--:-- Host: The DoD is a black box. And I say that meaningfully because we're trying to fix it. We want to overwhelm the system with how much better it can be.
--:-- Alex Miller: The US Army should never trade blood for blood in first contract. And it should always be blood for iron, and it should always be their blood.
--:-- Host: In 2025, when people think of government, speed or innovation aren't typically words that come to mind. But this was not always the case. In the early 40s, we built the Pentagon in 16 months. Shortly after, the Manhattan Project took three years. When the Soviets launched Sputnik in '57, it took a mere 84 days for America to respond with Explorer 1, kicking off the space race. In the 60s, Kennedy asked for a man on the moon, and the United States Apollo program answered within the decade.
--:-- And here's the thing: the very roots of our technological brawn as a country are deeply rooted in the federal government. But in recent decades, a gulf has been developing between Silicon Valley and Washington, D.C. A renewed interest is developing, and founders are building toward the national interest. But many have no clue how to navigate the black box that is government procurement.
--:-- This is not a technology problem. This is a culture and a process problem that we can apply technology to. So in today's episode, recorded live at our third annual American Dynamism Summit in the heart of Washington, D.C., we bring in two chief technology officers from two of the most important institutions in the nation. Alex Miller is the CTO for the Chief of Staff of the Army, and Justin Fanelli is the CTO for the Department of the Navy.
--:-- Listeners of this podcast may know what a CTO does in the private sector, but what does this mean in the public sector, especially when the CTO position within government came 30 years after it did in the private sector? And even later than that after defense.
--:-- Joining Alex and Justin is our very own a16z go-to-market partner focused on American dynamism, Leila Hay. Together, the three discuss important topics like how we accelerate, shifting from planning years ahead to software speed.
--:-- Alex Miller: The entire way we think is predict the future three to five years out, land the perfect shot, get a trick shot and go. And it just doesn't work. It does not work.
--:-- Host: Also, the reality of resources on the ground.
--:-- Justin Fanelli: It is shocking to me how much stuff we buy for soldiers that no one would accept if I handed it to you in your everyday life.
--:-- Host: And importantly, whether our leaders have the leeway to fix what's broken. We discuss all this and more, including the role of culture in bridging the divide between Silicon Valley and DC, plus the changes ushered in by the new administration. So do we need to rip up the system and start over? Or are we already on the path to a solution? Listen in to find out.
--:-- [Disclaimer: As a reminder, the content here is for informational purposes only, should not be taken as legal, business, tax, or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any a16z fund. Please note that a16z and its affiliates may also maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see a16z.com/disclosures.]
--:-- Leila Hay: You both are CTOs of two very important public institutions. Can we start off with telling us who you are and what you do?
--:-- Alex Miller: I'm the CTO for the Chief of Staff of the Army, and that is a big title because the Army is a big place. In my day job, it's a lot of education. I would say probably 60-40: talking to people about what technology can do for a mission, and then 40% actually hands-on, putting my engineering cap back on and actually doing things that move the ball forward.
--:-- I would say very similarly to a company, what I do as the CTO for the Chief is think about what the North Star should be for the Army in terms of how we leverage technology, how we employ it, and then what our missions are that can leverage it. And then actually going forward and helping people get there.
--:-- While companies have boards of directors and presidents on all these different things, we also have a board of directors that is Congress. We also have stakeholders that are the American people. However, most companies don't have 60 years of policy and rules that were written generally by enthusiasts and visionaries, but implemented by minimalists. And what I mean by that is risk minimalists. So most of our day job is just going through and educating, having a unique understanding of what our job is, what the mission is, and having done it to make that connection happen.
--:-- Leila Hay: And what did you do before you were the CTO?
--:-- Alex Miller: Before I was the CTO, I was the science and technology advisor for the Army G2, and that is the Deputy Chief of Staff that runs intelligence. So I was the guy doing all the technology for Army military intelligence. I got to do some pretty cool stuff, work with some really interesting people, and that was where all my downrange time was as an intelligence professional.
--:-- Justin Fanelli: Justin Fanelli, Department of Navy, so Navy and Marine Corps, Chief Technology Officer. Just like Alex said, Navy's a big place. Oceans are two-thirds of the world. We want to cover as much ground as possible.
--:-- And the CTO position within government came 30 years after it did in the private sector, right? And even later than that after defense. This is really a late acknowledgement or a nod that commercial technology can be a disruptive force positively for the government and that we can go much faster. But we need digital to do that.
--:-- So to Alex's point, finding innovations from nontraditional partners and at the edge and scaling those much quicker, and tying it to some sort of divestment because that money is locked in that three-year period where we're spending on existing tech. We're out there scouting, very similar to how private sector would do. We're working with hundreds of internal organizations, in some cases with different kind of sub-optimized or very specific goals. How do we make that a reconciled space where we can just bring the biggest outcomes to the hardest problems? That alignment piece to bring ultimately outcomes overmatch is our goal.
--:-- Leila Hay: So you both mentioned industry. You both mentioned startups. My colleague Ryan and I actually just wrote a blog post about selling into the DoD because there are these paths that startups can pursue. We all know about SBIR, STTR. There are great organizations like DIU and AFWERX that are out there looking to help pave these paths. But for a lot of startups, working with the Army or the Navy still feels like a black box. Can you walk us through what kind of pathways are evolving and new opportunities for industry to engage with your organizations?
