defense at the time said, we are going to go through a period of consolidation, budgets will not be growing, you guys will have to consolidate. I leave it to you to figure it out. And that entered about a decade-long period of intense consolidation, where now you end up with about five, six major players capturing the vast majority of the DOD's budget. So you have big names like Lockheed Martin, Northrop Grumman, Raytheon, lesser known ones like General Dynamics, Huntington Ingles, there's a variety of these companies that have consolidated down nearly every player in the market. And you have this big capture with just relatively few players winning most of the awards.

If you're an engineer going into this space today, you may work on one airplane in your career. If the timelines have gotten longer, you look at the F-35, this was started in the early 90s, I want to say, was when they started looking at this. It's going to cost over $1.5 trillion. over its life cycle. The timeline to get fielded is getting insane. They're putting a new nuclear attack submarine in the water with the expected timeline of 2035, with a lifetime running through 2085. I can't imagine doing technology on this time scale. So the desire for more predictability, having this very bureaucratic approach, I think, is incentivized both the government and the defense

industry to take this very slow, steady, measured approach. And it's really showing at this point. What shifted compared to 10 years ago was there was a belief at the time, early 2010 timeframe that the current players could really solve the problems. It just was a matter of different process or different acquisitions. I think today there's less of a belief when you start looking at modern software problems, looking at AI, how you would actually make the stuff more affordable, how you would have a more software-defined approach. There isn't that belief anymore. There's been a real shift that the current players do not have the, It's not even necessarily skill set and talent.

It's really the process, the way they think, the way they build is tooled to these very expensive, very large, very time-consuming, big airplanes, big ships. And that software necessitates a different approach and a different strategy. And that wasn't obvious 10 years ago, but now it is today. And everyone agrees with this, that there has to be something very different. There has to be a new model. And that's what will enable us to win. The speed, the pace, different way of building, different way of iterating, different

way of figuring these things out, separate from what we need to do. to build a fighter plane or something like that. So it's a very large scale shift that's happened. And this is very recent. This was just not the case 10 years ago. There was still a belief it would all work, but with the rise of more autonomous systems, more intelligent systems,

and seeing what Silicon Valley has been able to do with modern software approaches, it really has changed the way DoD thinks about how they will build and what types of companies will enable them to be successful on some of these different aspects compared to the past. What was that like being a part of those minds

changing. We were at dinner with a former J-Soc commander when we met, and it struck me that it was incredibly important that you had to try to use military terms, but get a small beachhead inside of DOD in order to demonstrate this new style of building, deploying, updating, et cetera, the Androo represents will come to that in a minute. Say a bit about what that early experience was like, I want to come back to why this all happened. It's such a fascinating history. But first, what was it like being around that and a part of that change? And what was that being. chat. How did you do it? So I think there's a lot of push for taking new innovative approaches and the government gets beat up a lot for not having more innovation and not doing things more innovatively.

They're certainly not blameless in the matter. But on the whole, people do want to succeed. They do want to move fast. They do want to field exceptional capabilities. It's just not obvious how. And our view to this has always been you're going to have to find these early adopters, the people with an urgent mission to solve that need to do something quickly. and are willing to take the chance on a new entry, or willing to take the chance on a new way of building because their problem necessitates it, because they need to do so.

And then from there, to get success, people start to see models that do work, it starts to become de-risk. And a lot of what we view ourselves responsible for is not just building the technology that can actually solve these problems, but actually helping think through the policy changes that need to exist,

how you can acquire this differently, how you can run contracts in a more faired, objective way, how to use more test and evaluation instead of our company investment directly. All these things we think are net good for the department. They don't apply to everything, but they apply to a pretty broad range of problems. So for us, it's always been about finding those customers who have that urgency on a problem that requires a different approach. Probably the first real beachhead within the Department of Defense was working on counter drone systems where this is a threat that was very urgent. It's one of the first times the U.S. has

had adversaries that can attack them from a distance where being at bases was not necessarily safe anymore. And we sought first with the special operations groups that are deployed actively that the folks who are going to see this threat first. And we started out, and everyone was looking at it from this problem of, okay, we have these small commercial drones, think DGI drones. That's what the threat is, an IED-style threat, dropping hand grenades from these things. That certainly was one of the risks for sure. But as we were deployed and as we saw this evolve, you start to see what you're seeing in the news now, where there's significant numbers of Iranian-backed rebel forces getting higher and higher on drones and really evolving very quickly on their tactics, how they approach this, what sort of

munitions they have on it, what sort of protections. And this necessitated a strategy, which was taking a very evolutionary approach, where instead of just saying traditional military approach for this would be we're going to write the requirements. We're going to define exactly what the problem is. We're going to then go send it out to bid. In five years, we'll have a system. We'll put it in production, we'll ship it, it'll all work out. We tried this with the IED threat back in the day, and it didn't work.

What we believe does work for a threat like this is you need to have a constantly evolving strategy. You need to say, I need a term they use as a system integration partner, someone who can look at this, understand the threat, understand the state of technology, and develop out a roadmap of how do you actually continuously address this, adapt, and push out new capabilities according to what we're seeing in the wild. And the key to this is also really doing this in a very software-first way. So you're not going to be able to get massive new hardware systems out in the timeline that's necessary. But can you push out new software modules over the air? Can you monitor

constantly how it's doing? Absolutely. It's the same model that we've come to take for granted with our phones, even with things like Tesla autopilot. It's just going to keep improving. And you you're going to learn and adapt as you move through this. So that model seems so straightforward in the commercial world, but it's very different than how the DoD has run historically, especially as you start talking about these integrated hardware software systems. So finding that first customer that had that urgent problem, actually really wanted to solve it,

and had the risk tolerance to actually work with someone like us, was absolutely critical. So we're hugely grateful to be able to work on this problem and the opportunity to make a difference on it. Through both your history and this other bookend, your recent experience, it just strikes me this idea of necessity being the mother of invention is just so true, and that the Manhattan project happening as fast as it did or some of the examples you gave were in arguably more dire conditions than when the U.S. was the clear global military hegemon and a lot of this

bureaucracy seemed to have set in. Maybe you can walk us through with that necessity being the mother of invention lens. The idea that the drone itself would be an area of first focus. I want to talk about some of the more interesting stuff, including the autonomous sub that is in the news recently for Mandrill. But the drone was the original thing that you focused on. Was that an easy decision? Were there other things that were potentially considered as the first piece of hardware technology? Just walk us through the story of creating this first product for Mandrill when you did. One of the key areas we've focused on throughout is what's the software platform that will enable us to solve a variety of DOD problems. So being kind of a very

software-heavy company, one of the views is, well, what's worked in, let's say, the AWS of the world and similar, or Amazon's or any of these companies or Googles, you have these very general purpose software platforms that you can apply to a variety of problems. These are very hard to build. They're very expensive, but they enable you to solve a wide array of problems in a very connected way and have everything accrete with each other. So from day one, we knew that a key part of this was going to be how do we apply the latest in computer vision and sensor fusion to actually understand what's going on in the world.

