The Startup Industrial Base: Building for the Next 250 — Washington, D.C.

Read the Portuguese translation →

Author: YC Root Access · Watch on YouTube

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: The Startup Industrial Base: Building for the Next 250 — Washington, D.C. · Channel: YC Root Access · 8:00:09 · Watch on YouTube

Topics: Enterprise AI applications, Robotics and autonomous systems, Health and science

Summary

The event brings together Y Combinator leaders, government officials, lawmakers, and startup founders to discuss building a broader, faster, and more resilient American industrial base for defense and space. Speakers describe startups developing lower-cost drones, interceptors, autonomous vessels, navigation and night-vision systems, manufacturing technologies, cybersecurity tools, and space infrastructure. Government participants discuss acquisition reform, testing access, production funding, clearer requirements, and the challenge of turning prototypes into dependable suppliers. Across the talks, recurring themes include domestic supply chains, rapid iteration, reliable manufacturing, and matching new technology to real customer needs.

Key points

• YC CEO Gary Tan argues that a 20-person company should have a fair path to becoming a dependable military supplier, with clearer requirements, timely feedback, and opportunities to earn orders.
• Founders present technologies aimed at lower-cost, scalable defense capabilities, including munitions production, drone interception, autonomous maritime vehicles, GPS-denied navigation, and digital night vision.
• Government speakers describe reforms such as competitive testing programs, simplified problem statements, faster decisions, and efforts to reduce barriers for new entrants.
• Officials emphasize that reshoring critical components—including motors, batteries, magnets, and other electronics—is necessary for domestic production and surge capacity.
• Several speakers stress that a successful defense startup must understand program offices, requirements, procurement cycles, financing, and how its product fits into existing systems.
• The event highlights the gap between a successful demonstration and production, including the need for dependable demand signals, manufacturing expertise, and financing.
• Space-focused founders discuss autonomous spacecraft, lunar construction, orbital servicing, spacecraft manufacturing, and the need for launch access and shared standards.
• Speakers repeatedly identify testing access, integration, safety validation, and operational feedback as important constraints on startup progress.
• The closing discussion returns to the event’s central question: how to help small companies become reliable suppliers while adapting faster than changing threats.

Chapters

Start	Chapter	Summary
01:59	Opening and the startup industrial base	Gary Tan introduces the case for opening defense procurement to small teams and asks what it would take for a 20-person company to become a dependable military supplier.
25:14	Defense priorities, competition, and acquisition reform	Under Secretary Emil Michael discusses domestic production, lower-cost capabilities, competitive programs, simplified requirements, and routes for startups to reach government buyers.
53:37	Autonomous maritime vehicles and digital night vision	Splash Industries presents autonomous resupply and other maritime missions, followed by Deep Knight’s AI-based digital night-vision technology.
73:25	Transportation modernization and airspace	Secretary Sean Duffy describes plans to modernize air traffic systems, support autonomous aircraft and drones, and strengthen maritime and commercial space infrastructure.
133:44	Budget flexibility and getting technology into service	Under Secretary Hurst explains how defense budgets, acquisition reform, program offices, and financing pathways can shape startup opportunities.
154:40	Planning software, rocket cargo, and defense-market advice	OneBrief demonstrates AI-supported operational planning, Hop Arrow presents rocket cargo, and Eric Alberg discusses how founders can navigate the defense market.
202:49	The defense industrial base and production scale	Senator Tim Sheehy argues for acquisition reform, sustained procurement, and an industrial base that can support both commercial and military production.
245:40	Counter-drone systems and orbital fusion	Founders present counter-drone capabilities and Zephr Fusion outlines its plans for compact fusion power systems for orbital applications.
276:50	Space startups and NASA’s commercial role	Space founders discuss robotics, lunar construction, spacecraft manufacturing, and onboard autonomy before NASA Administrator Jared Isaacman addresses commercial opportunities.
300:52	Space-based radar and congressional modernization	Array Labs presents its space-based radar concept, and Congressman Pat Ryan discusses bipartisan defense modernization, testing access, and oversight.
343:34	Manufacturing, electronics, and battlefield sensing	Founders describe robotic manufacturing support, domestic circuit-board production, drone-swarm sensing, and intelligence-analysis agents.
389:30	DIU pathways and the mid-batch founders	A Y Combinator panel discusses early-stage defense companies, followed by DIU’s Chris Murphy explaining competitive contracting, transition, and startup access.
423:11	Iteration, insurance, and modular drone systems	Speakers discuss rapid adaptation to changing threats, insurance barriers for emerging technologies, and modular autonomous drones.
458:12	Interceptors, AI testing, and closing remarks	Wardstone presents a modular interceptor, No Labs discusses synthetic data for AI training and evaluation, and the organizers close by summarizing the day’s lessons.

Quotes

What would it take for a 20 person company to become a dependable supplier to the American military? — 16:44

The criticality is we we can’t do this by ourselves. — 23:33

I’d rather compete and know if my product works and doesn’t work. — 38:14

Innovators move fast, government moves slow, and if we don’t move at the speed of innovation, — 57:30

What would it take for a small company to become a dependable supplier to the American military? — 58:02

Concepts and entities

Y Combinator, startup industrial base, Drone Dominance, Defense Innovation Unit, One Brief, Deep Knight, Perseus Defense, Metalware, Hop Arrow, Array Labs, GR Space, Orca Aerospace, 64, Reality Defender, Drone swarm, Decision dominance, Dual use

Full transcript (original language)

01:59 Ladies and gentlemen, please take your seats. Our program will begin in a few moments. Please take your seats at this time. Thank you.

06:54 Please take your seats. Our program is about to begin. Please take your seats at this time. Thank you.

09:08 Please take your seats. Our program is about to begin.

10:57 Ladies and gentlemen, please watch this presentation on the screen. Thank you. Heat. Heat.

11:34 Heat. Heat.

12:17 Heat. Heat. Heat. Heat.

13:05 Good morning everyone. Thanks for being here. My name is Gary Tan. I’m president and CEO of Y Combinator. It’s great to have founders, government, industry all in one room. This morning when I went through YC as a founder in 2008, uh I was building a blogging company. Software was the frontier and the big challenge was getting it into the as many people’s hands as possible. In 18 years since, software has left the screen. It can see, it can navigate, it

13:37 can fly, it can run entire factories. We’ve got computer vision, autonomy, batteries, chips, advanced manufacturing, and most importantly, the engineers who know how to make it all work. Most of it was built for commercial reasons, but all of it can be pointed at defense. YC funds startups at the very beginning, often when it’s just a couple of founders and an idea. We’ve backed more than 5,000 companies since the

14:08 beginning, including Airbnb, Stripe, Door Dash, OpenAI, and Coinbase. So, if you’re wondering why I’m in Washington talking about Interceptors and Munitions, the honest answer is we followed the founders and what they want to build. Over the last few years, more and more engineers have come to us convinced that keeping America safe is a problem worth spending their lives on. And we agree. They want to build hardware and they want to build it here in America. They want to get it into the hands of people who need it. I want to

14:40 be specific about who those people are. Right now, there are Americans deployed on ships, on bases, at the edge of jam a jammed network, and they’re all depending on whatever equipment they were handed. They deserve the best this country can build, not just whatever happened to be on contract. Some of the founders building that that next generation of equipment are here with us today. They’re working on hard problems that give the US a real strategic advantage with systems that

15:10 don’t cost a fortune per shot, per mission, or per unit. One team is building navigation systems that let drones operate when GPS get jammed. Another is building interceptors that can take down cheap drones without costing more than the drone. Others are using rockets to deliver cargo to contested areas or building aircraft that fly in the stratosphere and stay up there for weeks. Then there’s the supply chain. You can’t have an independent defense industrial base if you rely on

15:43 your adversary for critical parts. Which is why one company here is making high performance magnets in America. Another is securing the firmware inside the vehicles and aircraft we operate. To the founders, you’re asking people to rely on your work in situations where failure can cost lives. Reliability, security, manufacturing, support, all of it matters, and it matters deeply. A great demo isn’t the finish line. It’s the

16:13 beginning of the obligation. Institutions have an obligation to keep the bar high, but to make it possible to clear so that When someone builds a product that works, there’s a path to actually getting it used. That’s why we’re all here. And it’s what we mean by the startup industrial base. America should be able to draw on its entire cap capacity to invent and build. The industrial base should be open to a team of 20 people with a better answer. Small

16:44 teams have a particular advantage. They can learn something on Monday and change what they’re building on Tuesday. The person who heard the customer’s problem can walk across the room to the person who can fix it. That matters when the threat keeps changing. Drones get jammed. Comms go down. Adversaries adapt. The people using our technology need to adapt too. So here’s the question I want us to work on today. What would it take for a 20 person company to become a dependable supplier to the American military? The big

17:16 defense companies play an essential role. So do the manufacturers, suppliers, and workers who’ve spent decades learning how to build things people trust with their lives. What we need is more ways for new companies to join them, more competition, more suppliers, more manufacturing capacity, more opportunities for a good product to earn a real customer. So let’s talk about faster rewards, clearer requirements, and fixedpric contracts where they make sense. Let’s give

17:46 founders a way to prove their technology under tough conditions and a credible path to orders when they do. My ask today is practical. If you’re a founder, tell the people in this room exactly what’s in your way and be specific. What test do you need? What kind of feedback from customers? What contract? What would let you deliver? If you’re in government or you’re a buyer, bring a real problem. Tell these founders what success looks like and what it would take for you to feel confident enough to buy. And if you’re at an established

18:18 defense company, take a look around. A team you meet today could end up as a supplier or a partner who helps you solve a problem you’ve been stuck on for years. We have people here who are willing to devote their lives to building for this country. Let’s give them a fair chance to deliver. Thank you all for being here. Let’s get to work.