--:-- Alex Miller: Let's talk lexicon first. I know it's a little bit bureaucratic, but I think it's important. So startups generally focus on the R&D side. They're taking a concept up through some type of development into that golden minimally viable product phase. And the "P" is really important because it's not minimally viable prototype, which I think a lot of DoD thinks about in terms of, "Hey, that sort of looks like it's useful." But what startups are thinking about is, "Hey, what's that first product that I can productize and get to market?"
--:-- For that phase, the DIUs, the works—whether it's AFWERX or Softworks or Eagleworks, all of these different organizations—that's exactly where we want them to be. The reason I say that is because the DoD is a black box. And I say that meaningfully because we're trying to fix it.
--:-- That black box was set up on purpose because in the 60s and 70s, coming out of the Cold War and all the way up through the 90s, we had all these different things. The Packard Commission for how you think about buying things. The Clinger-Cohen Act for how you do information technology. Goldwater-Nichols on how you connect mission with buying things. And all of that is still the rules that we live under, and everything since then has been duct tape and bubble gum.
--:-- So what we are trying to do is instead of MacGyvering our way through that—which all of us have spent a lot of time doing—we are trying to sort of domicile, cut through all of it, get rid of all of this extra bureaucracy, work with the Hill so that it's not a black box. It's like the 12-step program: if you can't admit there's a problem, you can't actually solve the problem.
--:-- That's the left side of the equation. The right side is what the Army is doing: just bringing industry in and saying, "Hey, we know we have a problem, we want to help." Whether it's helping us write requirements better so that we know what's available, so we're not trying to divine that from nothing, or, "Hey, how do we bring consortiums together?"
--:-- I love the new software strategy from the SecDef because it tells us: use other transaction authorities. Use the things that Congress has given you and do it aggressively. So putting consortiums of teams together with industry to say, "I'm not going to pretend like I know everything. We've done that for 50 years, no more. You tell us what's available and we'll tell you the mission and how we can apply to that."
--:-- Justin Fanelli: We want more game changers in this space. When we have barriers, it stops outcomes. So in general, we have more companies than we've ever seen before who want to work on national security. There is a general recognition that this world can be a dangerous place. And the more secure we are, the better we are at this, the more overmatch there is, the more that we can have peace through strength.
--:-- This is, in my opinion, probably the best alignment that we'll have in our lifetimes. This is potentially a century anomaly for actually being able to divest. It's very hard to turn things off and then bring new players in and bring new waves of ultimately positive disruption through and through.
--:-- That translation piece has been very expensive. Product market fit is slow. And at times we turn away companies that would have brought breakthroughs. But only superior military technology can credibly deter more. And the fact that we have more people interested in supporting that mission—what does the go-to-market strategy look like?
--:-- Alex talked about, "Hey, we have some of these front doors." We don't want 100 front doors. That's confusing. So how does streamlining that look? If you're going to completely break through and change the game and you're selling something that we don't buy right now, you're swimming upstream. Anduril has made it through and fought that fight. SpaceX has, Palantir has. But number one, we want that to be more straightforward.
--:-- In that particular case, the PEOs are not your first customer. Sorry, program executive offices.
--:-- Leila Hay: Maybe we can just quickly talk about what a DoD program office is, because I think that's another black box challenge for startups. I think a lot of companies know that you generally want to be on that journey. You get some R&D dollars, you work with DIU, you get an OTA, but ultimately startups want to get to a program of record. Can you talk about exactly what that is and what that journey is like?
--:-- Justin Fanelli: One hundred percent. So people have heard of "program of record" if they've worked with the DoD or the Defense Department at all. We have 75 program executive offices. They buy at scale. These PEOs are made up of program management offices, and then you have hundreds of those. Navigating that to figure out where to sell is a scale game, and it's been too hard.
--:-- So one thing that we've started to do within Department of Navy—we have 18 of those 75—is we've said, "Wouldn't it be easier if we made data-driven decisions and started treating these buys not as sacred program of record, but let's buy capability per dollar?" Our move towards "military moneyball" has been to open up the aperture through converting programs of record and program offices to portfolios. We've shifted PEO Digital, and I believe that we'll shift even more of those program executive offices across DoD to that.
--:-- Now that's for a Horizon One capability. So if something improves the way that we're doing business, if we already have it budgeted, then you can go directly to them. And there's actually an index online that—I think Steve Blank put it out—that says, "Here's what each PEO does. Here's where they are. Here's who you can talk to within them." The goal there, because they have fixed budgets, is finding something again to turn off, because no one has runway to do three years of piloting until you transition something.
--:-- If you are doing a Horizon Two capability—we're working on a breakthrough, but it's existing technology or there is a budget line—we have now more Valley of Death funds than we used to. We have Small Business Innovative Research, but that's more of a door prize. You need to get to that operations and maintenance long tail funding. We have something called APFIT—Accelerating Procurement for Innovative Technology. It allows some breakthroughs to get pulled all the way through. There's just not enough of those, right? There's just not enough funding and not enough awards.
--:-- So the point is: if you can shrink the Valley of Death, you're doing this in the Horizon Two and Horizon One space. For Horizon Three, you probably have to do some guerrilla marketing and advertising—pull in DIU, pull in the Hill, pull in the requirements writers to make sure that you're not doing all the translation yourself.