So this can be detecting drones. This can be detecting things underwater. It really can be any sort of these applications where you're processing sensor data, you're making sense of it. You're trying to automate as many of the simple parts as possible to give humans the right information. Then the other side of this is then how do I actually affect this in the world? This could be deploying one of our interceptors, Anvil out, to go knock a drone out of the sky. This can be steering a sensor to go look and interrogate and get more information to tell me what exactly this is.

We can take a variety of forms depending on the process. problem set you're working on. So throughout this, the software layer we call lattice was something we knew from the beginning was absolutely necessary and was the only modern way that you're going to be able to build the interconnected future that you really need to have. And again, this sounds so obvious, but let me contrast this to how aircraft were built in the past. Boeing has multiple aircraft programs. That software stack is completely independent for every single one of these. The idea that you would share software across these is quite foreign. It's almost more incidental than it is intentional.

So this is not the norm. The incentives are not set up to do this. How they contracted, how they price this, it doesn't fit in the normal model, but it's necessary for how you go about these things. So then in terms of applications and problems to work on, there's a variety of things we kicked off in parallel, but I think what has worked for us

is recognizing where there's that problem set, where there's early adopters, and leaning in on someone who wants to move quickly. So recognizing that and being able, to say, hey, we think we have tech that can solve this problem. There's an opening here. Let's move fast against it. It's something that's very, very important for us. And I think that's easier said than done in a lot of ways. Recognizing where you will have that traction and being able to lean into it is not obvious. It takes some experience inside the department knowing where these

things can move, who can go against it, where there's a lot of people who work in the science and technology space doing early research that have that easy entry, this is where a lot of innovation stuff happens, but getting into fielding is very, very hard. So finding those windows is absolutely key. So for us, fairly early on, we saw this as an emerging threat. We knew this was something that traditional approaches of just a single sensor that'll solve everything or a single system that will solve everything will not work. And our style of integrating a solution together with this unique software approach was absolutely key. Then actually manifesting that into a successful working relationship took a fair bit of luck and a lot of hard work to get there.

Can you talk a little bit about what it's like to develop software and hardware together, understanding that you wanted lattice, the software platform, to be this fundamental infrastructure layer on top of which you could build apps, which would be the different kinds of drones, the submarine, whatever. Really interesting switchup relative to the Boeing example. You've seen other people talk about how the best software comes from people that also make hardware, Apple being the prime example. What's your experience been like there?

You ran engineering at Palantir before starting Anderals. So what lessons have you learned about co-developing? software and hardware together. There's two reasons we went after the hardware side as well. Probably the more exciting one is what you're describing, which is being in control of our destiny and being able to shape the hardware to really take advantage of what we can do in software and vice versa is absolutely key. So simple examples of this, being able to design the processing to take advantage of the cutting edge GPU. So everything we have to do is done on the edge in these very austere conditions and deserts

and very limited connectivity, often running off solar power. It's a very hard problem for AI inference, for example, to really work on. So being able to constantly be able to adopt the latest and greatest embedded GPU while it's really evolving what the software can take advantage of with that

has been key. Thinking about this for drones, or this ghost drone, we designed that thing to be as simple and modular as possible. Or you've got a flying stick and you can bolt things onto it that actually do interesting stuff. But being able to then design and say,

hey, here's what I think is achievable in terms of computer vision performance or what's the very simple API I would actually want to be able to control this drone. What information can I even get from my batteries? That's actually like really hard. Being able to pull that information out is very important. When do I need to land?

How much power am I consuming? How do I optimize for this? All these factors start to really matter. If we were to go at this from a counterfactual perspective and say we didn't control the hardware, we'd have to then go get permission, or get people to buy in to using our software. And honestly, what's very hard there is our vision for how the future will come together

with larger quantities of smarter systems that are cheaper. That is not the normal business model in DoD. So you'd have to convince someone not only that your software is compelling, but that their business model is probably not going to work, and that you need to take a different approach here. It's very, very hard. The timelines are long. And this is not an industry that's incentivized, generally work on IRAD and

pursue a vision. If there's not a customer signing up on a contract, it's very hard to get traditional players motivated to work on these problems. The other aspect of this is DUD still really struggles to articulate a software business model that is very effective. So where people have converged on software as a service and will pay what is comparatively a very cheap cost for licensing this product compared to me building it myself, I will then in turn get a working product that actually solves my problem like Slack or office or any of these things, that model is not very typical in DoD, where they typically want to, and often for good reasons, custom build the solution, own it themselves, and it ends up turning very quickly into a services and labor structure.

It's very hard to have a software business model in DoD. Hardware is very straightforward. They understand why it costs what it costs. It's countable. They understand the value it's providing before it was there, and now it is not. They understand why it would cost money in the future to continue to build this. There's a very straightforward model there that allows you to scale very, very quickly in a way that was very challenging just selling software in past life. So there's a lot of reasons why we went down this hardware path, partly business, partly to be in control of our destiny, partly to build the tech we think needed to exist. But I think it's proven to be very, very successful in showing what's possible and being able to just manifest something that otherwise would be

very misaligned and very hard to create. I don't want to lose the opportunity to really contrast this more SaaS-like modern business model of Andrel with the legacy defense model and very specifically the notion of cost plus. And maybe you could just walk us through, how does the defense business complex work from a business model standpoint? Because obviously, Anderle is very uniquely counter positioned against that business model. And it evolved the way it did for reasons you discussed already. But give us the detailed cost plus model and what you think it leads to, which sounds like it's super long lead times and very expensive unit costs on something like, say, the F-35.

So the cost plus business model, this is the significant portion of the DoD contracts, is you run this cost accounting system. So you track every hour, every piece, every PO, everything you're doing, and that tracks your costs. And the cost plus fixed fee is then on top of it, you get a fixed profit. So this is somewhere in the range of 7 to 12 percent, typically. look at a defense business, the operating margin will be in that range, somewhere around 10 to 15% sitting around there. So when you're developing a product, the government just reimburses you for what it costs.

So you spend engineering hours, you spend R&D materials, you get paid back for that. The advantage, well, depends on whose advantage it is. So one, it really shifts the risk of the contract delivery from the contractor to the government. So the government's absorbing all the risk. This isn't possible. It takes too long. They change their requirements. they change what they want, they want it more complicated.