18:57 Hello. Uh good good to see everyone today. Uh my name is Gorb. I’m the co-founder of company called Solugen. Uh we’re a winter 2017 YC batch and we were a strange fit. We’re this chemical company. We figured out this new technology to miniaturaturize chemical plants. And we went through YC, which was effectively a software incubator at the time. But what we realized was the same exact principles that you could apply in software could be applied to the chemicals industry and you could

19:28 move faster than anyone’s been able to do it. So today I’m going to tell you about a little work that we’ve been doing on the munition supply chain with our chemistry and what we have to offer. If you guys want to start the video. So in 2026, every missile, every munition, every platform is dependent on one plant in Tennessee called Holston. It broke ground 1942. The RDX pilot plant went up in 13 days, which is insane. Uh composition B from phone call to first shipment in three

19:59 three months. And in 1945, we had 1.5 million pounds per day. 2020 9 million pounds. So what we made in a year, we now take too long to make. So if you look at 1942, our munitions capacity was 300 million pounds for obvious reasons. 1976, 2016, everything kind of went sideways. We had to put in$1.8 billion to Tennessee. Uh but even after that, the the maximum volume that we got out of it was 30 million. Uh now if you compare

20:30 that to where China and Russia are, we’re probably uh two orders of magnitude uh off which is a problem. Even with all the announced expansions, everything like that, how do we do this safer? How do we do this faster? And how do we do this better? That’s really the fundamental question we had. And we discovered four problems in the supply chain. One, we talked about there’s just not enough munitions. Two is these unmanned platforms are getting smaller, so you need higher energ density uh energetics. stockpiles uh are uh sorry this is going too fast but that’s okay

21:01 uh the stock piles are are visible from space so we solved all four this is all to say that like we solved this problem we’ve put in $und00 million of private capital to solve this problem and the way we did it was by studying how it works today right now a chemical plant is a regulatory nightmare an energetics plant and an regulatory nightmare put those two together and you’ve got the worst combination possible so what we ended up doing was disagregating the two. We took the chemical plant, put it centrally, and we created distributed

21:31 production units for payloads. So that way you’re not taking the risk of an energetics plant and a chemical plant at the same time and allows you to build much faster. We got permitted in three months uh because of this technology. Um yes, thank you. So we we’ve actually executed on this across the United States. Uh our home base is in Houston, Texas. Um, but we’ve got two units uh uh in Houston and one unit in we’re building in Louisiana. Um, in the future we plan on having multiple sites, but

22:02 right now this is the the footprint that we’ve got. So the precursor is called Terraforge. It’s the Energetics precursor at scale. This is the plant that we’ve built. It does roughly 240,000 drones worth of precursor explosive. And then the Nitro Forge is the distributed payload production part of it. And that we uh is roughly 48,000 uh drones per unit. So one terraforge feeds five nitro forges fundamentally. Uh and we actually built a nitro forge

22:32 technology in in uh six months. So on top of that, how do we make a better uh energetic? So comp B is what our grandparents used to use. We need to make something much better. And what we did was we actually designed these plants to produce higher uh density energetics that allow your uh blast radius to increase. So, this isn’t theory. Uh, we’ve actually done this. Uh, we partnered up with the premier warhead designer, Kraken Kinetics. Um, and we actually built a fully KONIS advanced energetics

23:02 supply chain in 247 days uh for $100 million. And so, this is the actual buildout that we we did to to give you some context. Uh, this would take a traditional chemical company 5 years to build up. And so, we were able to do it much much faster because of the way we operate. This is the Nitro Forge unit that’s actually getting shipped to our arena testing site. So here the Nitro Forge actually makes the payload that gets incorporated into uh the drone. So with Kraken Kinetics, uh we actually said,

23:33 “Hey, these drones, they’re getting smaller and you need really more advanced energetics to keep up with that type of warfare.” And so fundamentally we we’re going to need dozens of terrafores and probably 100 to 150 nitro forges. So the criticality is we we can’t do this by ourselves. So we actually launched uh a company with Kraken Kinetics called Odin Energetics. Uh Odin Energetics is focused solely on creating uh uh payloads that we can scale across multiple drones. So if you

24:04 look this is basically where the Nitro Forge uh technology is born out of. Everything’s made in Houston. uh we deploy it uh with international partners everywhere in the world. Um we we do believe that with the advanced energetic you actually get a better uh kill um per per cost per kill fundamentally what we end up what we need uh to succeed is Emil if you’re listening uh I just need

24:35 someone one lead on the R& side just one lead that can help us over the next 30 days to do a sprint we’re going to show you guys that we’ve solved this problem and now we need to scale it and I can’t do that by myself so thank >> [music]

25:14 >> Good morning. Uh I’m Luther Low, the head of public policy for Y Cominator. So excited to see everybody here this morning. I’m going to be your uh part-time MC and I’m so excited to welcome to the stage uh Emil Michael and Jason Miller. Uh Emil Michael is the under secretary of war for research and engineering at the department and the department’s chief technology officer leading its research development and prototyping efforts. Before returning to government, he served as Uber’s chief business officer and held leadership roles at Tel Networks and D-Wave. He previously serves as a White House fellow and special assistant to the

25:46 Secretary of Defense. Jason Miller is a longtime senior adviser to President Donald Trump and a veteran political strategist and communications adviser. He’s held senior senior roles across President Trump’s presidential campaigns in transition efforts helping shape political strategy and public messaging. Uh please welcome Under Secretary Emil Michael in conversation with Jason Miller. [applause]

26:19 All right, good morning. Good to see you. >> So, it’s great to be here this morning with one of the rising stars in the President Trump’s administration, Emil Michael, under secretary for research and engineering. So, I want to dive right into it. Yesterday we were all watching on TV on whether it be Fox or CNBC uh with all of the tech leaders and the president. Uh president made it very clear where he is on super intelligence with data centers. How does something like yesterday then immediately impact what you’re doing at the Department of War starting today.

26:49 I mean it was an extraordinary collection of really some of the if not all of the country’s most important technology leaders um from Jensen to Elon to uh to to open AI’s leadership anthropics leadership um Google I mean it was really a collection of people who are are really going to drive the super intelligence race for the country and essentially for the world given where the United States is and to get them to

27:20 agree on a framework for how we uh develop these technologies for the benefit of the American people and do it safely um without sort of an overbearing regulatory scheme uh was really quite quite amazing. and and then having the president be able to sort of not only gather these people but get them to agree on on core principles that they may have disagreed on uh individually was really kind of really fantastic act and I think we’ll look back on this

27:51 moment where they signed this White House compact for the safe development of super intelligence as a watershed moment where we decided we’re going to win and the industry players decided we’re going to work together to win to do the all the things that you want to do auditing and independent valuations and those things. So, you know, just having being been there and watched it was was pretty pretty incredible. >> And so, if you’re communicating with the the founders that are here today, how will what happened yesterday then impact

28:22 them as they’re in the this early stages of their company’s lifespan? Well, I think it I what I hope it does is ensure that we’re not headed for some notion of regulatory capture by one or two companies in the AI in the SI world. We are headed for a world where we’re going to have a lot of competitors. We’re going to promote open source. Um we’re going to ensure that um companies that are t talking about risks internalize those externalities and make sure they’re spending enough on on the safety

28:53 part of their model development process and model deployment process. And that’s good for the ecosystem. What you never want is one or two companies sort of driving regulation that keeps the small companies out. And hopefully we’re on a path to make sure that does not happen. So, a moment ago you talked about needing to win. Of course, when we talk about needing to win, we’re talking about beating China in the SI race. Uh, make sure it’s not I said SI, so don’t get me in trouble here. Uh, but the we have to go and win this. And so much of

29:24 what’s critical for the companies that are here today is how do we go and win in their respective fields, whatever companies they’re doing. You’re one of the few people that I know who’ve actually gone into China and won. Uh, is the chief business officer for Uber. went in there with approximately I believe was it roughly $2 billion investment. Uh received a lot of criticism and I believe you guys exited with a $7 billion valuation roughly. Okay, I might rounding right about there. What was it that you learned going in there that you that now applies directly to what you’re doing the department of war about going headtohead

29:55 and competing with the Chinese? Well, one of the things I learned doing business in China, um, during the Uber days, and this was about 2014 to 16, um, the the the Chinese are great at copying and and we know that, right, the using American IP, but there is this tipping point where where there’s they’ve run out of stuff to copy, they start inventing. Um, and that’s not something to be underestimated. and and I do think

30:27 even more so now with some of the things that where they are now almost 10 years later that they are inventing things um and they have spare capacity in their manufacturing base. So they have the ability to produce and mass things we we don’t here. Which is why this reshoring of critical supply chain technologies and items is so important because if we don’t have that then we can’t do mass either. And that’s why you see us doing critical minerals reshoring batteries

30:58 for drones, brushless motors. We’re going to have to go after robot parts, actuators, all the things that go into the things we want to build up. uh we’re going to have to have those be in this country in addition to re-industrialization so that we can compete on mass and quantity and the reason mass and quantity means you’re seeing it in Russia and Ukraine mass attraitable lowcost drones and counter drones those are the way sort of territorial battles will be fought in the future so no longer be just battleships and so on um so we need the

31:28 ability to do those things too and and and you know we’ve let China do these things for a long time and don’t underestimate their ability to invent when they shouldn’t underestimate our ability to reshore and ensure that we have our our own domestic supply chain. >> So I want to stand that point for a moment as you look at out here all these young faces and people are starting these companies. What is it that the founders in this room can do that their contemporaries uh in China cannot right now? >> So there is there’s still uh a few

31:59 things that are uniquely valuable about the American ecosystem. Number one, the entrepreneur uh entrepreneur class themselves the ability to taste risk. It is not shameful to try and fail in this c country. It is in some ways a badge of honor. Sometimes VCs want to bet on you after you fail even more than your first thing. Not that you should be trying to aim to fail, but it is it shows you’ve got grit, that you’ve tried uh and you gave it your all. And there’s still some social um sort of negative about failing

32:31 over there. And that means their risk-taking appetite’s different. I’d say the second thing is our capital markets are still five times as robust as theirs are. Um, our VC community, they they take risk on small on smaller companies. They bet on people. Um, they understand sort of again what losing means and what winning means. and the the just the sheer dollars of capital how quickly those dollars come out in our country um are again a huge strategic advantage for the nation here.