--:-- Alex Miller: So what if we didn't do all that? The cool thing is everything that Justin just said is a workaround. It is a workaround to a system that is fundamentally broken. The cool thing is PEOs are not actually in statute anywhere. They don't exist in law. We made them up because the rule is that the SecDef gets to determine what that looks like. So as we think through what the future is, I don't want to have to know that acronym soup. Why would we put that on anybody willingly? The only reason it exists is because the process doesn't work.
--:-- Just a short story on how this works for us. Justin talked about how we build these requirements, and forever they've been super gold-plated. They're this thick. They're every word known to man because people are afraid that if you don't get it in the requirement, you can't move to this program. Thousands of pages. What if we just stopped doing that? Because it is not actually what we have to do.
--:-- I'll give you an example of that for countering unmanned systems. And there's lots of startups in this. I know that a lot of your founders are not in the threat headspace that I'm in right now. So think about it: you're at an NFL stadium or a soccer stadium and you see a quadcopter. In my world, that's a horrifying prospect because that means that it's intruded into your airspace. You can't do anything about it. It's really hard to bring them down. Now think about thousands of those. And that is the real threat that we're under.
--:-- Well, all of our requirements writers—they're not operations. They're not the guys getting shot at. They're just trying to do the right thing. And you write these sort of black box requirements. I found one the other day as I was going through them that was coming up through the process because it wanted to go to a program where if you took the combination of statistics that they put in there—your operational availability, your operational accuracy, and your operational reliability—what you ended up with was a system that only had to work 51% of the time. What person at what company is going to go, "Yeah, I'm okay with my engineers building something that only works 51% of the time"? Or if you flip it, "I'm okay with wasting 49% of my resources."
--:-- So as we think about what the process should be—and it's real, this is a "with our shield or on it" type of moment for the Department of Defense—I just want to blow up all of that. Anything that's below law, we should be able to reshape. No startup should ever have to go, "I want to be a program of record," because what that really turns into is a 30-year long set where Microsoft might have been thinking Office is going to be around forever. But inside of them, they still had product managers and program managers who were making sure that Excel was really good. Please don't stop making Excel good. The entire world runs on Excel.
--:-- But they're not thinking, "Oh, if I just get it here, I can slow down. I can take my foot off the pedal," which is what happens in programs of record. As soon as you make it, as soon as you've got that long-term budget, there's a perverse incentive just to maintain and create inertia. So my theory of the case is I'm going to try to blow up all of that.
--:-- Justin Fanelli: The question is: how do we increase yield? We have examples where every good idea the government has piloted—DoD has piloted at some point. In this particular case on requirements, I worked with a 4,000-page requirements document. I'm fairly certain no one read—
--:-- Alex Miller: Do you read it?
--:-- Justin Fanelli: No one read all of it cover to cover. You look at it, reference it like the dictionary, but it's a little bit longer. In this particular case, we said, "Hey, we're going to use capability need statements because this isn't going to get us what we want anyway." And so we fast-tracked an 18-month process in three months with a couple pages between CNSs—capability needs statements—and top-level requirements. The software acquisition pathway allows us to do that.
--:-- So we already have examples where the right thing is working way better than the old thing. It's a K-curve. This is a matter of scaling what's working better. And the only way we do that—because there is some risk aversion—is emphasize that the outcomes and the trade-offs are for our warfighter. When we have the best software developers in the world, when we have the best capability coming in, let's not cut it down with 1,000 paper cuts, 10,000 paper cuts. Let's make decisions based on the impact that they're making. We're doing more of that, but it's pockets of excellence.
--:-- Leila Hay: That's huge. So there's this acknowledgement about the black box and then a solution to managing the black box. It sounds like there's also major changes happening when it comes to capabilities and being able to say we need to experiment because the pace of technology is so fast. We can't write a requirement for a capability that's going to be developed three years from now because of the pace of software development, AI, all of these things. It's moving so fast that you need to be more agile.
--:-- Justin Fanelli: Especially with horizontal capabilities like artificial intelligence. How would we pull in artificial intelligence across 75 PEOs? We need to name enterprise services. Defense Innovation Unit actually has a digital on-ramp where they're trying to streamline that so that we can make buys based on impact. The chances that a bunch of horizontal organizations were going to do the same thing was always high.
--:-- A quick example: we had edge compute ashore. We had a hyper-converged infrastructure, small box that we could divest a billion-dollar purchase and just move to this in every way more capable solution. And we said, "What about that for afloat?" One of the captains of a ship said, "Let's put it on mine. Let's figure out how this works." We put it there. He deployed to the Red Sea. It's on other ships right now. This is enterprise moving to edge, specifically afloat, in a case that we didn't need to go down two different procurement paths.
--:-- There was another example recently where we had an electronic warfare solution, and we heard that the Army was going to go out and do a request for proposal on something that was really similar with the same long requirements document that we had. We said, "Can you just paint this brown and throw it on a Humvee?" And that's exactly what they did. They did it in three months for $300K, and then that went to scaled fielding. So these things are happening. It's just a matter of: can you make the best thing the normal thing?
--:-- Alex Miller: And Justin said it's treating the value as the actual goal. Everyone has heard this: PMs love this concept of cost, schedule, and performance. They want to make sure they hit their timelines at the cost that they were given. And performance to me is like, "Hey, it meets my warfighting needs." But performance is actually what's written in this requirements document, which they might not have actually had any hand in doing, or a soldier might not have had any actual hand in informing.