It just costs the government more. So the contractor doesn't end up bearing any risk. They're going to get a fixed profit regardless. But you can see the perverse incentive here where there's no risk and has no particular incentive to move faster and actually get these things done on time. So the alternative to this is what they would call a firm fixed price. So this is what you'd expect in the commercial world. I write a contract.

You're going to deliver me X and I pay you a fixed fee. If you can do it cheap, great. And if you can't, you lose money. It seems like a very straightforward model. So that transfers all the risk to the contractor to actually be able to deliver. Fortunately, I think a larger number of contracts are being written in this way where there's pros and cons to this. There was recently a news article where Boeing has bid a number of these firm fixed price contracts, but because of a strategic calculus of underbidding, which works when you do a cost plus contract,

then they're losing a lot of money being able to actually deliver these things. And when you're tooled to sort of not require that efficiency, it's corrosive. And it's not malicious. It's not people trying to do better at their jobs or anything. It's just the nature of the incentives. And incentives do matter on these things. You even see the director of NASA saying these cost plus contracts are a huge anchor on them right now. These are causing a huge problem for them.

And you have companies like SpaceX doing things from fixed price and they deliver on time and it works and everyone's happy. It just seems like so clearly a better model in the long run. And that's how we work. The other idea here is we want to invest in our own research and development. Most defense companies spend a very low percentage of their budget on research and development. The even crazier thing is you get to build that as a cost. The government pays for your internal research and development. Your fee is on top of that, which is wild to me.

So most of these companies do relatively little internal research and development. And to me, this is indicative of you have this belief that the money I am taking in, that profit I'm taking in is a higher return for the company by just straight giving it to shareholders than me driving growth, new contracts, new ideas that will make my business successful in the future. And I certainly understand why a lot of companies have gotten there. When this period of consolidation, unclear if there's a real threat, I certainly understand why these companies have moved to more of a financial engineering and dividends view. But as the threat changes, as the problems change, it's hard for me to believe that not investing

in these areas that not trying to lead the way on new technologies is the correct move. So for us, very much, everything we've done is we're not going to work on a problem unless we're willing to put our own skin in the game on it. We'll have to have a lot of conviction on the roadmap, what's going to be successful, and what will be able to scale and what's going to be a real capability. And everything we do, we have to have that conviction in because we really do want to put our own skin in the game on it. I think it just aligns our incentives, our goals, and gives us the right motivation to succeed. You said something there, which reminds me of a question, I should have asked maybe as the first question, which would be for you to give us a state of military

technology and military conflict as it pertains to technology today. This is especially important with what's going on in Europe, with Ukraine, with Russia, with our involvement with it. You see bills going to the Congress floor about providing tons of additional aid. I would love your take on the state of conflict and where you think it's going, because obviously all of this only matters if things continue to evolve and you're there to evolve along with it, which I think is the whole point of Andrel, precision and technology and warfare is an important topic that we'll talk about a little bit later, but the stakes are high. Give us just your state of conflict and technology's role in it today because I think that's a really important area to explore. There's two views to these

conflicts. Maybe there's like a day one framing, which is you're launching a campaign. How do you mass a ton of force to be able to overwhelm the adversary, take out key communications and missile systems and defensive systems, how we did Gulf War, where it was a very large-scale day one campaign, and you just roll in with an overwhelming force. And that's where things like stealth bombers, stealth aircraft, long-range precision munitions, all of these things become very, very relevant, where you're going to have to fight at a distance. And the U.S. by necessity is often projecting power. We're not building up on the California coast. It's always we're doing something forward, and that is a very, very hard problem to solve.

That's the day one view. A lot of technologies built up around this day one view of how do we have that overwhelming force to make that battle very decisive very quickly. There's a good calculus for that, which is you want to make it so that any sort of large-scale conflict could be decided quickly to be very deterrent. You want to make it so that the adversaries in the world think twice about whether this will succeed, whether my aims will be met. So there's a very good argument for having these very exquisite, very high-end technologies.

So a lot of the stuff you see around the next generation pieces are really about that. So how do I have that overwhelming force, that overwhelming mass, and early days of a conflict when I have the advantage? Then I think these conflicts tend towards the extended phase of these things where now it's about ground forces. It's very tactical. It's very extended. And I think that's going to be any sort of conflict in the future is going to end up looking like that. So areas of technology that seem to continuously surprise people. surprising that it's surprising, is the efficacy of unmanned drones at all levels. I want to say it was about two, maybe four years ago. I'm bad at dates. But there was a conflict in Armenian-Azerbaijan, where the Azerbaijani's had Turkish drones. So this Turkish birakhtar TB2 drone was one of them. They had Israeli munitions, loitering munitions, precision munitions, things like that. They even had some very low-tech things. They took old Antonov-A-N-2 biplanes. made them remotely piloted and used those to then fly in and stimulate the Armenian air defense systems, which were Russian air defense systems. And the Armenians had tanks. They had armor.

They had air defense systems. They had all these capabilities there that were very traditional way of viewing how warfare would play out. And it was a one-sided conflict. The Azerbaijani's were completely decimated, the Armenians, and took very few losses because they were able to fight this in a very remote way, being able to use these advanced UAVs and munitions and things like that to conduct strikes, to take out adversary positions in a very effective way that was very hard to hide from. It's very hard to defeat. There's a lot of these things. They're cheap. Your missile systems are going to get targeted early. It's a very hard threat to deal with. You look at the Ukrainian fight, and in a lot of ways, it's similar, where I would not want to be someone in a tank

in a modern conflict. It is not a good place to be. The ability to use these conventional, large land forces is very hard. You look at what's effective. It's highly disaggregated. It's having small units be able to have those precision munitions, be it javelins or stingers, utilizing small drones to be able to target, strike the adversary very, very fast. And you see a lot of the electronic warfare showing up as well, where there's constant jamming, interfering with GPS, all these things that have been reported, those are all playing out. So you sort of have what I think any of these protracted conflicts looks like, which is this very tactical, very disaggregated fight. And I think that's going to be the hallmark of what

a lot of this looks like. Now, if I'm America, if I'm Taiwan, if I'm Ukraine, if I'm Poland, the Baltics, I know I need to have that defensive capability to fight in a very broken down way. my big central infrastructure is going to get attacked on day one. I need to be able to disperse. I need to be able to put up a huge fight, and I need to be the biggest bone to choke on possible. And that's something that I think we've believed from the beginning, and a lot of the technology we're working on

is very dialed towards that type of fight, where it's very defensive in nature, it's shorter range, but it's the sort of things you need to pull back an invading adversary to really actually defend yourself. And those are the sorts of things that we think are going to be very necessary. hearing the future. It's going to be less super expensive, relatively few systems that you want as the fight progresses. On day one, those are still very relevant. But as these actually happen, these are the capabilities you need. And I think demonstrating that, showing that training to that, we'll give

these countries pause when they think about taking aggressive action because the cost will be so high and the likelihood of success low, where it's very unclear, will you be able to sustain an offensive? Will you be able to actually meet your aims. And I think Russia miscalculated that. And I don't think there's any reason for them to believe it would be any easier with any other country around them. I certainly wouldn't want to go against Finland. I don't think it would work out very well. That's the type of tech. That's the type of problems that we think we need to solve. You hear a lot of talk of sending like eight mig 29s to Ukraine. It's like, it won't matter. They will get shot down fast and that will be it. And it's just not the mass and quantity you need. The logistics train, the fueling, the maintenance.