33:01 So I want to stand that with regard to efficiency because one of your u reputations is someone as a chief business officer Uber and a lot of time in in startup type companies is efficiency now that you’re inside the building at the department of war inside the Pentagon. What is that gap between what the Pentagon and the supply chain before you came in was producing versus what we actually needed? >> Yeah, I mean it was and maybe every administration says this, but when I

33:31 when we got in there, the sheer neglect of what was happening was [snorts] was staggering. It was neglect on critical minerals, neglect on the supply chain, neglect on ship building things we’d already spent money on, neglect. We canceled hypersonic missiles. We cancelled I mean there was just an an allergy to business. Um an allergy to AI and technologies. When we got in there, we had less than 50,000 people out of a out of a department of three and a half million people using AI

34:03 when the rest of the world was using chat GPT on your phone. uh we were just not using it at all. Today we have have about 1.7 million out out of our three and a half million. So just bringing the basic technologies to make people more efficient uh was totally ignored. We were single threaded on one AI provider. Um we you know if you’re a defense contractor working on an important program that we were spending $50 billion of taxpayer money in, you got one meeting every four years and your program was failing. So all these things were happening and we had to get our

34:34 arms around uh all these things in the last year and a half to get them on track and we’re doing I think we’re doing an excellent job getting these things on track and new entrance not a shot. Not a shot right there was no way we were going to buy a lowcost cruise missile like we are from Androl and a few others. A lowcost hypersonic uh drone defense you know drone production we were producing hundreds not millions like the Chinese. So none of that stuff was even under underway. >> So I want to stay on that point with regard to efficiency because you made it

35:06 very clear that we have to be efficient. We have to get the cost down. When you look at something say such as the u uh the cost per kill when you talk about shooting down a rocket that’s coming from Iran. How much of that dollars and cents versus the the sheer scale debate comes up as you’re trying to do both at the same time? >> Yeah. So, so my counterpart in acquisition and sustainment, he’s he needs to build more Patriots, which are the expensive way to take the these things down. I’m doing the cheaper way. I’m trying to do directed energy um

35:39 things that cost thousands of dollars, not millions of dollars, because it’s otherwise it’s asymmetric, right? You have a a country like Iran that could, you know, send a $50,000 drone if you’re spending millions over time. The math does the math, right? So, I’m trying to figure out other ways. So we’ve launched uh and transition a joint weapons laser system golden dome that architecture is built on directed energy lasers high-powered microwave things that could take things down at a really cheap rate and that’s that’s critical not just

36:10 because of the money but just because then I could deploy these systems in a lot of places and that deters the threat right so you think about deterrence too what do you have to have so that they’re deterred from even doing that well if your systems are all expensive they’re not going to be deterred and they’re bespoke token small number if they’re proliferated small cheap well then they’re deterred from doing that so that’s the that’s the end end of the spectrum that I try to work on and so if you’re one of these founders and you have a startup what type types of things should the founders be saying no to as

36:41 they’re looking through the uh the life cycle of the business and where they’re growing to uh where do you look and say here are obvious inefficiencies or things that people need to be and and again look it could be founders it could be the uh the primes who come in, where do you look and say this is this is wasteful. This is not going to work for the long haul. >> You know, it’s um a lot of this is on us at the Pentagon. We’ve got a lot of rules, qualification standards,

37:12 um things that were written 20 years ago about what the product you might be building should do that haven’t been modernized for for the modern era. So, we try to when we find those things, try to kill them, right? Um but we also try to create competitive environments um that give companies a real shot. Like for example, we built created this program called drone dominance. So what is drone dominance? It’s like not you just applying with your drone product one-on-one. We have a competition. We invite companies. You can apply. We take

37:44 the top 20. We put you in a flyoff situation. We give you objective metrics. We measure you against other companies. And at the end, the top three get some money. So there’s something for you to invest in in entering that context. And we keep narrowing it. We call them gauntlets. And at the end, we have half a billion dollars of orders to give um to to companies that have won. So you get transparency in how you’re doing. You got, you know, objective metrics on how you’re doing. And then you have an

38:14 order, you a financial order at the end state of each one of the things. So if you’re a small company, you’re like that works as opposed to knocking on 20 doors at the Pentagon not knowing if my product is what you want for what theater when. Uh and so we’re trying to do more of that in every in every category. Uh and for me, if I was on entrepreneur sitting in your seat, I was like, “This is great. I’d rather have this. I’d rather compete and know if my product works and doesn’t work. I’d rather know if I lose and why. And I’d rather if I win, get some money and know that I’m building towards something.”

38:45 >> So on that point, we have drone dominance. number of the other programs you’ve helped to launch. Some of the folks who are in this room though won’t naturally fit into a certain category. So what was your advice then to these founders as they’re trying to decide am I trying to solve a current need or is this a great idea that I think will go and transform things? Where’s where’s that balance? >> It’s a good question. Um you know what we’re trying to do what we’ve done is we’ve done something called requirements reform which sounds really boring but I’ll tell you why it’s it’s not that boring. Um so typically

39:17 um if the military wanted you know some munition or some piece of equipment they write a 300page RFP and that RFP was designed for a lawyer or a lobbyist and a prime contractor to know how to fill out and they can just do it over a weekend whereas a small company like reads it and you’re like you know I have to figure out what all these things mean and if you get one thing wrong you get kicked out. So, um, we’ve tried to simplify it and say, well, the new way we’re going to do requirements is we’re going to do here’s my problem. I need to

39:47 fly a missile 500 miles in this kind of environment, and I’m going to kind of have it be a one or two page description of the problem I’m trying to solve. You come back to me with solutions. So, now that gives entrepreneurs a creative way to uh to answer with different technologies that we might not have thought of. So, that sort of takes that barrier down there. And um after that then you get a more quick well that’s that’s not going to work in this scenario or the timeline doesn’t work for us or you know the payload’s not work for us but then you have some

40:18 feedback right away. And so instead of having that process be 3 months and then six months and then nine months then twice. So you’re now you’re two years in trying to shorten that all into to 90 days so you know what you’re doing. If you have a totally novel thing that we haven’t thought of, there’s enough people who will listen um like me and like people on my team who will put it in the back of our minds and make sure that when we’re looking around what we need, maybe maybe it comes up later. >> So you talk about um lobbyists for example in DC, we have some smart people

40:48 in this room, people can figure out how to uh you know in their spare time figure out how to get you to Mars and back in the same afternoon. Just geniuses. Uh but then I get the question all the time, but I don’t understand this budget cycle and dealing with you have you have the Pentagon and then you have the OM and then you have Capitol Hill and you have these influencers and you have the media and all these things going on and you have a lot of uh young companies who are saying, “Do I need to hire somebody? Uh do I need to uh if if we’re so thin on dollars, we’re trying

41:20 to scale and do and innovate at the same time.” At what point does a company look and say they probably should hire someone in DC to help them? >> You know, it’s a hard one because in theory, you want it to be open so that small companies aren’t spending an amount of their dollars on these kinds of things. Um but and it is true that people think, oh, there’s this, you know, huge defense war department budget. Uh but really Congress every

41:50 line item they tried to prescribe where that budget goes and try to constrain us into well that’s what you said you wanted well and that was 18 months ago when we started passing this budget in committee so we try to get broader definitions they try to get lighter definitions um and but I I do think early on it’s a little not that useful and I’ll tell you why we we used to have you there the innovation ecosystem was spread meaning we had the cyber sitter which is sort of a small grant program

42:23 and then we had DIU which was reported somewhere else and then we had appetit and then we had the office of strategic capital all over the department now they’re all in my world so whether it’s a small company we’re doing a small initial thing a medium company doing a scale up a bigger company doing a prototype or or an important company that’s got a product that works now they they’re eligible for a big loan that’s low cost so that can fit them in where they are and hopefully that reduces the need to do some of the lobbying uh

42:54 because at the end of the day you want to be a program of record and that’s where you sort of become sort of a you know a stalwart in the Pentagon. So maybe until you get closer to that point you could you could hold off hiring a lobbyist. >> So one of the things that you said when you came in is that you want to have a a series of fast yeses and fast nos. Um, so tell the folks here, don’t have to necessarily name companies, but what types of things have been presented to you where uh this has been uh whether it be a demonstration, whether it be a pitch, this was a fast yes. And then

43:24 what types of things has been put in front of you where you’re just like, this is a fast no, Trevor, how this person even get in front of me. Get him out of here right now. >> Yeah. For, you know, one of the things that we were had a real deficit on was was hypersonic weapons. you know, missiles that go Mach five or above and the cost for the primes to develop one was like $50 million and it took years and years and years to do it and we essentially stopped during the last administration. So I said we need

43:54 to find some lowcost high volume ones. So we found a company Castellian and there wasn’t a lot of competition for them which was a fast yes. is like, well, let’s get them in testing on our test ranges, you know, with some with some contracts for some test articles. And that sort of made sense because what they were doing fit the time, the time that we needed and the cost structure. Um, and they were using a new engineering approach. Um, from a no standpoint, you know, we are getting to the point where there are so many drone manufacturers

44:25 that if they’re not in the drone dominance program, um, and we haven’t been coaching them in how to, you know, get their supply chains more Americanized, um, and they’re just their capabilities are the same. We get to the point now after a year of spending it where we’re, you know, it’s it’s tough to go back on that unless they have done something unique that we haven’t seen before. So on that point, I want to stick with this for a moment because if you’re I’m sure we have a lot of people here who who make drones are involved in the uh the counter drone space. If somebody’s not already in the drone

44:57 dominance program, how do you get in front of that? How do you enter that? We’ll have multiple rounds of this because we’ll go from, you know, the group one, group two small drones and then we’ll do a competition from group three and four and we’ll kind of meet you have to cover the range and then we’ll have the counter drone and, you know, we’ll have different flavors of it, but we’re going to keep evolving after we find vendors that work in each category, you know, and we’re not going to go down to one, it’ll be two or three. Um, we’re going to keep evolving it because ultimately the unmanned systems arsenal is pretty broad. We just

45:29 started with the easiest and then we’re going to keep moving. >> Uh, so you posted something on social media this summer. There’s a a picture on maybe it’s on Instagram, but you were only in America. Of course, you’re someone who was u born in Egypt, grew up in New Relle, someone who went to Harvard, went to Stanford and just very very successful and America is really the only place in the world where you you can do this. I was at dinner with a number of the founders on Monday evening and realized that of the I think I was with four of the founders uh that I was the only one who was born in the US. Of

46:00 course, I haven’t founded anything and it was remarkable that everybody wants to come to the US and wants to do business here and start a company. What are the the things that these founders who are not from the US and again regardless of industry that they need to be thinking about as they’re setting up a company in the US or they’re redomiciling to the US uh where you’ve seen folks find pitfalls or other traps things they need to be looking out for to make sure they can succeed as an American company. Um, I mean, if they’re, you know, if they don’t have a US citizenship yet, I mean, it’s hard,

46:31 you know, as you get, as what you’re building becomes more central to the Department of War needs, you’ve got to, it’s likely someone’s going to need a clearance if not you, uh, at your company. Um, and those things, you know, could get more complicated if you’re if you don’t have US citizenship. So, I mean, that’s the only thing I can see is is making sure that uh that you’ve got some, you know, sometimes we take a board member, sometimes we take the GC, but someone has to be able to do this if what you’re doing is is, you know, falls

47:03 over the line in terms of a classified program or classified materials or science that we want to classify. But generally speaking, um we haven’t seen that challenge just yet. I mean, we talked to a lot of found every founder um from here uh probably come, you know, has talked to someone at Department of War. I’d be surprised if they found a roadblock, you know, if they’re a foreignb born. >> I certainly haven’t. >> So, one of the things that you said you want to have uh before we open up to a couple of questions, uh you said you want to have five new primes pop up uh

47:34 before President Trump’s uh current term is finished up. Um, if you’re a founder and you’re here today and you’re doing something with the Department of War, you want to be doing something with the Department of War and you want to you want to be that prime. What is it that a founder should take away from here today to go back and tell their team to tell their investors say we if we want to be the next prime, this is what we have to do. Yeah, I’d say that the uh a prime is is in our parlance a company that brings

48:07 multiple product and multiple disciplines to the table. Um and therefore they’ve sort of mastered maybe the selling process, but uh they also have a range of things that make them a prime. Otherwise, they’re a one product company. There’s nothing wrong with one product companies. Um but when you look around the table uh they they want to have you want them to have enough engineering talent, enough production capacity that as our needs evolve, they can be bidding on the next thing even if

48:37 they haven’t invented it yet. And that makes them, you know, another prime that you can go to that can be competitive on a on a certain project. Take Golden Dome for America. You know, we’re we’re building that and we just put out an RFP for space-based interceptors. That means something in space that could take down a a hypersonic missile in the first few minutes of launch. A very complex phys, you know, physics question, something that’s hadn’t been done before. Um, would I like do I, you know, are there the prime primes bidding on that? Sure.