--:-- What we're trying to do is flip that and say, "No, your value and time to delivery is your actual need." Now, everyone has permission to do the right thing. So as you think about venture funding all the way up through Series X, Y, or Z—where a lot of the companies that we work with actually still live in those spaces—they need to have a front door. But what we're doing in the Army is going, "Hey, if you're ready, let's go to the field right now."
--:-- I say this a little bit parochially: the Army is the service that can do that because you can move around the edges of some of your ships. The Air Force can move around the edges of some of the aircraft. The capital assets, those are fixed. What I have the luxury of in the Army is we can modify our organizations, we can modify our kit, we can tailor it to missions. That's what we've been doing with some of our brigades that have been in Europe and in the Pacific and frankly going in and out of the Middle East still.
--:-- So this is an open invite for all of the founders and all of the companies. If you have things that work, we have a front door. We want to try to get you in here. We want you to be able to engage with Army Futures Command. We want you to be able to engage with our units directly, because only the soldiers will actually be able to tell you, "Yes, this works at a time of crisis."
--:-- I'm super excited because I remember during the Global War on Terror, we had quick reaction capabilities. And as a customer, I could go, "I need this right now." And somebody would go, "Cool, I will get you that." In 30 days, I had kit in Afghanistan that was working for me, which told me it was legal, it was moral, it was ethical, we could do it. But we never actually went back and fixed the system. We just moved off to a side chain and everything was a side quest.
--:-- Now we're saying, "Hey, we got to fix the main storyline." And everybody's like, "Oh, we don't know how to do that." This is a period about scaling what actually works. So I'm excited about what Alex is doing. That transforming in contact in execution is a big deal because it's tech informed.
--:-- Justin Fanelli: If you read "Origins of Victory," they say it's not about who necessarily just has the best tech, it's how they're adapting to that tech and using that as advantage. The way that we're measuring outcomes is we're now using outcome-driven metrics to figure out: do you move the needle? Not, "Is this the best tech?" or "How does this fit into our current operating model?" but "How does this transform us?"
--:-- We are more open and listening to and working with and pulling through more of those Horizon Three capabilities than we were. But we need outcomes overmatch. We need it to kick the ass of whatever's there right now so we can show that difference. It's defensible and we can move off of the old and onto the new.
--:-- Leila Hay: You've mentioned "overmatch" a couple of times. Can you double click into that and share what exactly that means?
--:-- Justin Fanelli: It's very hard to make a change for a 15% improvement unless it's something that you're doing a lot. We want to make the case that there are orders of magnitude leaps, like power law leaps. Most recently, we have a thousand different identity management solutions.
--:-- Leila Hay: Just a thousand?
--:-- Justin Fanelli: So we've said Naval Identity Service is the enterprise system. That service is what's moving forward. That opens the door to repurposing some of that money, repurposing some of that attention to harder problems. To this point, we are looking for A-B testing that shows that A is no longer a viable option and there's much more value from B.
--:-- If a company comes in—like Serana comes in—and they said, "We have tests. We use Rescale. We use the cloud. And we show that based on running this for 100,000 reps that would have taken 10 years, we can fast forward this." And that is proven. It was proven on paper. We didn't have to put that into an exercise. We can now with a lot less risk reduction—but that is an easier A-B decision because they've translated it into our language, which is outcomes.
--:-- Alex Miller: I'll give an Army example. We have this concept of mission command. I don't want to go into Army terms because it gives us this false sense of security, like we're doing something unique. We had 17 programs of record in this mission command enterprise. And each one did similar things in the fact that each one of them had a map. Humans like pixels. We're really good at seeing pixels. They make us happy. So we looked at maps.
--:-- Well, each one of those came with its own mapping server and its own way to do tiles and its own way to deliver the maps and its own way to store the maps. And about 2015, 2016, everybody went, "Hey, this Google Earth thing is pretty legit. We should probably just try to use that." And it took five years for people to go, "Oh, we can just start sharing KMLs and KMZs with each other and start sharing it."
--:-- By then, you had really, really good software companies that had said, "Cool, we're going to give you these enterprise services. I don't care what map you use. I don't care who provides it. You just tell it, you point us to it and we'll serve it to you." And all of us kind of went, "Yeah, I want that." But we couldn't because we had 17 programs of record owned with their own vestigial architecture. And we had to do a bunch of architecture archaeology. And we had a bunch of people who were protecting their rice bowls in terms of those programs.
--:-- It's only just now in 2025 that we are actually trying to hack and slash and kill some of these systems to go, "No. If you want to look at the map, just look at the map. If you've got some weird, unique data—like we do fires, like we do artillery, like all those unique things that are unique to us as a military—we'll figure out a way to get those on the map."
--:-- But we, the Department of Defense, are certainly not in 2025 who is building the best maps. We're not building the best data platforms. We're not building the best API gateways. So let's not pretend it. Let's just get all those pieces together and collapse all of these other vestigial organs. I cannot describe how excited I am because we're doing it. We're doing it live and I love it.
--:-- Justin Fanelli: It's a leverage play. The reuse piece—we're even using some of what the Army's doing and vice versa.
--:-- Leila Hay: It's interesting because I'd never thought about it from this angle, but it's not unlike venture capital where we're making bets on companies. It's not about, "Can this company move the needle by 10x?" Because to your point, there's this switching cost. So we're investing in 100x solutions.
--:-- Justin Fanelli: Yeah.