It's so hard to sustain these things. You need cheap, tactical, very disaggregated capabilities, and that's really what's going to enable to fight once you get past this initial volley. Do you think that in the future, most conflict will just be machine on machine? Is there some point that we reach where little to no loss of human life? And obviously, I'm probably being naive not thinking about the people in power that could create loss of life as a strategy. But is that the arc of military progress, do you think, that? 50 years hence, let's say, a conflict like this would be entirely waged for machine supremacy,

and then that's clear to demonstrate and that's sort of the end of the conflict? Do you think that's where this is going? I certainly think that's one outcome, but you get to the question of what's the ends you're going for. At the end of the day, you look at Russia, you look at any of these things, it's not merely to show that they could destroy Ukraine military capability. It's to invade and take over Ukraine. And I think invariably you get to that place where the ends of these conflicts often involve change of political leadership, changing of territory, all these things. In those ends, I think frequently there will be a loss of human life, unfortunately. I think you will end up with troops on the ground. You will end up with some sort of actual campaign to affect the change you're

looking to affect. And that will invariably require some sort of human engagement at that point. I think you can make the likelihood of success of that very low. Having that as your deterrent effect is often how we think of this. The ends for all of this military tech is not to kill more people or anything like that. It is always about how do you minimize the probability that a conflict happens in the first place. So I think the reality is going to be, yes, there's going to be a lot more machine on machine that's going to happen. It's going to change the nature of how these things are fought. But I think invariably there will be a human element.

and there's still going to be the aspect of destruction of buildings, property, infrastructure, refugee crises, all these things are still very bad. There's this view that, okay, once it's fully machine-on-machine, there's no risk to human life, there'll be more conflict. I'm not sure I agree. I think the human suffering will still be very bad. I don't think there's any world where anyone should believe that these conflicts become less problematic.

I think you want to get to a world where one, countries can truly defend, themselves, the more powerful invading countries do really understand the consequence. They probably will not achieve their ends. That's what I think will create real deterrence on this. It's been amazing to watch the sometimes decades-long shift in policy choices or decisions, whether it's in Germany or Finland or even in the U.S., around this conflict in Ukraine. I think a great way to really drive that point home is maybe to compare the ghost system with something like Predator.

The comparison, I think, could be an interesting way to understand these two differences. People, I think, will be familiar with the predator drones, especially around our Middle Eastern operations post-9-11, and viewed that at the time, probably rightly so, as this cutting edge of technology. So I think it would be interesting to contrast a drone against a drone, to understand what the difference in the systems are, how they're operated, etc. Could you draw that contrast for us between what ghost represents and what predator represents? Yeah, it'll be a little bit apples to oranges simply because Predator was meant to be very long range, very long endurance, carry a lot of sensors and munitions and things like that.

Ghost is a small helicopter fits down into basically a gun case, can be carried, set up very simply, and launched, think in the order of a dozen miles sort of range, much shorter range capability. It's probably worth highlighting the state of how these larger drones are operated today. So you have Reaper, you have Predator, you have a global hawk. These are all the well-known military drones today. It's better to think of them as remotely piloted aircraft. So these things have some sort of radio connection, be it through satellite, through a line of sight link, and they are flown by someone sitting in a container-sized

ground control station where they have a stick and they're flying it like a plane. There's a sensor operator sitting in there as well, moving around the sensor, looking with the camera, moving these things around. So that's two people. Then there are about a dozen more behind that. So then you have people sitting there

and exploiting all of this sensor data. So they're taking screenshots, they're building up analysis decks, the building up products of all of these things and disseminating them out, then for each of the sensors you'd have on there, you'd have a specialist managing that. So maybe you have some specialized payload

that deals with radio frequency information. You'd have someone exploiting that specifically, repeat, repeat. So you end up having this very large tail to actually operate these things, plus the ground crew, the large scale infrastructure to deploy these things for larger drones to a degree in necessity,

but certainly from the operation side is the very manual process. When we've thought about this, particularly you start getting down closer to these small units, this company level, maybe a dozen people size, say, you can't afford to have multiple people pulled off to be manning and operating a drone. So what we've really focused on from day one is how do I make these things more autonomous and more intelligent so that you can just absolutely minimize the manpower required? On top of it, you want to minimize the logistics and complexity of how do I deploy these.

So when we think about flying ghosts, you tell it what you want to go look at. You just say, here's a building, go look at it. Plots, of course, flies it automatically, takes off, just goes, gives you footage of it. You can tweak it and adjust and task it to go do different things. But it's closer to you tasking a pilot than it is a pilot flying a drone with a joystick is then the idea. And then when we deal with the sensor data, it's like,

how can I make this as simple as possible for you? Maybe you're looking for tanks. Okay, I can go out and autonomously scan an area. with a team of drones to go find tanks. That seems like a very straightforward problem that software can solve at this point, and that's how we view this. So we've tried to have this very autonomous,

very light footprint view from day one, and everything we're building has that view, which is often your limiting factor of how can you deploy these is how many people can I have supported? What's my communications bandwidth? That's often your limit, even beyond just simply budget and how expensive is it, which is a big factor as well,

but it's often the manning, the operations, logistics complexity that the U.S. faces, especially since we have to do this power projection, we have to move forward. That becomes the driving constraint. That's a big part of what we thought of. Whenever we design a UAV, we've got some larger UAVs we're working on as well, nothing quite in the same size and duration that these bigger ones can fly, but a big step up from where we are today. And again, same idea. Just because it's bigger, should still be operable with just one or two people. That should be the goal all the time. They should operate a team of these. one person operate 10 of these instead of 20 people operating one.