49:08 But should new companies that are doing space stuff, if they have capabilities and understand how to get into space and have done things in space, could they then stretch to have another product line that does something like that? Yeah. So that that’s what I consider a prime that they have domain expertise and can extend it for based on the needs of our department. >> You guys heard it here. Space is hot. The secretary likes space. Uh Luther, do you want to open up for a couple of questions? >> I think we got time for about two questions. Uh let’s maybe in the back. I’m going to run it back to you.

49:40 >> Hi, I’m Mike Stone from Reuters. Thanks for coming in today. Great meeting yesterday. Question. Uh yeah, of course I’m a reporter. Anybody else want to ask a question? I’ll do it for you. >> It’s okay. >> It’s okay. We’re going to do it now. So, sir, we talked about drone dominance. All the drone manufacturers that I speak to, they have a problem with uh brushless motors. The winding machine costs $300,000. In order to get the cost per brushless motor down to whatever China can give

50:11 us, 50 bucks, they need to do way more volume. Should the United States government get into the brushless motor business? Should you be a merchant supplier? >> Yeah. So, the the Office of Strategic Capital just announced um a drone components deal with a company called PDX that’s going to be developing brushless motors that and supply that that product to uh drone manufacturers. And we’ll probably see another few deals in the coming months for drone

50:43 components like brushless motors batteries and the things that do that that are done here in the United States. Um just for that reason, right? Because there are some key components in drones that that uh are heavily China based in terms of the percentage produced and those perfect example of the things we have to domesticate here if we want our drone companies to be on American supply chain and if we want to do surge capacity for those things over time. So, we’ve our master spreadsheet of things that that that uh we have to do things

51:14 like that for. Um, and it’s hard because we’re starting from a sta standing start at the beginning of this administration and with President Trump’s leadership, we are looking at all those things and uh trying to pull them in. >> Luther, one more. >> Yes. Uh, under Michael, special adviser Miller, my name is Yuki. I’m retired Ukrainian diplomat uh that spent some time in Silicon Valley uh working with Ukrainian tech, Ukrainian defense companies. So I have some experience there. Uh from from that perspective, I

51:46 would like to ask you uh how do Ukrainian companies and Ukrainian defense ecosystem in general uh can uh be a become a part of the uh startup industrial base in the US. We know about drone dominance program. what is else is out there for for them? Thank you. >> Yeah, we we’ve opened the drone dominance program to Ukrainian companies who um if they win, they would have to domesticate their operations in the US. Um and a few of them have participated

52:18 and have done well in the competition. Some of it’s tied up in export controls by by the pres president Zilinski and letting that happen. Um but I’m optimistic. There’s certainly learnings that the Ukrainian operators have um that would be valuable to us. So, you know, we’ve been pretty open on inviting those companies to participate in drone nominance so long as ultimately they have to domesticate here in the US. >> All right. Well, I know we’re at our time, so join me in thanking Secretary

52:49 Michael for coming to work with us here. [applause] Thank you. Excellent.

53:37 Okay, good morning. I’m Ivan, CEO of Splash Industries, and at Splash, we develop lowcost USVS for contested logistics and maritime domain awareness purposes. Now, I know what you’re probably thinking. There are a thousand other USV companies out there. What separates us from the rest? Ultimately, it comes down to our highly vertically integrated approach. So, we develop all of the autonomy stack in-house. That’s from the camera systems to the electric actuators to every single circuit board

54:07 on the boat. And of course, the software, the most important part, that is all developed by us. What does that actually mean in practice? What it means is that we’re able to deliver a vessel that is 5 to 10 times cheaper than current similarly capable solutions. Our goal from the very beginning was not to just build another million-doll boat, but really to build something that can provide affordable mass to the fight. Uh they are a lot faster to build. So it takes one of our technicians 8 hours to build one of our Typhoon USVs, the one

54:39 shown here. Uh and they are incredibly easy to use. There really is no learning curve to our vessels. We can train people in a matter of hours. Now, I want to be careful and make sure that this does not sound like a sales pitch because it definitely shouldn’t. The U uh the Navy, the US Navy already owns and operates a number of our vessels. The Army recently placed their first orders as well. So, instead of just listing stats, what I thought we would do here is that I would go through some of the cool missions that the Navy has gotten us to do uh over the summer. So,

55:11 we’ve done probably 12 exercises this year so far, and I’ve gone through and selected three of the ones that really stood out to me the most. Really, what I want to do here is to show rather than tell. Uh, so for the first one, we’re going to be hopping over to Hawaii where we were tasked to deliver cargo to a number of warships off the coast of Honolulu. Now, traditionally, this would be done using uh helicopters, larger ships, but in any case, it would be manned assets. You would have people on

55:42 boats or helicopters, of course, very expensive, and in case of conflict, you’re actually putting war fighters at risk, which is not great. Our goal here was to demonstrate that we are able to deliver smaller, but well, smaller components quickly to these warships. And you’ve probably seen this video at some point. Uh this is us approaching LHD2 uh doing an autonomous beach landing. And it’s really important to note that this was not remote controlled. Uh this was all controlled from a different location using autonomy. Uh so no one

56:12 was holding a remote controller next to it. That’s not really the point. That is not scalable. Uh so during this exercise, we proved that what it is possible to autonomously resupply a warship. Uh that’s the first time that a USV has ever done a fully autonomous resupply of a warship while it was on the move. So it wasn’t stationary. Uh they were transiting at around 10 knots. So after we completed that, the Navy said, “Cool, let’s do that, but 20 times further away.” So uh on the following day at Rimac, we went out on a grueling

56:44 170 nautical mile mission to deliver cargo to CVN71. Uh so that was in SeaTate 5, mostly at night. Uh you can imagine that out there in Hawaii, 170 mi out, uh you don’t get many opportunities to take photos. So really we got some good onboard footage, but that mission once again was a success and really showed the reliability of our vessels. Uh this is what our C2 looks like. Just a quick sneak peek in there. The operator can see all of the vessel stats, the live feeds, and then be able to plot out

57:15 missions on the map. Uh now, so it showed just how reliable our vessels are. A lot of our competitors when they saw that mission brief, they thought we were crazy to even attempt it because usually USVs a lot of time are used with chase boats. This was without any chase boat. If anything would have failed on our USVs at that time, we would have never seen them again. Uh so for the next one, we’ll hop over the quad lane, I guess. Quick show of hands. Who’s been out the quad? Okay, we got one. We got one. Oh yeah, Max. [laughter] Uh yeah, so one thing that is important

57:46 to note about Quad is that it is very, very far away. So getting any equipment out there and doing any sort of uh missions out there is hard and that is where another important part of our USVs comes in. They are just about small enough for a single person to launch and deploy. And since they are using an internal combustion engine, we can dump that fuel and ship them in commercial aircraft. So quite literally, you can have an unboxing experience for your USVs. Uh they’re also lightweight enough to even hook up to a golf cart. We did get into trouble for that at Quad. So,

58:16 don’t do that if you go to Quadr with your USVs. But that all goes to show just how easy it is to deploy. You don’t need docks, you don’t need harbors. All you need is a beach uh and a dolly uh to move the vessel around. So, what were we actually doing out in Quad? Um first, we were training local participants on using our USVs. Um these uh so uh the participants here, they have never used any remote control system in their life. Uh so the fact that we were able to get them trained up in a matter of hours to

58:47 successfully execute uh medical supply delivery missions that really speaks volumes about how easy it is to use our USVs. You don’t really need a PhD uh to operate them. So once we got the training out of the way, we embarked on a few other medical supply delivery missions across the ATL. Uh so those were about 150 mi. The weather was varying here. Uh we got some great weather and you can see just how clear the water is. That’s probably 60 ft of water, but you can see straight down to the bottom. Uh, and on the way back from, oh, actually, I’ll not jump ahead

59:19 here. That’s what a beach landing looks like. So, the vessel gets out on shore. An operator on the receiving end is able to unload the cargo, grab whatever medical supplies, cargo you have on there, and push it back in the water. So, really, it’s quite easy to use. And on the way back from one of those missions, we actually stumbled across a sunken World War II German uh cruiser that was used for nuclear testing. So that was a just a cool thing that I thought I’d include. Uh and those are just the rough mission tracks of what we did uh on that all going both outside

59:51 and inside of it. Uh so that mission really showed just how easy it is to transport our vessels. Out of the 44 companies competing in this, we were the only USV company to complete all of those missions successfully. Uh and of course the low cost of the vessels also helps as they are going to be getting procured by RMI ideally uh Royal Marshall Islands. So a million-dollar boat doesn’t really cut it for using them in a medical supplies delivery use case. Uh now for the final one I have to be a little bit careful. Uh I have redacted pretty much everything I need

1:00:22 to redact there. So this was during a force on force a war game exercise with the Navy. But I have done a very nice graphical recreation here to kind of get across what we were able to do. Uh yes, my art career is just getting started. So uh in this mission, we were tasked with surveying and if possible destroying enemy assets at a island, uh here marked in red. Now there were a number of rivers connecting these assets to the main bay. Uh but the enemy being a smart enemy blocked access to that

1:00:53 channel. So, what we were expected to do is go in the harbor, go in the bay, uh, survey the enemy outpost, and then get out of there. That being said, we did not take no for an answer and try to get in. So, the way they blocked it off was using these thick steel cables, so normal boats aren’t able to pass. If you pass with a normal boat over this, you’re going to foul your props. But, we don’t have props. We have a water jet, which allows us to do the beach landings. So, what we were able to do is after a little bit of moving back and forth, get over that and get into that

1:01:26 enemy outpost. Once we were in, things got easy. So, our precise mission autonomy allows us to navigate at fairly high speeds in decently narrow channels. Uh, and when the channels got really narrow, and I mean like 15 ft narrow, I could jump across that and I’m not that athletic. Uh so even when it got as narrow as 15 ft, our vessels were able to navigate in that river and setting uh which is pretty unique. They’ve never seen that before and definitely that was not an outcome that was expected from

1:01:56 that exercise. Here is just a picture of what that looked like uh towards the very end. So you can see this part here that was 15 ft wide. Other USVs obviously couldn’t get in there. Uh and this was also not remote controlled. So we were sitting 30 m away. uh no one was remote controlling the boat because that would have been easy. It was all done using our autonomy. Uh so that showed that we have a very unique river capability, an attribable price point and the exposed props the lack of exposed prop gives you a pretty unique

1:02:28 advantage in some of these use cases. Of course, your draft is a lot shallower. You’re able to get in places where no one else is able to get in. Uh here I’ve only selected three of the cool missions. We’ve done quite a few. So, I would be happy to talk to any of you about more of our exercises, what we’ve done, including towing a water skier behind our USBs. I think that’s the world’s first as well. Uh, so I’d be happy to talk to you about that. I’ll be hanging around today. And if you want to reach out, that’s my email as well. So, thank you so much. [applause]

1:03:34 Hi, good morning. My name is Eric Bradom and I’m from uh Deep Knight and it’s my pleasure to uh talk with you this morning. Today I’m going to uh talk about a problem that we actually solve. Everything these days needs to see in the dark. Drones, unmanned vehicles, dismounted soldiers, not only the soldiers that are fighting, but the soldiers in reserve, the technicians and logistics support chains uh for them um need night vision. uh marine vehicles

1:04:04 there there’s a whole uh uh even man- operated vehicles everything needs to be able to see in the dark and it needs to be able to do it digitally. So the revolution that I’m going to be talking about today is AI based night vision that runs on commercial imagers and commercial processors that gives you uh night vision surpassing the latest analog tubes. But again, it does it digitally at a very low cost.