--:-- Leila Hay: And then you all need to be able to validate that this new solution is going to provide 100x outcome. And you also need to be able to replace and sunset capabilities as well. So it sounds like our organizations are not that dissimilar.
--:-- Alex Miller: They're not. And what I appreciate is I've learned what venture looks at and how you grade and how you assess risk. You do not expect every company to be good at everything. And that is a failure that we as a Department of Defense—and really anybody who does federal acquisition—falls into the trap of. If you came up and you won a program, I expect you to deliver every part of that top to bottom, even the stuff you're not good at.
--:-- Well, what that does is it creates an amalgamation of things that people aren't good at. And you can actually multiply those tolerances across lots of systems until you get something that looks like, "Oh, there's a lot of not good stuff here." What we're trying to do now is just figure out—I'm a Lego person. How do you find the right Lego pieces and how do you put them together so that if Justin Corp is really, really good at serving data, I get him serving data. Leila Corp is really, really good at UI, you're doing UI. I don't make you do the things you're not good at because frankly, your engineers are focused and excited on doing the things you are good at. And I can put all these things together.
--:-- That's what we expect in our everyday lives. It is shocking to me how much stuff we buy for soldiers that no one would accept if I handed it to you in your everyday life. And that is a mindset that we are changing. I love that the Secretary is on board, the Chief is on board, all of these senior leaders within the Army are just really going, "Hey, common sense. What's the value and what's the opportunity cost of making really dumb decisions? And how do we stop doing that?"
--:-- Justin Fanelli: One of the unlocks there is they're acknowledging—the Secretary of Defense is acknowledging—that software-defined warfare is here. And if you can use one component that is higher performing than others, should we have a fully different artificial intelligence stack for edge compute for sensor integration than we do with no leverage as an enterprise use case? No, of course not. What does the governance look like versus what does the compute look like? We need economies of scale.
--:-- We're working together more with our vendor partners and even signaling to some of the investors to say, "Here is our hard problem. But can we land the plane on buying them? Can we land the plane in sustainable cost?"
--:-- Justin Fanelli: Whether it's in a program executive office or not, everyone should have to sing for their dinner every year. The idea of bringing those breakthroughs in and then Zambonying out the legacy capabilities so we can unplug them for the first time in a long time — these work together within the valley of death. This is actually happening within our budget cycle if we do it well.
--:-- Host: It sounds like you guys are really trying to break down a lot of these silos with the program offices. There are 75 of them. There are all these different programs. And you all are thinking about how we can find some of these horizontal capabilities that are relevant, not just within the Navy or the Army, but across the services to be able to find economies of scale for your organizations and that also make it easier for startups to partner with you all and scale with you. Am I hearing that right?
--:-- Alex Miller: Yes. And one of the things that I've been really big on is we need the enterprise to do enterprise things and get us out of the business of doing things that are common to all. In the intelligence community, we had this concept of services of common concern. What that means is it is something that everyone should be able to use. Therefore, don't let every individual agency or organization do it themselves. It's sort of a common sense on the face of it thing. However, on the combat side of the department, we've had a really hard time adopting.
--:-- I'll give a couple examples. The CDAO and DIU and some of these other organizations have said, "Hey, we're going to pull capabilities up and deliver it in a common way for everyone." That's great for things that are common. So that when Justin and I go back to our offices, if we need to share data, we can do it in a common way. For things that are unique to me, or unique to him, or unique to the Air Force, or unique to the Space Force, or the Coast Guard, the Marines, whoever — if it's really, really unique, like can only be done by them and should only be done by them — that's where we want to focus, because that's the niche.
--:-- But I will keep harping on this: there's opportunity for startups and small companies in all of that. Palantir was an In-Q-Tel investment. We found them as a very small company and now it's driving a lot of the way we do warfare. So even those big companies will go, "I'm not good at this, but that company, that really small company, they're really good at that. I want them on my team." We want to unlock all of that and we want to enable it.
--:-- I'm trying not to be super technical just because this is not a technology problem. This is a culture and a process problem that we can apply technology to. I had a chance last week to sit with all of the big LLM providers. I loved the conversation because they were ruthlessly honest with me and with each other. All of these things are available — whether it's Llama or Claude or ChatGPT or even DeepSeek. What makes them unique and good and competitive is the unique data sets that their customers have and the unique ways that they're going to use them. That's us. That's my job: to be able to go, "Hey, here's the unique mission for my soldiers and here's how we're going to do that."
--:-- Startups can't solve a fundamental deficiency in us not being able to describe our problems, but they can help us identify where those problems are. This is why it's really important to be close to the warfighter. If you are a startup CEO, don't hire salespeople before you understand the ecosystem. Move to San Diego. Spend nine months near the group that you are trying to positively enable and empower to do their job 10 or 100 times better than they are right now. That proximity piece is real, especially for unique or specific problems.
--:-- Justin Fanelli: And then more often than not, there is some aspect that should be evaluated as dual use. From a shared services perspective, Alex already said it: we're doing more common and similar than dissimilar from industry. In some cases, we're not making the tough decisions to scale what's working fast enough. So if we have more shared services at the department level, how can we do autonomy together as services? How can we make sure that that is well integrated from the ground up and we're not looking at this as an integration problem because we made separate decisions three years later?
--:-- From an artificial intelligence perspective, right now we're leveraging an Air Force and Army and Navy solution. Can we single up for common enterprise problems in one place? And in other places, can we have narrow models where we share and all of them feed into some of the more unique problems like command and control? We're starting to architect in that way and make decisions based on what's available, not based on what we've done.