That should always be the goal. Again, this takes a lot of expertise on software. You have to look at this as a software-first problem. How do I minimize the overhead and complexity of this? I think there should be a rule that no new unmanned system can be built with a stick and rudder. Like, nope, we just won't do this anymore. We're going to have these be all autonomous. I don't think anyone would go for that rule, but I think it's a good idea.

That's our view. That's our lens through all of these problems, is how do we automate as a much of the more manual tests to enable what it's just going to be a limited number of operators deployed forward to actually be able to do the intelligence collection, define what they need to find, do the targeting they need to do. I'm really interested with that as a backdrop in how you think about what to develop and work on next. And maybe this is a good opportunity to get into some of the guiding principles that you have as a firm, especially around things like ethical considerations. I remember taking just war theory in college and having that be one of my favorite

courses, it's such a hard, really sticky, very high stakes problem. And obviously, you're right in the middle of it. So as you think about principles that dictate what you choose to build, what requirements like the one you just said, more and more autonomous. Talk to me about that. Your product roadmap is obviously very, very different with higher stakes than a difficult software company. What do you think through autonomy, there's this dystopian view of the sort of slaughterbots view of the world, which is somehow someone's going to make these robots that go out in autonomously decide to find bad guys and kill them. And it's like literally no one is proposing that. The government, nobody wants it. It's a bad idea. And it wouldn't even work if you wanted it to.

It's just not believable. What we've thought about consistently is we want to have these systems provide humans with more agency, more control, and more ability to actually make the right decisions. And this is how any policy framework will work at the end of the day with the DOD, which is We already have weapons that you shoot forward. You say there's some sort of target in this area, be it a radar that's emitting, go find it and blow it up. That is something we do today. That is something where we've said, yep, these are legitimate military targets.

We'll have the right targeting process in place to make that decision. But in a lot of ways, we will have the systems make decisions about what they will and won't do when they get there. So there is actually a lot of framework around these. But again, the policy framework is the human, the guy that pulled the tree, trigger has accountability over the actions, what happened there. Whoever gave the order has accountability over this. So everything we're doing is very much based around the idea of any policy framework we live in and any sort of ethical framework we live in will be based on humans having

accountability over the systems. Therefore, they need to have the right information to make those decisions. The other side of this is AI will be awful at being able to assess contextual things that can't be digitized, all the societal information, what aims you're trying to trying to have the political objectives of these things. What's an appropriate use of force? What's proportionality? That will be very hard to codify into an objective function.

That will be very hard to structure into something that a computer can work with. So humans will be very necessary, rightly so, in every decision-making process. So that's one view to this. Another is the privacy and reasonable use of data and technology perspective, which is when we've done these technologies, be it for surveillance or anything we're doing, we always look through the lens of can this be used responsibly. This shows up with how do you retain data, how do you control access to it, how you make all those pieces go?

But then how do you adopt things like we've been asked from time to time to do facial recognition? Our view is it probably won't do what you think it's going to do. And I don't think this is something we can responsibly employ. There's uses where that will work, but not the uses we've been asked for. So we say, I'm not really willing to work on that part. is probably not going to be effective. And I don't think if we gave it to you, it would be possible to use responsibly. So our view to this is, as technologists, we have the responsibility to inform the government on what is possible, what can be done responsibly, what cannot, and

understand the policy frameworks and the ethical frameworks that largely already exist through a very rigorous process, through a democratic process, and be able to reinforce those, not circumvent them and not short-circuit them because you can. The goal is always to live within this framework and handle these things responsibly. From an ethical framework perspective, there is a lot of policy and thinking on this already, and our responsibility is to reinforce that

and show what can be done in a responsible way. In terms of next technology areas to work on, one of the guiding through line for everything we've done is how can we bring more autonomy to these systems, and what that really enables is more scale. So how can I have cheaper and smarter and lower cost systems deployed that change your calculus from, okay, if I have to have 20 people managing this, my incentive correctly from the government is I need this to be as expensive and specialized and exquisite of a sensor or a platform as possible

so manpower constraint. So drive up the cost of my platform, that's the right move. If you can have more AI and autonomy operating these things at scale, it changes the way you think about that cost calculus. And this makes it now so that I can have larger quantities of cheaper systems that I can put at risk. I don't mind if I lose them. They don't need to be as specialized. Maybe I don't even care if people know they're there. It really does change your view on how we can employ these systems. To understand the progression of how it's actually worked, maybe you could just list off the actual things that have gone into production that you've worked on that are publicly announced or known, just to give people a sense for how it's evolved from that original platform through where

where we are today. I mean, looking at the pictures of the unmanned submarine. It's a very different looking piece of hardware than where you started. So maybe just list out what you've built so far and what each does, and then I might have some follow-ups on the lineup. So the first problem we worked on was actually base protection and border security. So when you look at, again, same story of how is this done today. You have cameras, you've radars.

And then you have typically several people sitting behind video screens, moving around a joystick, moving the camera around, saying, oh, there's something. It's easy to miss things. It's very manpower intensive and it's just sort of a very inefficient system. So first problem we worked on was could we bring automation intelligence to this very, very quickly? So we built a prototype in three months. We had a pilot on Southern Border and then I think shortly after with Marine Corps to deploy the tech to learn what worked, what didn't. I think that was within six months.

And then within, I want to say about three years, we were in a full scale program of record, which is one of the fastest periods that this has happened in federal government history for quite a while now. It was the first system we deployed scale. There's hundreds of these deployed now, and it's working very well. It's kind of very mature tech. Again, all based on that same core software baseline. The next system, we were able to roll out and get through to production, very similar paths, so the counter drone work, where we solve this problem, we said, all right, well, I think a lot of these techniques for defeating drones stay, mostly around jamming or intercepting their communications. Those aren't going to work for very long. People are going to figure out that's how this is working and adapt very quickly. So we thought a straightforward solution for knocking out these small drones was to make a very fast racing quadcopter be able to fly out and knock these out of the sky. Again, same sort of timeline. We built a proof of concept and I want to say it was about 12 weeks, went to a flyoff sponsored by DIAU, so Defense

Innovation Unit, one of the innovation groups in duty. We're one of the best performing systems there, knocked out a large percentage of the drone threats that were coming against us. From There, we're able to roll that into a pilot proof of concept deployment, learned a lot, refined the system, had massive evolutions to how we detect and track and identify these drones, like massively increased radars, visual ranges, optical ranges, all these things. And then from there, we're able to then roll into a program of record, and I think that was about two and a half here. So again, very, very fast timeline from first touch through to something at scale. The third one we've had was this ghost drone, so this helicopter drone.