1:04:36 And you don’t have to take my word for it either. I’m actually doing demos uh in a small room over here. So please uh stop by. So the the problem in in using uh CMOS sensors is that a lot of them are blinded by the dark. You know, when it gets dark outside and you want to take a picture, you hold your phone up, you try to take a picture and it says hold still and it counts down 5 4 3 2 1 and then you look at it and the picture’s not even that good. Anyway, um what we’re talking about is the best

1:05:09 level of night vision. 90 frames per second, 120 frames per second, multiple channels on a little tiny embedded processor. And that’s the that’s the paradigm shift we’re talking about today. So, if you take a a standard off-the-shelf CS CMOS imager, for the in most cases, the the image on the left is what you’re going to see at night. And this particular setting is 1 millux, which means there’s no moon in the sky,

1:05:39 only stars, and it’s hard to even see your hand in front of your face. Uh there are some specialty camos imagers that the the army created with uh different companies. And if you use one of those, you can get a pretty good uh view, but still, it’s not it’s not clean enough to be able to use downstream AI. It’s not really clean enough to understand the uh the surroundings, but as soon as you apply dominite to it now, you can see the rocks that you’re standing on. You can see the water’s

1:06:10 edge. You can see across the bay, the beach, the stars, and the sky. Completely different level of performance and context. Our AI is actually recovering true information. There’s no hallucination. There’s no guessing at what you see. We’re basically using the AI to remove the noise that is actually generated by the sensor. So, it’s revealing the signal that’s that’s in the scene. So,

1:06:40 if you look at the raw output uh from an a dominite camera on the left and again, this is a high performance CMOS lowlight camera that’s available off the shelf. Um, you can kind of see there’s a fence in front of you, but you don’t know if it’s a defense installation or or what may be ahead of you. And you can’t really tell what you’re walking on. Uh but as soon as you flip on the uh the deep night solution now you can see you’re walking on pebbles. There’s a sewer grading ahead some obstacles. The fence is for a tennis court. There’s

1:07:12 people on the tennis court. And if you want to run downstream AI to do uh object uh classification detection tracking uh you can do that now because the the image is full HD and it’s so clean. And again, this is this is no moon in the sky, only stars, and you have a tree canopy. So, this is the darkest that that it gets. >> Um, the army’s actually been testing our technology at Fort Belvour in uh C5 ISR,

1:07:43 the night vision labs, and they maintain demonstrations of our technology if you’re ever there and would like to see it. Um, on the left you have a Gen 3 night vision tube. This is only 1,800 FOMO, but we’ve tested up, the Army has tested us up against the 2376 FOMO tubes as well. Uh, and you can see how that compares to running Deep Knight again on a CMOS imager. Uh, so here you can see in full HD. Uh, you can actually read what’s on the sign. Uh, you could do facial um, recognition if you needed to.

1:08:16 Uh, and run all kinds of downstream AI. So it’s it’s exciting because it’s an order of magnitude lower cost, higher resolution digital and you can run downstream AI and communicate the signal. It’s also more cost effective than thermal. So uh a lot of uh drones for example will use thermal imagers. They’re much lower resolution and much more expensive. Um, so if you look at uh

1:08:46 this is a 640p thermal imager on the left. Um, and again you can see it very well, especially as an indication of where the target is. But when you turn on deep night now, you’re getting a full HD image. You can tell that the person is on grass. You can see there’s a cat there. Uh, you can read what’s on the sign because we can see the reflective bands. Uh so again it’s it’s a it’s a shift in thinking and and one of the things we’re trying to do is help people understand that instead of using a thermal imager um and struggling with

1:09:18 the cost and export restrictions you can use a CMOS imager and run deep on it. Um and in fact um we uh will be in production with the drone company in fourth quarter uh this coming quarter and uh uh testing early testing by um by one of the military programs shows that by using low light instead of thermal they gain 20% better um navigation and target accuracy

1:09:48 while reducing cost. This is another image from a drone. On the left is the raw cos imager output and then on the right is what you get with deep night. Um this is the worst case. This is a 1 mlux environment. Um and the bright lights on the ground don’t wash everything out. They stay they stay in place. Um and this just shows you what’s possible. I’ll also mention that the AI takes the movements of the drone and is able to counteract for motion blur. So

1:10:19 that’s why the the image is so crisp. Uh so we are currently pursuing our round B funding. Um we are working with customers and dismounted solder soldier applications. Uh we’re on we’re in the supply chain for multiple SBMC um uh primes. Uh we’re doing drones, vehicle vision, counter UAS, and and much more. Uh we have a lot of new features coming too. Um, one of the new ones coming is color night vision. And

1:10:50 we think that after people get used to color night vision, they won’t they won’t want to go back to monochrome. It gives you so much more context of your environment uh especially when you’re trying to navigate a vehicle. Uh, so that’s what’s coming. Uh, we were just awarded an OTAA contract uh for the Dutch program um to apply our processing to thermal cameras. Uh, so that’s just getting started. And I mentioned that we’ll be in production on a drone in fourth quarter. Uh and at the bottom is

1:11:21 just some of our uh customers that we work with. Um and I guess the uh in terms of what we’re looking for uh from from the government uh we are developing we are funding the development of a lowcost super high-performance digital binocular that can be upgradable to color. It can be upgradable to add a thermal sensor. It can it can be it has a long upgrade path um but a very low

1:11:52 cost of entry uh so that night vision can be made available um to all the all of the troops and the supporting troops. Uh we’re doing that on our own dime with the investor money uh and we’re looking for potential customers. uh that’ll prototype in the April time frame and uh then be it’ll go into testing and uh hope to find customers uh shortly thereafter. So with that, I’d like to say thank you and please come find me in the uh the demo room across the hall here. Thank

1:12:22 you. [applause] Can uh my uh the voice of God person encourage people to uh come out from the uh reception area uh for Secretary Duffy. Um I highly encourage everybody

1:12:53 to check out the deep night demo. Uh they’ve got it set up in a a private room over here. Um please check it. It’s very cool. Um, so it’s my pleasure to introduce uh the conversation between uh Y Cominator CEO and President Gary Tan and Shawn Duffy. Shawn Duffy is the 20th United States Secretary of Transportation leading the department responsible for America’s aviation, highways, railroads, and other

1:13:25 transportation systems. Before joining the administration, he represented Wisconsin’s seventh congressional district in the United States House of Representative and served as an Ashlin County District Attorney. Please welcome Secretary Shawn Duffy in conversation with Y Cominator President and CEO Gary Tan. Hey everybody.

1:13:56 Vibrant morning. >> Oh yeah. >> Or not. I don’t know. You guys need some coffee? >> Yeah. Everyone needs a little coffee. Secretary Duffy, thank you so much for joining us. >> Good to be with you. >> Uh let’s dive right in. I mean, the America’s transportation system reflects decisions made really generations ago. uh if we get this decade right, what should Americans be able to do that they can’t do today? >> So, first uh at at DOT we have uh so I have uh pipelines, I have ports, I have

1:14:30 rail, passenger and freight. I have roads and bridges. I have cars. I have airports, airplanes, airspace, uh commercial space launch. Uh so NASA is one thing but whether it’s uh uh blue or X or anybody else. So it’s a it’s a really big department and there’s been uh I just think a long time where it is kind of on autopilot and the world has changed a lot um over the course of the last 5 10 20 years and so not a lot of

1:15:02 innovation has been injected. There’s a lot of conversation about how we change the way we do things but not a lot of implementation. And so one of the biggest hardest projects that we’ve undertaken is uh is modernization of the airspace. And what I mean really what we’re doing is we’re going we need we need new radar and there might be new ideas on radar. We were having a conversation Gary and I were before of like what does America accept as new technology and what won’t they accept? radar,

1:15:32 digital radios. It’s like really exciting. Voice switches, a surface awareness system so you can see the airfield from the um from the tower. We we have air uh uh uh our air traffic controllers will pass an airplane off from one controller to the next by using literally a paper strip with information on the flight. We’re going digital. We can actually swipe uh the information from one controller to the next and the cockpit can see it at the same time. Um

1:16:03 we use like really old copper wires. We are going to go to fiber. It’s really exciting fiber. Uh we have a old uh telecom architecture. We need to build a new architecture. You guys see what happened on the east coast about 10 days ago. Whole east coast shut down. I had one of my lines cut. uh we have this redundant system with that should be redundant but it all has a single point of failure which means it’s not redundant and some of the lines don’t always work and so one line gets cut the redundant line doesn’t work and so the

1:16:34 east coast shuts down and all the while all these foreign leaders are trying to fly in for the for the UN week and so we are going to we are fixing all that and I got 12 half billion dollars from Congress to uh to to do this infrastructure build but how we manage the airspace case is is uh you’d be shocked at what we what we use. Um and so we have humans that sit in our command center and they will they’ll have six different screens and they’ll have a a paper flight chart

1:17:07 above them and they try to assess all this data and make decisions on weather and on flight trajectories. And we have really wonderful tools um that are actually old tools now, old technology that we could use AI and and predictive analytics to determine what weather’s going to do. So instead of having a ground stop at an airport for say two and a half hours, it might only have to be 45 minutes, but the ground stop you get today is for you know two and a half

1:17:38 hours. We have So if you did anyone flu a da here, we’ll have Yes. So, we’ll have there’s one runway, right? We’ll have uh four airplanes get to DCA at the same time, and only one airplane can land at a time on a runway, right? So, um we would say, well, if we could predict that two hours before they’re all going to get there, we could speed up two of the airplanes and we could slow one of them down so all of them get there when the runway is open for them