--:-- Host: One kind of an older comment you made was we have some things in common with VC. Sure, from a funnel perspective, that would be the ideal state. One of the tricky parts for us because we are such a big organization is finding people who think that way and fanning the flames and empowering them. Every organization has latent risk takers, even if they're conditioned to not take risk. The acquisition community teaches us that lowering risk, reducing risk is the goal. So essentially, playing not to lose. It doesn't really keep pace in many cases with the current trajectory of technological advancement. We have unleashed people, mavericks, who are willing to work against their self-interest to bring breakthrough capabilities through. I know Justin and I could sit here and talk really deep tech all day long. And we will.
--:-- Alex Miller: Don't do it. There's two things. The silos. And they just put something in my brain. I want everyone who listens or sees this to understand: we haven't missed the signal. The department hasn't missed the signal.
--:-- In 1968, something fundamentally changed the way that we do buying in the government. And it has not stopped changing since then. In 1968, there was a program put directly in the United States budget. It was the Minuteman 3 program. That was the first time that a program by name was put into the budget that we had to spend money against. Before that, there were just these big pots of money. The Army had vehicles and ammunition and soldiers. All of the money was just, "Hey, do what you need to do to run the Army."
--:-- Alex Miller: A portfolio, yeah, absolutely. And then in 1968, bam, Minuteman III is in there. And I get it. From a national security perspective, you want to make sure that money goes into a long-term project and you want to make sure that it's successful because we were the only people who were going to do that. We, the government, were the only people who were going to do that.
--:-- Now, today, the Army is only 21% of the defense budget. We are everywhere. We, the Army, are everywhere. And again, this is going to sound a little parochial, but we're on the border. We're the Global Response Force. We're the Immediate Response Force. We're the Homeland Defense Brigade. We're in Haiti. We're in Africa. We're in CENTCOM. We're in PACOM. We are everywhere. 21% of the budget and less than 50% of that — do we have any flexibility to spend to change how we're spending? Even less than that. There are thousands of budget line items that are directed to us in the National Defense Authorization Act everywhere that tell us exactly how we're going to spend that money.
--:-- So as Jess was talking about earlier, that three-year spin cycle. What that really means for us is we have to predict the future three to five years out, hope for the best, and then work in the year of execution to move money around the edges to try to do advanced technology. Otherwise, what happens is we give money to our labs and they try to do the right thing and they try to pull advanced technology in and they try to land this magical transition where it goes from something that works in a lab to something that works in the field in one go. And we know that that's not real because technology evolves so fast. You could go from something that is a concept to someone writing it up and doing it in the software space in days to weeks. And our entire system, the entire way we think, is predict the future three to five years out, land the perfect shot, get a trick shot and go. It just doesn't work. It does not work.
--:-- Justin Fanelli: This is all the more reason if someone does have a capability that is so much better than what we have in the field right now. We have stocked up the pipeline, but the data and the stories associated with that — we want it to be undeniably good so that we can then make the case. When people come to me and they say, "Isn't zero-touch AI better than how you're doing unified endpoints right now?" Yes, of course it is. We are trying to make the case that this will save us money. We need a little bit of capex so that we can get the opex right, but the total cost of ownership and the impact of this from a mission perspective is best for those who are protecting us.
--:-- Here are ways that we can lower the tail on ineffective things, but we have to aim at the exact solution that we have, even if it's a manual process. If we can find severable things, we can find enough room until all the workarounds are turned into scale and solutions that Alex is talking about. We can still pull that through. But if you can make the A to B not close — blow it away — that's what we want to see. We use world-class alignment metrics to show how a divestment is possible through a game-changing investment. That's a really big change. We want that to be 90% of the portfolio.
--:-- Alex Miller: You know, like a big percent of the — like they're each is right. When we still talk about DIU, it's less than 1% of the budget. These innovation activities are still very small. So it is a good signal that we're doing them. The impact of scaling what's working — again, once in a lifetime opportunity. Imagine how much more secure our country and the world would be if we were making the exact, ruthless, highest impact decisions across the board.
--:-- Host: So I want to double click into the culture piece. [Looks at door] Is that door locked to protect us? Yes, it is. The world I'm in, venture capital, we're in the business of risk. We're all in on risk. That's how we think about the world. Obviously, with the Pentagon, that's different. You all are making a lot of strides to change that. But I'd love your perspective on how do we get closer together between Silicon Valley and Washington and get that risk appetite up a little bit?
--:-- Alex Miller: We're assessing risk on different things. And it's at a necessity. I'm using big words. There's a lot of folks in the department who actually understand venture, whether or not they came from that or they've just been working with the venture ecosystem for a long time. I mentioned In-Q-Tel earlier. That's a really good, very public facing example. In-Q-Tel's ratio is 25 outside dollars for every federal dollar spent. Dual use takes care of the rest of that. This is one example where we've proven this model.
--:-- What we have to work through — and Justin said it really elegantly earlier — the people who take risk as their job, and I don't want to disparage anybody because they are hardworking Americans trying to do the right thing, their job is to minimize the risk to the government. The way that they measure risk is the potential for waste, fraud, or abuse of dollars. So what they're trying to say is, "Hey, if I give you a dollar, do I get a dollar of value out of it?" That's a good metric. What they don't do is: if I give you a dollar and I take 10 years to spend that dollar, is it still worth a dollar?