That one took a slightly more circuitous path where we built some prototypes. It turns out early prototypes for things that fly, people aren't super keen on something that flies 90% of the time. It's got to fly a lot higher percentage of that. So went through and got that fully productized. And now that's a very robust, reliable airframe where we're just starting to see production deployment there. But even there with some of the earlier versions, which were still quite good, We had significant adoption with the UK Royal Marines, where they adopted this tech to basically be their squad level intelligence drone very, very early on, and learned a ton there,

deployed there, got a lot of feedback early. We've also grown through the acquisitions as well. So this is like, I think, a little bit unique for startups to be as acquisitive as we've been. But in our space, that's, A, it's kind of the norm, but B, there's a lot of companies who have built themselves up through bootstrapping and growing on their own. And what I think has been unique with a company like ours is we're able to do a lot of the things that maybe they were a part of the total of the solution, but we're able to help them solve the totality of the solution, where we acquired a company that makes air launch, tube launch drones, company Air AI, very, very cool

product. It's deployed a number of different places. So these things can shoot out of a helicopter, out of plane, off the ground, off a ship, be able to go out and conduct intelligence video imagery, things like that. Very, very cool product, but I think what their interest was is getting to the next level, having company that can do the business development, the government relations, all the software pieces, integrating field service, all these things you need to sell to the DoD. It's quite unique. Dive was another one that we recently acquired that makes these large underwater vehicles where these things go hundreds of kilometers, weeks at a time, very, very cool, about 20 feet long, six feet diameter, pretty big on man.

vehicle. They come out of defense world. They'd previously been through a defense acquisition. It's a bad time. They left to start a new company. And I think looking at it, they said, this gives us the ability to accelerate a lot faster where we can have more holistic solution. And that's worked out very, very well for us. One of the things that's been surprising in studying the company is how incredibly open you and the other leaders have been about innovation in distributing into DOD and open sourcing the playbook there. Ultra selfishly, you'd think, okay, you figured out a way to to get into a pretty impenetrable place. You've got that beachhead.

You can sell a lot of product into that for business purposes. Maybe keep that playbook to yourself. Why that decision then to be actively open about, okay, no, here's how we've done it. Because it seems like you're trying to educate others that maybe even would be competitors of Androl. I think our view is, this only gets better if there's more people playing in the space. There's more success. There's more patterns of what works. Our selfish view to this is DoD being a better buyer.

and having more bites at the apple of how to do this, and that helps. The more selfless view is, for a lot of us, this is very much about a national security problem. There's a lot easier ways to make a buck than working on this. You get punched in the face every day. It's always like climbing uphill. It's hard working in duty.

And for a lot of us, it is very much a mission problem. We do want to see this succeed. We do want other people to succeed in the space. We're not going to solve all the problems. We're fairly focused in what we do. And other people do need to succeed here. there needs to be fresh blood coming in. I think for us, that is a big part of it. We just want to see the space get better. And our belief is then if it gets better as a buyer, that's just better

for us as well. How do you handle politics? I'll just use that broad word as it relates to building and running the business. I think naively, people might look at Andrew and say it's probably a certain political affiliation or comes from a political tradition or something like that. I think that's not true, starting with you, but maybe describe the nuance there because I think it's important to understand. Myself, I'm a lifelong Democrat, gift of Democrats all the time, have always sort of identified

on more of the liberal end of the spectrum. We've got folks like Palmer, sort of not a lifelong Democrats, and has attracted a fair bit of attention for his views. The interesting thing is, especially in defense, it's a massively bipartisan issue. People really do believe

a stronger U.S. security position is net better, not just for the U.S., but for the world. You see what happens with Ukraine when we don't have our allies with the ability to really defend themselves. At the end of the day, I think really there's going to be a spectrum of ways we support allies. It's going to be economically. It's going to be diplomatically. And ultimately, they do need to have the hard power to be able to withstand an aggressor. It just gets there sometimes. I don't think on the whole defense is actually particularly controversial. I think it was for a while.

There's a reason we're not in Silicon Valley. It was a period of time around 2017, where I think it was just very controversial to do nearly anything with the federal government. Our view was doing this in places where there's a more broad view, very similar to the rest of America, where defense is one of the most popular things in America. It's one of the most wildly supported aspects of American life. Internally, we've actually been extremely apolitical. There's no political chat on Slack. There's no conversations about these sort of issues. It's about the mission we're working on. And it's pretty straightforward. That's what people talk about. And I think we've said that that's how we're going to run the company. We've been very open

about what we do here. We work on defense. We work on weapons. That is a reality. We talk about it thoughtfully. We care a lot about what we do. We try to be very ethical and reasonable in our approaches. We're not cowboys with this. We take it very seriously. I think that's worked incredibly well for us. We've just been very open and honest about what we do. And I think people really do respond to that certitude and confidence and clarity. And I think people really do respond to that leadership. What are the most difficult classes of decisions that you as the leader of the business have had to make through the business's history? Is there some consistent thing that is just always hard and probably always will be hard? Probably the hardest is really picking on what to work on.

There's a lot of things we could do. There's a lot of problems to solve. And these things are very expensive. You have to have a lot of conviction that, A, there's a buyer here. If we build this, will someone show up and actually want to scale this with us? Are we solving a real problem? This has led us into sort of a mindset that I think is pretty unique in the space where we're not working in R&D for hire. We're not just responding to proposals. We're very much looking mission first and saying, what are the real problems that need to be solved and how can technology actually move these forward? And if we brought this forward to the customer, would they actually do this? And I think a lot of this has been informed by a pretty broad base of our team where we have folks that are something like 20% vets.

We have engineers from big tech, from defense primes, from everywhere. We have folks from the hill, from government all over the place. So it's a really pretty diverse view of folks who can look at this problem in a lot of different ways and say, this is what needs to be to solve, this, what's not. But these are big consequential decisions. We'll dedicate a team to these things for years. It'll cost us a ton of money. And we just have to have an outrageously high batting average with success on these things. And the nature of our business is very much going to be there isn't one product that will scale to billions of dollars.

I think we'll get there with some of these bigger hardware products where they will be very large successes in and of themselves. But we know we're going to have to have this sort of approach of multiple products that sort of each sell into some part of the market. And that's a very hard decision to make. We're sort of constantly evaluating it, reevaluing what we've invested in, how can we get better signal? and then how quickly can we get to something to prove that there's actually interest in traction in the market? If you think about the recent news where you were awarded a billion dollar contract, I think by Socom, you know, come a long way. That's obviously a huge number. If you were to rewind time before you had had this established relationship with DOD, what was the absolute low point in you trying to

penetrate them as a buyer for the first time? What is the story of how you overcame whatever that challenging period was. Maybe I can hit one or two vignettes that I think highlight sort of challenges in working with DoD. One was we had a great early adopter with Marine Corps on using our towers for base protection. And I think this is the pattern a lot of people fall into where these were the operational users. These are guys who were responsible for protecting the bases, managing the bases. So we did a pilot with them a year or two where we deployed, I think, 30 different towers across four or five different bases. Users loved it.