1:18:09 to land. Like, not complicated stuff, but we don’t do any of this. So, we’re trying to, by the way, I’m I’m getting I got some crap for saying you’re you’re turning over air traffic control to AI. I’m like, this was a really bad like it’s dumb AI, not smart AI. >> But >> moving from paper from paper and pencil to something even just digital, >> just go digital. So, the systems are so old, it’s really hard to to to rebuild. um while you’re keeping a airplanes have to fly. The system has to be safe and

1:18:39 all the while you’re you’re you’re changing spark plugs while you know your your car is running. It’s happening and I just the pitch that that I have is and especially in aviation I need answers. I need solutions. I need really good partners. all some of the some of my partners uh I’m shocked that um not all of them lot I’ve got a lot of really great partners but a lot of partners will have contracts with us for years sometimes decades and they just don’t they don’t prod it’s it’s a it’s a

1:19:10 payment that they get that they don’t really produce for us. I need I need companies to be really innovative to think of what we do and think of what solutions they have for us um and how we can use the best innovative tools to make our system work better, make it safer, but make it more efficient. I think everyone is sick of the cancellation and the delays. We can do a way better job. Um and especially when, by the way, you guys all probably know this, we have the busiest airspace in the world, bar none. like we fly there’s way more commercial, way more

1:19:41 general aviation um in America um compared to anywhere else. And so what the rest of the world does doesn’t really work for us. Um but what we do here actually leads the world. Um so what the tools that we’re going to use is, you know, the rest of the world is going to buy a lot of what we do. We were the first in flight. We led we, you know, ushered in the jet age. We like we we’ve led aviation and this is one of these critical points. And so, um, we’re looking for good solutions and good partners. Um, and but just if anyone

1:20:11 wants to come and see me, if someone says, “Hey, I got this tool that’s pretty cool. How could you use it?” Like that doesn’t that’s like I’m like, “I can’t think for you.” It’s great if people go, “Hey, I got a tool that could solve this problem for you, right? This is a pro. This is this is how we could, you know, fix this.” And so this this is a ASI has built this smart tool which by the way small company that competed with the biggest companies move fast really innovative and they got the contract a big contract with us that again they’re focused on the US but they’ll I’m sure

1:20:43 they’re going to sell that around the world. Um we don’t care about the size we care about the speed and the innovation that comes from the company. So um that’s exciting. I think that was the maybe your question was along this but this is this is exciting stuff that um we’re doing that will change the experience I think for the American people. Um and it’s smart companies that help us make that happen. >> I mean it sounds like it’s uh a little bit a matter of will. You know can we obviously assuming that we can keep everything as safe as it ever was or increase the bar of safety. Um do you

1:21:15 want your flights to be more efficient? Do you want, you know, people to move from this part of the country to this other part uh 30% faster, 50% faster just by better coordination? Like these are things that are basically readily available today. And that’s what you’re trying to >> We’re not we’re not we’re not breaking the innovative, you know, ceiling here. It’s like this is these are tools that just aren’t used, but it’s also tools that maybe haven’t been used in aviation that you can think through that how how can we deploy them here. If by the way, when we’re more efficient, just think how much jet fuel we save um and how

1:21:49 much staffing costs we save. It’s billions of dollars a year we save um in dollars, but the amount in carbon emissions is substantial. Um so, no, it has a real impact. One other thing, and again, I’m this I’m not this other the tool when you fly, um love the airlines, they’re wonderful, but they they they’ll Delta sees Delta United Seas United. And what they do is, let’s say the an airport can handle in a 15-minute time period 15 airplanes.

1:22:21 Well, the airlines will schedule 32. Well, if I can only handle 15 in a 15-minute time period, but you schedule 32, you can have delays. You’re going to like you’re you’re all going to be lined up, you know, trying to get to the end of the runway and take off. It’s just we’re building in delays to the system. We’re burning fuel. We’re burning time. I’m giving the information to the airlines to go, hey, could you guys do a better job on your scheduling? Could we actually spread this out a little bit? Because you got time before and time after. If you spread it out, we don’t have this chaos at airport. So that’s

1:22:52 not me dictating, but that’s me asking and giving information from the tools that we have. Anyway, so aviation has consumed like 80% of my time. >> Yeah, it sounds like there’s already headway being made on, you know, smart or the modern skies act. So smart is the is is this ASI tool we got at the command center um in uh at DOT. So we have you have a tower, you have a Treycon, you have a center. Um a lot of this infrastructure is very close

1:23:23 together and we’ve divvied up the airspace based on the technology we had in the 50s60s. um with new technology means you could radically transform how how we think of the airspace, but then also how we how we structure uh what we’re doing on the ground. Um and so with with ASI, we are uh we’re starting small. We’re starting in the DC region just make sure everyone’s getting comfortable and then we can expand it from there. And I think by next summer we’ll be driving uh more

1:23:53 efficiency. And so, um, we’ve asked the Congress for 30 billion more with our modern skies coalition, which by the way seems like a lot of money, but if you think about the cost to the economy of the inefficiency that the current system drives, um, this I think it’s I think it’s a great investment. I’m a fiscal conservative and I find myself asking for money all the time. Uh, and there’s a great return on that investment. But, um, yeah, we we’ll have more money coming. We want more tools, more software, and we need, I think, better. We can

1:24:25 help we can help medium-size airports um find more resources uh to build more. Listen, if you travel the world, beautiful airports around the world, we should have that same kind of beauty and infrastructure in America. I think we deserve that. So, we want to help facilitate that. I need private equity to get involved and help. There’s bonding authority that can make that happen. Also, the a lot of the towers that we use, they’re they were built in 1960s, 1970s. A lot of you get a leaky

1:24:55 roof. You’ll have toilets that’ll leak down to the break room, you know, down below. Um, but we’ll we’ll we’ll replace every tower in America um over the course of the next 300 years on this plan. So, uh yeah, not great. There’s also opport and again this is a conversation we have to have internally, but do you need a tower? Are there some places that we should have um you know infrastructure where we put cameras and sensors and then we can have controllers

1:25:27 see that in a in a you know whether it’s in the tower or in the Treycon or some other location to control um control the airfield. Well, you probably get better vision, you get better data, and you probably get some better results. But that’s a conversation I have with, you know, air traffic controllers and how we do that. It’s there’s a whole bunch of components but the technology is there to make that happen. It’s just a matter of is that where we want to go uh in this country. >> Yeah. If you get redundant networks then you know as you modernize the different systems they actually make the system

1:25:58 safer and then you don’t have to have direct line of sight >> because you have you have you have you have cameras and you have and and >> backup cameras >> and and and um >> you have systems that’ll give you awareness of where every plane is on the runway. um you know from ADSB or um other tools that we have in the works. >> Absolutely. I mean down that road um you uh autonomous aircraft and drones they’re becoming useful in infrastructure, inspecting pipelines, moving cargo, supporting national

1:26:29 security. And it sounds like um one of the things you guys have been thinking about is uh how to help ensure safe drone operation beyond that pilot’s line of sight. as sort of a a next frontier of what drones would be capable of doing. >> Yes. So, we’re working on beyond visual line of sight for drones. So, I mean like airplanes, drones are similar in the sense that uh airplanes are used commercially and they’re used for war, right? Those two things are how we use airplanes. Drones as we see what’s happening in um Russia and Ukraine.

1:27:00 Drones are uh weapons of war, but they’re also weapons they’re they’re tools of commerce. And again, whether it’s yet uh pipelines, train tracks, delivering your Uber Eats uh to your home much more efficiently, taking uh taking uh vehicles off our roads. Um and so would maybe this is an overall philosophy for us. I innovators move fast, government moves slow, and if we don’t move at the speed of innovation,

1:27:30 uh and again, we have a bedrock of safety. We have to make sure the systems are safe, but we have to move faster. So innovators, if you’re too small and and you’re trying to innovate in America, you’ll you’ll go under. Or if you’re big enough, you’re going to just you’re going to go somewhere else. You’re going to take your business somewhere else and you’re going to innovate somewhere else. I want the innovation to happen in America. I want the jobs in America. I want the technology in America. And I want American companies to deploy that around the world. And that means we have to move at a different speed whether it’s with uh drones um but also and

1:28:03 we’re by the way I think we’re we’re we’re behind in um drone technology. We’re catching up very fast. Um working with um our our national security assets to make sure we’re we’re we’re seeing what is what’s commercial and what’s not commercial um is incredibly important. But you know, Evatolls as well, these the vertical lift uh um advanced air mobility, Jetson, you guys know what I’m talking about. Uh Joby, Beta, Archer, uh some of the bigger ones. Uh this is I

1:28:34 mean I think this is the cool technology that we’re we’re trying to advance. Uh and we’ve got a EIP where we got eight test zones where we’re testing allowing these test flights to happen. And then we’re getting data off the vehicles that’ll help us certify those vehicles. Um, which is very cool. There we have a pilot in those vehicles. I imagine at one point they completely can be autonomous. The question is, does America want that? Do you want to have vehicles that are autonomous or do you

1:29:04 want to um do you want to keep a pilot in those vehicles? But it’s the same thing with autonomous vehicles. are moving quickly to make sure uh the our autonomous um technology can continue to evolve here. Again, if we if we import someone else’s um autonomous vehicles, the data that comes off those vehicles is substantial and we send that ve that that data in America somewhere else. Probably not great. We’d love that data to be here, which means the innovation has to happen here. The companies have to be from here. And so um we’ve moved

1:29:34 uh I think pretty pretty well at the speed of innovation to allow them to continue to grow in this country. So >> yeah, autonomy is pretty wild one in that uh I don’t know if you saw this study around uh the amount of bodily harm and injury to human beings in the course of being you driven around or driving around uh can be reduced by as much as 90% by things like Whimo or self-driving cars. And when you look at that and you know autonomy broadly like this is sort of the promise of this technology obviously it’s uh you know a

1:30:06 long time in the making and you want to do it very safely but you how do you think about that balance it’s like you know when is when is it ready certainly like in the next we’re talking about like this generation like this next 10 years. This is so I would say that that um it’s it’s not a science, right? It’s I kind of call it my Goldilocks philosophy. We we can’t go too fast where we don’t have the safety bedrock because then people get hurt. Um and then innovation will be shut down by the public and we go too slow and uh innovation leaves or or it dies. So we

1:30:39 we have to get it just right. Uh, and it’s a continual analysis of making sure we have systems in place to drive that safety culture, but also we’re we’re we’re we’re giving um a long runway uh for those innovators uh to continue to expand their technology and and and grow at the same time. >> Sounds like that’s a a conversation we’re having across all parts of autonomy really in in sort of >> intelligence. Can I say just AI is exciting, right? It’s or frightening. I