--:-- What we are trying to do is figure out: I'd rather lose a dollar right now. I'd rather try it on a company and they go, "It just didn't work out," than spend that 10 years. Because that is also super valuable time that we could be trying to figure out what all of the players in the startup landscape and all of the players in the venture landscape — and frankly, all of the businesses, because once you sell your first product, you're not a startup, you're a business — how do they get us the best war-winning capabilities right now? Maybe it sounds cheesy, but I mean it.
--:-- The risk calculus — we're not comparing the same things. What would help is that those venture folks talk to our leadership, continuing to engage, because we're going to work up. They need to work laterally to go, "Hey, the way you measure risk is fundamentally incoherent with where technology is going." If a quarterback was measured on not making mistakes versus throwing touchdowns or winning games, the performances would be 100% different. If your goal is to hedge against anything happening, then you're fighting against inflation, as opposed to against disruptive technology.
--:-- Justin Fanelli: We used to have something called cost as an independent variable. We make anything an acronym, so CAIV. This was to say, how do you back in and make sure that you're spending the budget and then optimizing for that? If we had speed as an independent variable — SAV — and we measured based on outcomes, then I think that looks pretty different. That looks more like executing.
--:-- When we say, "Hey, unleashed folks, what do you have?" They're not bringing 20% better capability. They're saying, "Hey, we're doing cross-domain solution right now through a bunch of different program offices. The Army's doing that as a service. Can we jump on cross-domain even if it'll affect or make some people that created something in-house upset?" Yes. For the sailor and for the Marine, yes.
--:-- Alex Miller: I just want people to understand who's listening how ridiculous this is. Cross-domain as a service is an API for binary XML between two networks. That's all we're talking about. Now, there's national security implications, but that is a solved problem that we are fighting over. We want to keep solved problems solved, get higher up the stack in general. I think it's academically interesting to resolve the same problem. That's for school.
--:-- Justin Fanelli: We have Marine Innovation Unit that is a way to keep really talented Marines in the reserves who have venture jobs as their full-time job. We're using them to scout technologies. We have folks in Europe right now working with some of the startup founders to say, "Hey, here is a breakthrough. What would scaling this look like?" Well, it would mess up our program of record plan. That's not what we're optimizing for. We have contracting officers sometimes say, "Well, I want to do fewer contracts." Well, I want to win. So there are really good contracting officers who get that.
--:-- Short of changing the way that we fuel incentives, measuring how much better — 25 times better — the "get after it" folks are in different organizations, that K-curve fuels us and it actually gives us more leverage to do more of this. Ultimately, send your success stories if you are doing this really well and you're already in and you've made it, because we want to double down on this flywheel of "this can happen." We don't want two successes a month. We want to overwhelm the system with how much better it can be.
--:-- Alex Miller: If you're not ready when the window of opportunity opens, sometimes you're waiting a really long time. We put Cape, a mobile virtual network operator company startup, in Guam. Then we needed them. The pilot was ready for actual use. If we are out hands-on testing with the best stuff, then we have an opportunity to scale in a significant way, much shorter than any of these timelines you hear about whenever you read about the government.
--:-- Host: Very optimistic. This is really exciting. It feels like we're in a lightning in a bottle moment where there's a lot about to happen. I'd love to just get your quick take on the future, the year ahead, what you both are most excited about.
--:-- Justin Fanelli: I'm excited about the idea that we don't have to prove that these things work anymore. If we scale half of what's working well right now — the software acquisition pathways — we have gotten more permission, significant permission, to do the right thing. Now, how many people do it and what does that look like and what wins do we get out of that? Ultimately, I am excited about participating in the race to have the best technology solutions and the companies that are offering the best breakthroughs to be fielded into that operations and maintenance pot and getting that to production to our warfighters' hands sooner. I'm excited about the best chance I've ever seen in the next six to 12 months to get capability that wasn't under government contract into full scale production.
--:-- Alex Miller: Hell yeah. I'm actually really excited about a lot of things. This year, we are going to put the consortium for the Army's next generation command and control into its program phase. Right now it's been in software pathways. Army Futures Command has been piloting this. We have a consortium — and I keep using that term deliberately — of just great companies who are going, "Here's what technology can bear."
--:-- I was at the National Training Center out in the middle of the Mojave Desert last week. I crawled into an M1A2 SEPv3 tank. Eight-year-old Alex was very excited because I was in a tank. But that tank commander — he was wearing the same socks. Eight-year-old Alex was wearing the same. This unfortunate soul all day long. I was sitting there with that tank commander and he showed me — because we have taken a 20-year-old tablet that's mounted in that Abrams that's really hard to move — and we've given him everything on his TAC device, which is an Android app on an Android phone. Now he can command and control his tank company. We're going to take that next generation command and control, get that on contract, start moving forward, and get that fielded to one of our divisions this year.
--:-- In Europe, we're doing this thing called Project Flytrap. I welcome anybody to come try it. It's a countering unmanned system. We are learning from Ukraine and what's happening over there. With 5th Corps and with UCRAF, we're actually going to work on how we scale up adding non-conventional sensors, non-conventional compute. How do we do automation so that if you were driving — and it doesn't matter what part of the world you're in — if you're an American warfighter and there is a drone threat, there is a bunch of automated decision mechanisms that are just going, "Hey, I sensed something in the electromagnetic spectrum. I hear something in acoustics. Here's what you need to do," and you don't have to worry about it.