The success was great. Efficacy was high. There was a lot of demand for it. And then the thing that I think everyone finds very surprising about DoD is the people who have the operational mission are in no way in charge of the budget or buying decisions whatsoever. So they said, hey, we want this. This is great tech.

We want to roll it out. And then I think working through the process of what is the requirement. Okay, if this requirement validated, signed off on and that takes time. then where does this fit into the budget cycle? The budget is decided two to three years in advance of the year where you will actually be able to do this. So when we were starting as a company, they were deciding the budget for when we actually wanted to get funding

two or three years later for this technology that was then proven to work. So it wasn't even invented at the time they had written the budget for that year. And again, it's like everyone's well intention. They're trying to manage this in the same way, but the system has gotten so slow and structured to be so slow, that when you start having these more innovative technologies, these things that move very quickly, the ability to react to respond,

even on pretty low dollar value things compared to the $700 billion budget of the DOD, it is very, very hard to move quickly. I think that's probably one of the best versions of DOD in a nutshell, which is like great tech. It works. People want it, wait three years, probably five of the time requirement,

then budget, then everything shows up. That's the timelines you're really looking at, which is quite hard. Now, there's people trying to move this faster, a lot of innovative approaches to making this go, but it's slow. And that slowness can be very disheartening. And I think the part that's probably hardest for folks in DoD to understand is they have not worked at a startup. They have not had to show top line numbers. I don't want to say they're unsympathetic, but it's certainly not something that resonates clearly around.

If you want innovative companies to show up and stick around, these are the things they have to put up. These are the numbers they have to show. We've been able to work through that and still be able to put up great numbers, but it's certainly not without its challenges, and we are in a lot of ways working against the system as design. Do you think that ultimately you will change that, change the way that DOD does its job so that the deployment of useful operational tools can happen on the tighter feedback cycle? I don't want to take too much credit for it. It's going to be a lot of different people working to solve this, but our hope is we can demonstrate successful models. We can put forward new ideas. Nearly every idea we've put forward has had a lot of traction and uptake.

Our GC put out a proposal for a different model of software IP rights and has gotten a ton of interest in running a pilot on this. Even very weedsy things like that, we thought 20 people would read the article. Turns out maybe it was 30. They were all pretty interested. There's a lot of interest in pushing these things forward. And I do think there is a lot of more macro reforms and things like that people are considering.

There's a commission to look at the PPB. I think it's the planning, programming, budgeting, and execution process. They established a commission to look at how can we go faster, how can we solve these things differently. So there's a lot more momentum. I think a lot of that is due to China, honestly, and the pace that they're able to develop new technology. And I think it's not as obvious to most people. I think the top folks in DoD really get it, but the pace is everything. The Silicon Valley has figured out anything.

It's compounding is very good. If you compound twice as fast, even if you're a little bit worse, you're still better off. And the pace of being able to evolve, change, get new tech out, that is everything. So if China can get new capabilities out in three to five years and it takes us seven to ten, it doesn't matter if we're starting at a better position. They're going to blow past us very, very fast. I think there's a recognition that that speed of compounding, that speed of iteration, past the pace to fielding is so critical.

It actually may be the most critical thing. there is a lot of energy and momentum to change around this. Even though the world's biggest company is a hardware company, the well-worn phrase is hardware is hard. And to invest in hardware businesses has been, sometimes it's burned everyone that's tried. And it's just really difficult. It's more difficult than moving bits around. So since you're in the business of both bits and atoms, what could you teach us about

hardware innovation compounding and doing that effectively? What are the big lessons that if you were to go to just some random hardware startup, you would bring with you on working on hardware, given your experience at Andrel? Coming from a software background, you're a SaaS company, what do you hire? Software engineers, maybe you get some front-end specialists. You think that's really different. And then you get designers and some PMs. Probably four disciplines.

You hire some sales guys. Pretty straightforward. You got a hardware, it is like dozens of specialists. The variety of folks you need in terms of electrical design, electrical test, mechanical design, testing pieces there are manufacturing, specialists in manufacturing, supply chain. logistics, it just gets very complex, very fast. It's certainly not for the faint-hearted. It's much harder operationally. The thing I think in the valley that has worked so well was, and why velocity has gotten so high is the cost to change, the cost to rework is basically zero. You just roll it out.

And a lot of the tech investment has been like, how do I move faster and make stuff cheap to roll out, make changes cheap? And that's worked very, very well. And again, get that velocity up. It's very, very effective. The hardware world, you can do that depending on the problem set. Maybe the biggest lesson we've had is how do we as cheaply as possible get to test out product market fit? That's the main lesson we've tried to drive, which is how can I cheaply get to a prototype that is not fieldable, that is not robust, but it is something I can see, is there a buyer?

Would they actually move on this? Are they trying to pull me along, or am I pushing uphill and trying to fight to get this adopted. And there's dozens of companies making these mid-sized drones at this point. There's a lot of questions of like range and payload and all these things. And it's hard to sort of test those until you get something in front of somebody flying and say like, what problem would this solve for you? Because they just don't believe you until the things flying or working or doing something. So a lot of the approach we try to take is very cheaply getting to that first prototype, that first article, and bringing customers along the journey with us to say, is this interesting to you? Is this

solving a real problem and getting that buy-in as early as possible. The extreme form of it is what the DOD has done historically, which is they come with the idea, you just do cost plus contracts and they absorb all the risk. But the intermediate form that we started doing with the Australians, for example, is this co-investment strategy where historically we just built a big sub, get to a prototype, and then get it to production quickly. There where we were able to work with them to validate, yeah, this is actually really interesting, and if we can hit the price points and features that they're really looking for,

they would 100% want to scale this. So it allows us to shave years off the development timeline and allow us to go much harder where we can skip to production much, much faster. So this is something we've been seeing with the U.S. as well, is trying to find these blended strategies where we still have skin in the game, we still have an incentive to move fast, and we're not trying to make an R&D business that was more successful, the longer and worse these things go. But to make a business that we've a real ability to shape roadmap, deliver something that works, but also get a customer really bought in on the journey and see us as a partner to actually solve a real problem for them instead of having to guess and then go hunt.