1:31:10 don’t know what what wherever you sit is how you think about it, but >> this room would probably say it’s very exciting. Yeah. >> So, um probably the most exciting thing that’s happening in America right now. So, I seed that to you guys. But the next mo next most exciting thing is all happening at DOT. So, and I’m not biased when I say that either. Um a little bit maybe. But if you think autonomous vehicles, you think of drone technology, you think of evatals, the way people and products move, the way uh innovation affects your everyday life. Um, no

1:31:42 department is going to impact you more than what we’re doing at DOT. And that’s why it’s so important to get it right and and to have smart people, smart companies, um, help us, you know, make sure that the the government’s not the impediment. And I think government so often is the impediment. It’s slow. and it’s not very insightful. Um, and everyone as safety is our safety is the bedrock is what I talk about. But people can wrap themselves around safety and and and that can be their excuse for

1:32:13 doing nothing. And by doing nothing, um, it actually becomes far more risky. It’s less safe by by not advancing the technology. Um, >> it’s not a it’s not an economy. It’s both. >> It’s both. It’s both. That’s right. I mean, NASA is all about safety, but if all they do is safety, you never put people on rockets and launch them into space. Like, it’s it’s it’s inherently risky and you have to manage the risk, right? Because we’re putting people on rockets, but but it’s we want to send

1:32:46 people to space and so there is some risk that we have to tolerate. And if there’s zero tolerance for risk, I mean, just chain up your airplanes, you know, chain up your innovation, shut it down, and like send it somewhere else, which we don’t think is the answer. Uh, I want to ask you a little bit about maritime. We saw a splash just now. Um, ports, shipyards, uh, domestic manufacturing sit at the intersection of uh, both economic strength and national security. Um, how are you working to build a stronger industrial base on the maritime

1:33:17 side? So, um, just I mean, a couple things. Uh, I think that it’s a national security issue that that America doesn’t really build ships anymore, at least in the commercial space. We were the greatest ship builders in the world. Um, bar none, by far, going through World War II. And we kind of gave up. We we gave it up. And we look at what’s happening in the Middle East right now, and you go, “Well, shipping’s pretty important. We don’t really think about shipping a whole lot, but it’s it’s a really

1:33:47 important part of commerce. Um, and as we see, there’s a there’s a correlation between national security and and commercial. And they there’s there’s dual use. Having a vibrant ship building industry is incredibly important because that feeds into our ability to build military ships as well and and and vice versa. And basically, the I’m a free enterprise guy. the rest of the world has subsidized their ship building and they stole it from us and we gave it up. Um, and so though I believe in free enterprise, if you want to build ships,

1:34:18 you you you can’t think that you can play the free enterprise game and beat the guys that are subsidizing ship building. And so I do we’re work we have a really good coalition in government that’s trying to think through how we bring ship building back, which by the way maybe doesn’t seem like the most innovative space in the world. Um, I it’s interesting like like rocket rocket companies when we talk about ship building like they get really excited about ship building. There must be some cross-section between rockets and ship building. And I don’t know like that you guys probably get that. I don’t really get it. You’re smarter than I am. Um, so

1:34:51 there’s like we have to bring I think ship building back to America. Um we’re deal we uh we’ve uh we think that and we’re in a little bit of a race with China but uh uh small modular reactors we should have nuclear on ships. Um that is on the commercial side. That’s the that’s the we we’ve we’ve uh we’ve sent that out to industry. We have three ports that are willing and working with us to accept um uh uh uh SMRs. And but I think we have to send the signal that we

1:35:21 we’re open. we’re going to work with industry and that is I think on the commercial side the next uh generation of innovation that can give America a a leg up and again it’s not just ship building obviously this is a conversation that’s happening in a in a lot of spaces especially with uh data centers um but we think there’s there’s a good correlation between um that and ship building obviously uh we’re trying to develop this at NASA and put that on the moon as well as a power source so a lot of uh multi-use uh sources for for

1:35:52 nuclear >> uh secretary, if it’s okay, we wanted to open it up for maybe one or two audience questions. >> Sure. >> Anyone? Yes. >> Only good questions. >> Only good question. >> Good questions. >> At YC, we have a rule. It has to be someone who has a question that helps everyone in the audience. >> Hi, my name is Kush. I’m in the garden batch. I’m making unmanned airplanes for fix uh inspecting oil and gas pipelines, but eventually I want to get into making large aircraft for moving cargo around. So I was curious to know what’s your opinion on how that regulation would

1:36:22 open up for large part 23 aircraft over,000 pounds u which can actually move around cargo. >> I’m getting odd feedback. I didn’t I can’t I What was the question? >> Sorry. Can you repeat slow more slowly? Sorry. >> Yeah. So I want to build like large cargo aircraft that can move cargo completely autonomously. Um so over few thousand pounds maybe. And I was curious to know what’s the regulation going to look like. >> Is this in shipping you’re talking about? >> Uh no airplanes. autonomous airplanes. Oh, yeah. Okay. What’s my thought on that? >> Yep. >> Um, listen, I think it’s I run into a

1:36:55 whole bunch of issues, right? I uh pilots have issues with autonomy um is coming, right? Uh and I think as I as I look at um kind of where do we deploy first, flying in Alaska is very dangerous. Uh Alaska doesn’t have a great road system, so they fly um supplies to a lot of these uh these uh outpost communities. I think that’s the first place we go is um using autonomy

1:37:26 autonomous uh airplanes to to fly in. And again, we have we have great companies that are working um on autonomous aviation. Uh and so listen, we’re we’re we’re we’re open and again we drive safety on this front, but there’s a lot of work that’s being done right now in the space and and and we’ll be good partners when that comes live. There’s a different political conversation that has to be had, but in regard to, you know, strict technology, drive safety, um we’re all in.

1:37:57 >> Last question. Dennis Willie with a company called Astroscale and we do inspace logistics. So that’s my question. uh what is what can we expect from the department of transportation’s 28 budget with respect to developing and investing in in space logistics not just the launch complex but the ability to do things in space to

1:38:30 support the economy there [snorts] >> so we I mean so so commerce has the in space um we have uh earth to space so the launch licenses um I just from from From from my view, I think we have to make sure the industry is robust. Uh we need a lot I think we need to have a lot of companies. SpaceX is unbelievable. They’re remarkable. Um it looks like Blue is uh after their uh launchpad um issue in at the Cape.

1:39:01 They’re they’re they I think they’re by the end of the year, starting next year, they’re going to be able to launch again. They’ve rebuilt their pad remarkably fast. Um the president signed a uh an executive order uh telling us to to build more launch capacity. So we did an RFI request for information. Where should we go? Uh and I see us as investing in the infrastructure uh of this new spaceport. Obviously the the companies can they have to build their own pads and I think there’ll be

1:39:32 multi-use pads as well. But we need the capacity to make sure that um smaller companies have the ability to launch. They don’t get pushed out of um uh uh launches because we don’t have the facilities available to them. So, you know, Wallops is getting full. Obviously, the Cape is is full. Um so, we have to see over the horizon on what’s coming next. And I think we’re going to have a lot of companies that want to get in the space. By the way, all you all you wonderful uh

1:40:02 billionaires or future billionaires, uh you used to buy horses and blow your money on horses. Now you’re all buying rocket companies, which is pretty exciting. And uh it’s it’s a cool it’s a really cool space, but it needs to be more vibrant. We need more competition, more players in the space. Um and so that’s that’s my role. and and just as just so in our commercial space off uh office it’s it’s difficult because a lot of talent can make a lot of money at at uh

1:40:34 at SpaceX or at Blue or you know pick your space company we we we’re trying to hire another 70 people if you look at the cadence of launches that’s going to happen not just now but into the future we got to bring in good people that can help us um with one we need regulatory reform that we have to move move much faster we have to license much faster and it’s hard for me to hire the right people because the industry is soaking everybody up. Well, the industry can’t be successful unless we’re successful um in doing our part as well because we’re

1:41:05 we’re you know, we’re regulators, but we’re also partners in making sure that America continues to lead in space. And by the way, this is one where I want to make sure these launches happen in America. I don’t want to see them go somewhere else. So, that’s um that’s what we’re working through right now. But it’s an exciting time in you know in commercial space. Do you want We’ll do one more. Did you have a question? >> Yes. >> Hi, my name is Theo. I’m actually industries. What you mentioned about airspace tracking with optical redundancy example you’re building. So thanks for the pitch. Um our question is

1:41:35 I’m sure for a lot of sensing companies FAA is something we’re very very grateful for keeping American sky safe. But at the same time risk tolerance might be too high for a lot of companies who are coming in trying to secure more airspace doing more activities in the sky. What sort of government incentives slash strategic can we work on to improve sort of the openness into working air? >> That’s a that’s a great question. Um so uh again it’s we take this on a casebycase um analysis. So again we have to deal

1:42:06 with drones right. So drones are you know 400 feet. How do we how do we manage the the the growing commercial use of of drones in in cities? Um how’s that going to be managed not just with one company but when I have four companies that are playing in the same airspace? Um and then you know with then the Evols come into play and they’re going to they’re going to layer right into you know the current system that’s used which by the way if you look at all the growth that’s going to happen um again the most complicated airspace the busiest airspace that’s going to grow

1:42:36 what you know two you know two and a half times in the next 15 years especially when you add in new technology uh new new vehicles um we need we need the tools in place to to effectively manage Um, we’re going to have supersonic flight over land, right? Very cool. Um, and again, that’s us being able to to uh to move with industry um and and move more quickly. I guess my as as we as we wrap up, uh, my pitch is

1:43:08 we want the best ideas. I don’t know what I don’t know what the next I think a lot the next administration, no matter who it is, a lot of the work we’ve done, they’re going to continue with like we’re doing we’re doing because what we do is for America. Like this is not there’s nothing partisan in a town that’s all partisan all the time. What we do is really nonpartisan. Maybe 5% 10% you could say feels partisan. But this is this is American infrastructure. It’s American regulation. It’s American safety. And so I think what we do is going to live on past, you know, past my

1:43:40 tenure. Um, but then thinking through I got really I I I see your problem. I have a solution for you in coming in and talking to us or I see a problem in the airspace. This is my problem with my vehicle. Can you uh can you accommodate us? Can you help us? And so I’m not trying in the air like airspace usage. I’m not trying to go like how like what’s the new technology that’s going to come out and how do I have to think? I can I I’m just trying to think through what I’m doing right now. But if you have issues, you should you should flag them for us. One last thing I’ll mention

1:44:12 is um if we’re going to continue in aviation to to uh uh to to be the innovative hub in the world, we we have to be able to certify airplanes. And if you look at what the EVA what what the Evols are going through, the pain they’re going to go through um uh in in the certification um of of uh of their of their vehicles, it’s too laborious. you you know you it’s hard to certify Boeing aircraft where’s a new innovation in in aviation

1:44:44 and from from the big producers it doesn’t happen because it’s so hard to certify so we need to think through again we need to make sure we have safe aviation but can we move faster a lot of I think there’s a lot of private capital a lot of innovation in aviation that’s not happening because they’ll say we just we just it’s going to take too long too much money to get um to get through the FAA um and and and and allow this craft to fly. We have to do a better job of that. And the same kind of energy

1:45:14 we’ve put into modernization of the airspace, we now need to put in we got to have two. We’re going to keep the same push there, but also think through how we uh move through uh the the certification process much faster. And that’s going to take us time. anybody wants to join us at DOT or FAA, we probably don’t pay as well. Um, but if you know anybody that wants to partner and come in and serve their their country, we are we’re open for business and uh you can be an SGE or you can be

1:45:46 an employee. But we need and this is what’s different in America people and I think Elon did this. This is when Elon was willing to come and serve with Doge, all of a sudden people said, “Well, maybe I would maybe I’d be willing to give back, too.” And we’ve we’ve got good people helping us out, but I need everyone likes to bash their government. I join in bashing my government um and how slow it is and how inefficient it is. But if there’s people who say, “Well, I’m I I don’t want to bash it. I want to fix it. I want to actually help.” And you’ll be frustrated by it. You’ll be government is frustrating.