--:-- In the IED fight, they taught us something. When you're going to combat readiness, they said: down, out, up. You have to look down to make sure you're not stepping on an IED. You have to look out to make sure you're not going to walk into an IED. You have to look up to make sure that you're not getting shot at. Now people have to look and make sure that they aren't seeing drones flying. I want to automate that.
--:-- We are going to put together a consortium for autonomy. We're going to take a lot of the vehicles that the Army has and think about what an autonomous formation looks like. How does that pair with people so that — and this is true — the U.S. Army should never trade blood for blood in first contact. It should always be blood for iron and it should always be their blood. We have to make sure that we can actually get the autonomy pieces together. It's not one company. It's not Miller Corp that's going to solve autonomy. It's how do we get the right tool chain in place from SIM all the way through VNV? How do we get the right actions and the right libraries? How do we get ourselves off of this sort of archaic microcontroller-based autonomy kick?
--:-- Those are three technology things off the top of my head. What you're going to do is just mask on the problem. All of the policy blockers that have stopped us — it's a "with our shields are on it" type of phase. We are going to body block our way through this to do the right thing because we can't not do it.
--:-- Host: I wanted to ask you guys a couple lightning questions. I could talk to you guys all day.
--:-- Alex Miller: Okay, very quickly.
--:-- Host: Very quickly. Red flag, green flag. What do you love startups doing? What do they need to stop doing when they come talk to you?
--:-- Justin Fanelli: Red flag: Don't work with bad requirements. These long requirements documents — there are enough capability needs statements out through software acquisition pathways and commercial solution offerings. Don't bang your head against the wall. Find a top level requirement or something easier to work with and people who get outcomes. Green flag: Keep building the product that you started on the path to build. Have your North Star and then work with us to figure out how to do roof shots as you're on your way to your moonshot, rather than trying to tailor it to us. The DoD — we are going to be a better customer. I wholeheartedly believe that. But if you had a vision, go after your vision and then get it in the hands of our warfighters to help tailor it.
--:-- Alex Miller: Red flag: Stop trying to use the Department of Defense as your only means of revenue. I mean that because if you get complacent with the Department of Defense, then you are no longer actually on the quest that you were on. It's just not good. We want to make sure that we have the best war-winning dual use technology. If we become the only provider to your technology, it's neither dual use, nor are we going to actually maintain the goal of partnering with industry. You were just a part of us at that point. Green flag: Keep imagining a safer world. We want the brightest minds to help every aspect of this. The goal here is to be so effective that we are getting as much return on investment as possible. That translates to protecting those who are protecting all of us every day. Technology is the advantage. If we have the best of America's talent working on these types of problems — our hardest problems — like we've fixed the alignment problem.
--:-- Host: What book should every defense-based founder read?
--:-- Alex Miller: *100-Year Marathon* by Michael Pillsbury.
--:-- Host: What is the number one misconception about your service that you would want to dispel?
--:-- Justin Fanelli: This is not your granddaddy's DoD anymore. We will be more flexible. We will be more open-minded. If you bring data to us, then we will work our butts off to make sure that if the policy is the problem, then that policy will get excepted. We've gotten 12 exceptions to policy because we brought data and said, "This is hurting us." These sub-optimizations that we've always had — we can do better than that and we will.
--:-- Alex Miller: I'm proud of the Army. I make a joke all the time. The Navy and the Army are the only two services in the Constitution. We've been doing this since 1775 and everyone knows what the Army brings. I say this just unemotionally: when the world dials 911, the phone rings at Fort Bragg. It doesn't ring at Naval Station San Diego. It doesn't ring at Anderson Air Force Base. I love my Navy and Air Force brothers. The phone rings at Fort Bragg, North Carolina when the world dials 911. The biggest misconception that I would tell people is: we haven't missed this. We're in the fight still. We're in the fight globally, and we're going to transform in contact and continue to transform in contact because we've been doing this since 1775. We are teammates and we are working together on these problems and we're going to start using that buying power to make more effective purchases. Because when the president calls for the carriers before they call Bragg, we need to be fully integrated and teamed. We fight, join every day.
--:-- Host: Love that. Last one, fun one: favorite military movie.
--:-- Justin Fanelli: *Top Gun: Maverick*. That happened on CVN 72, which has one of those hyper-converged infrastructure stacks that we talked about. So we're moving out.
--:-- Alex Miller: Is that how they did the post-processing? I am a big fan of *World War II*. I used to watch it with my dad. This is going to sound corny: I love *Starship Troopers*. I know it's a bad adaptation of the book. It is the mobile infantry. It is just people on the ground getting it done. So in the spirit of doing things a little differently, I'll go with *Starship Troopers*.
--:-- Host: Thank you both so much for your time. So great to know that there are folks like you working at the highest levels of government to help make some of these changes.
--:-- [Music begins to fade in]
--:-- Host: All right, that is all for today. If you did make it this far, first of all, thank you. We put a lot of thought into each of these episodes — whether it's guests, the calendar Tetris, the cycles with our amazing editor Tommy, until the music is just right. If you'd like what we've put together, consider dropping us a line at ratethispodcast.com slash a16z. Let us know what your favorite episode is. It'll make my day and I'm sure Tommy's too. We'll catch you on the flip side.
--:-- [Music swells and fades out]