We've kind of been playing with both strategies. We'll continue just to do things that we have conviction in, totally independent of customers, but then to the extent we can really get that buy-in early, I think it really helps. But the reality is just getting something in a prototype, you can do that pretty quick. So if someone's out there starting a hardware company, unless it's consumer hardware where nobody cares about a prototype. You can really try to learn the space fast and learn what you don't know. Then really getting those customers bought in, they would run through brick walls for this.

Then you know you have something. Is there a moment in Anderil's history as a business so far that you are most proud of looking back on? I'm very happy that the government is kind of moving at the space. So being able to see these bigger programs move in such a timeline and get the tech adoption, despite all the past conceptions of these things, so be it with customs and border protection. so calm, that's always really impactful. Probably the things that are most exciting

are really just mission wins. That's the stuff I get really excited about when it's deployed and it solves a real problem. Those are the things that I really feel the most motivated by. They're consequential problems. When you move the needle on it, sounds corny, but it actually like, well, that mattered.

That mattered a lot. Those are the things that I personally get most excited by seeing the pace at which we've been able to get people excited, get government really moving in a way that it was different than I thought. I thought it was going to be a lot harder. I thought it was going to take a lot longer. It's still hard. I thought it was going to take 10 years.

What about the other side of that coin? Looking back, any major moments of regret or things that you would have done differently? Honestly, not many major decisions that I can think of that we felt like we made major mistakes on. I think a lot of the learning along the way has been, for me, as a software person, has been on the hardware side of how do we actually build these things for production? and what does that really mean and how do you get them reliable? I do think that's very hard, and we're learning a lot of hard lessons on that as we go along.

Like, how long does this take to get a drone to a point of reliability? You're really happy with it. What sort of testing do you have to do? What sort of rigorous kind of process do you have to do? How do we adapt that to being able to move really fast and balance those things? I think those have been hard learning. I don't know that anyone has the answers. We know a traditional aircraft program looks like.

It's like five to seven years of systems engineering and then out pops the other side. Hopefully you got an aircraft that works. But taking this more iterative approach and really learning what's the right people, processes, and right way to blend both worlds, I think has been a really interesting learning experience. If you think about the very small but hopefully growing list of companies like Anderil that face the government that are tackling these hard problems, maybe the two that preceded it that are important, one of which you worked at Palantir, are Palantir and SpaceX.

What did you learn from those two companies either working there or observing them that you think are useful lessons for people thinking about. this class of problems. The Pallenter case was really a lesson on how do you sell and work with the government, both from like an operational view, how to security work, how to contracting work, all those things, which are extremely not obvious. And if you don't know, it will take years to figure it out. Then really from a business development perspective of what works, what doesn't, where are they going to help you, where are they not? And what is something they're willing to buy? The really hard lesson is they buy systems and capabilities. They don't buy parts. And that's

quite a bit different than, you know, if you're a cybersecurity company, it's very hard to just go sell to the government because they want to buy a secured network. They don't want to buy another tool that they have to then deploy, manage, install, and operate. The second piece we really learned there was going through the being a subcontractor, doing a small part doesn't work. You don't have control over your destiny. These larger companies are well incentivized to minimize your revenue and contribution and maximize their ownership in unique positioning. The very rational incentives. They want to be a unique value at. They want you to be as not unique and as commodity as possible. So you just can't go through these traditional players easily unless you have

something truly unique that they cannot get any other way. But even then, it's like how fast can you scale? You're on their timelines. They're disincentivized from making you the centerpiece of why this is the right technology. I think the SpaceX lessons a little bit different in that that is a class of technology that nobody thought someone outside of the DoD world could do. There was this beliefs like, oh, they can't build these rockets. There's no way. Only Boeing and Lockheed consortium can do it. Obviously not true. It's like, well, they can't do it cheaply. Obviously not true. They went from not having any satellites to being the largest constellation operator in two years. I'd be like, oh, it won't work. It's like, it obviously will work. The degree to

which they've been able to completely shame everyone at the pace and scale and innovative approaches that they've been able to do on what is considered some of the most difficult and hard to access technologies is incredible. It's just wild. And I think it shows that there's really no space that these legacy players uniquely will be able to control. I think it is something where everyone should look at it and say, well, this is one of the hardest areas of tech, and they did it.

And they did it on their own dollar. it faster than anyone thought was possible. The story we often tell is both of these guys had to sue the government to succeed. In Palancher's case, it was they had the right tech and the government refused to consider commercial solutions and instead kept rolling their own tech, which was expensive, not delivering. And there was just a very willful disregard for looking at alternative approaches. I think in the SpaceX case, pretty similar, a willful disregard for looking at alternatives. that still exists. If something is locked in and on an approach, it's very hard to get them

changed course because the incentives in the system and the government are, if you admit that your approach was wrong, you will be punished. Not that you picked a better approach, is that you were wrong before. So it's very hard to drive this change because Congress will yell at you, your superiors will allow you, you will not get promoted because you were picking a prior bad strategy. So it's not set up to adopt new approaches on the whole. It requires an admission of failure, not an acknowledgement of success. So it is quite hard. So largely where we've tried to focus is then don't require an admission of failure. If you go after those areas, you will be fighting forever and probably will still lose. It'll be so expensive and so painful. So we've tried to

target areas where it's novel. There's not an existing approach. And this is something truly new, and technology has changed that enables something that you didn't think was possible for. and that's something people can really get behind and I think it's enabled us to succeed very, very well. I think it's just an incredible lesson to close on. Such an interesting business that you're building and have built already. I've learned a lot from that it's completely unique in the lineup of businesses that I've explored for sure.

Just like you said, an end of one type company. I ask everyone the same traditional closing question. What is the kindest thing that anyone's ever done for you? Kind and nice are very different words. one of the kindest things in the course of our business was when we were working on countered drone systems, we had kind of an early champion. And he beat us multiple times to perform better and give us a chance and say, you need to step it up. And that was actually one of the best things that has happened to us in terms of really putting real pressure on to succeed. But in a way that he wanted us to succeed. And that to me was that belief and that willingness to trust us and work with us.

That was one of the best things that's happened to us and really was one of the kindest things that took a lot of courage on his part. Not the nicest, but definitely one of the kindest things. I love that distinction. There is a category of these answers that is tough love that wouldn't be there if the question was nicest thing. I don't think anyone's made that exact distinction, but I love it as an example. And Brian, I'm so appreciative of your time today. Thanks so much for having this conversation. Thank you. To find more episodes of breakdowns ranging from Costco to Visa to Moderna or to sign up for our weekly summary. Check out join colossus.com. That's J-O-I-N-C-O-O-S-U-S-U-S-U-S.com.