1:46:17 It’s slower than you move. But um this is making your government work is making the private sector work as well because there’s so much overlap. And so if you guys know anybody that has made a little bit of money and wants to come in or wants good experience, uh we would we we need the brightest minds helping us out. >> They might be watching online. We have uh several thousand people online right now. >> Reach out. We want we want great we want great people. Um and then in the end it’s way better for everybody else. And that’s how we that’s how we change

1:46:47 America is by letting our departments move with all of you. And instead of stopping you, let’s all move together. But it’s good people helping us out um making that happen. >> Secretary Duffy, thank you so much for your leadership. >> Thank you. I appreciate >> appreciate you.

1:47:34 >> [music] >> All right. Is this thing on? Cool. Hopefully you could hear me in the back. My name is Dr. Jason Cornelius, and today I’m talking about making missiles as cheap as bullets. So, this is our Mark III interceptor, a two foot long rocket meant to shoot down group one and

1:48:05 group two UAS. We build these for about $2,500 each. And we just tested it 3 weeks ago at Yuma Proving Ground uh with JITA 401, Northcom, and NORAD. So, here’s a little bit about the company. And we have a booth in the back, so feel free to stop by later and check this out. I’ll just set it here for now. Um, we joined YC Summer 25 batch and we did um All right, just to make a note, the timer they’re just adding it now, so all good. Um, we did YC Summer 2025 and then around August, you see the Wall Street Journal article on the left side there.

1:48:37 Um, April 2026, we did testing at Fort Hood with First Cav and US Army DevCom. That was our Mark 2 interceptor, a three- foot long, very not maneuverable missile. And we got feedback from the end users, specifically Jaya 401, saying, “Hey, can you make this a little bit more compact, a little bit more agile?” 4 months later, we did an entirely new clean sheet design of a missile. In 4 months time, and demonstrated it just 3 weeks ago at Yuma Proving Ground. That’s the photo you see on the far right, uh, September, obviously this month. So that missile very much more maneuverable. You’ll see

1:49:07 a little bit more about it. So, this is the entire team, Perseus Defense. Right now, we have eight full-time folks, and this team built a fully active guidance interceptor missile to shoot down drones. So, you can see we have all the folks listed on the bottom there. Maybe most important is our chief happiness officer, Ariel. Um, she keeps the keeps the guys, you know, in good shape at the office. Um, a lot of background in aerospace and defense. You know, we have a few folks from Aervironment. Um, Phillips, actually, Nick is somewhere in the back. He could tell you about our

1:49:37 flight computer later on. Um, and you know, we have kind of a lot of experience from aerospace and defense. You know, [clears throat] counter UAS, as you guys know, lacks affordability and range. We do have a lot of solutions that are out there. Very exquisite. We’re using coyotes. We’re using Patriots. We’re using A9X’s to shoot [clears throat] down 3,000 and $4,000 drones. Not great. Everybody knows this. And so what Perseus is working on is dropping that cost curve. So you can see the normal curve if you do a curve fit to all the existing solutions, it’s very

1:50:08 steep. Our curve fit is about 1/5 the cost per range of those. So you can see on the bottom we have the Mark III on there, the MarkV, which is the new one that we’re going to start developing. That’s a little bit of a larger missile. Think about like a three and a half foot missile to shoot down specifically group three UAS. And really the idea here is that we’re not only building very low cost, we’re doing it very quickly. So four months from new design to when we get it out there on the range. And then also um the speed of which we can build them. So scalability, it takes us about

1:50:38 a half hour when you have a pile of parts in front of you to a fully assembled missile that’s ready to go. So if you have five technicians, a regular 40-hour work week, you can do the math. That’s about 1500 interceptor missiles per month. So that brings us to some of the problems that we’re working to solve. Um obviously one of them here is cost. You kind of get the idea there. Um this is the more important one I would say. Though on the magazine depth, the entire rest of the United States builds about 1,000 interceptor missiles per month. Our near-term goal is to match the rest of the United States. Now,

1:51:09 obviously, our missiles are not as exquisite as a Rathon Coyote or as a Patriot missile, but they’re meant to do the role that those missiles are being used for right now, which is to shoot down very lowcost attraable drones. So, we’re doing it at a price point that’s more reasonable since we’re designing it specifically for that mission. And we think we’re pretty bullish on being able to do very quickly 1,000 interceptors per per month production capacity and longer term getting up to 10,000 interceptors per month. Um, so we are based down in Austin. That’s where we have a 18,000 foot manufacturing

1:51:39 facility. And I think I kind of mentioned some of the high levels here, but group one and two counter UAS is what this missile that’s right here is designed for. High rate production cost. Um, so sales price $5,000 per shot, which is a good order of magnitude cheaper than any other interceptor that’s out there right now. And, um, with that, you know, you could see some cool photos here on the slide. So, I’ll kind of kind of gloss over this, but how we can do this when no one else can, uh, we really started from, okay, what how are missiles built today and how could we build them differently and cheaper?

1:52:09 And so we threw out a lot of the you know aerospace has done to make it like 10 to the ninth redundant so that out of whatever that number of solutions that is or or flights that is you have only one mistake. We took the approach that if we can make it so cheap like 30 times cheaper than a coyote it’s okay if we need to shoot two of them to take out the target. And so that means that we use a little bit lower cost components. We use additive manufacturing which also enables us to move very quickly. But using some of these newer technologies and leveraging them allows us to build an interceptor missile that flies fully active waypoint guidance for $2,500 per

1:52:41 shot. So the left photo there, you could see the parts broken down and then 30 minutes later you have a full missile. Uh we do have our own in-house developed flight computer which is a pretty cool aspect too. So April 2026 Golden Shield at at Fort Hood. That was a really good test event for us. That’s the left and middle photos there. Really all the photos on this slide. And then just a couple weeks ago, we were doing some additional testing with JITA 401 with this new interceptor. Um, and then we’re currently working through contract the contracting process right now with the Army Applications Laboratory to do some tech maturation and get this thing from

1:53:12 where it is, which is like a TRL5 system up to a TRL6, TRL7. Uh, modular missile launchers. This is kind of one of the the last parts I’ll talk about. Um, we do have what you see there is an eight shot missile launcher. I wasn’t able to bring it with us today, but um eight of these go inside a Pelican case that’s about this tall. Two of us, any two of us, it’s really only about 100 pounds, could lift it up and walk with it. So, it’s very modular. Um we’ve had preliminary discussions with a lot of USV providers, UGV providers, even UAV providers to take our system, whether it’s this platform that holds

1:53:43 eight, a larger platform for maybe say a destroyer, um or even individual missiles and mount them under the wings of a UAV. So, there’s a lot of different options from that side. Um, and then, you know, this is kind of the last slide here. First startup to build an actively guided missile, I think. Um, I don’t know if anyone feel free to correct me on that one, but um, yeah. So, it’s working pretty well at this point. We get within 5 meters of the drone. So, that’s not bad. Um, at several hundred meters range, we can get within about 5 meters miss distance for a $2,500 missile. We need to obviously tighten

1:54:15 that a little bit um, if we’re going to put a warhead on it. What we’d really like to do and what we’re aiming for is to make this direct hit to kill so that we don’t even need a warhead and then we can use it domestically to protect nuclear facilities, airports, the capital, the US Mexico border and so on. Um, so that’s obviously very hard to do. There’s really one missile that’s endoatpheric inside the atmosphere that does direct hit to kill reliably. It’s Patriot. Um, we’re going to try to do that for $5,000 a shot. So that’s Perseus Defense. Uh, we’re in the back. Please stop by and check us out. Thank you. [applause]

1:55:12 Hi everyone, I’m Dante Vice, co-founder CEO of Albore Inc. Uh we make drone submarines. So basically we have this way. All right. So Albaore makes drone submarines. Uh we started the company in January of 2025. Uh looking at how America was starting to mass-produce unmanned aircraft and unmanned surface vessels at low cost. Um

1:55:42 the challenge with undersea assets has historically been uh they’re extremely expensive and they’re used for very brief intelligence gathering tasks. Um but uh given the challenges the US is facing uh we need undersea assets that can perform the same types of missions as unmanned aircraft and unmanned surface vessels. That is you know there these are traitable systems. They can be produced in large quantities and they can perform uh kinetic missions in order to deliver that same sort of cost asymmetric threat in the undersea environment. Okay. Uh not sure what has happened to

1:56:13 these slides, but yeah, basically uh the envisioned conops for these types of underwater vehicles are uh if you can make vehicles that uh are lowcost and can travel oceanwide distances, you’re able to do uh portions of the role that manned submarines in very large, very exquisite underwater vehicles, we’re talking about, you know, tens of millions of dollars in vehicles the size of garbage trucks, uh are currently tasked out with. So um that’s everything from intelligence gathering um to uh detection of mines which has been very

1:56:44 relevant recently uh to survey uh in commercial applications as well as you know long range precision strikes. >> Sorry. >> Okay. Thank you. >> There should be a picture of a a drone submarine here. There is one literally directly outside if you’re curious. Okay. Uh yeah, but basically the the challenge right now is underwater vehicles today,

1:57:15 uh they’re super limited in terms of their capabilities. You either have these really expensive systems that are really big. Uh we’re talking, you know, large percentages of the size of a man submarine, tens of millions of dollars. Uh and those are the ones that can actually travel entire oceans. uh or you have these these smaller vehicles that are basically, you know, incremental improvements to torpedoes where they’re fire and forget and they can maybe, you know, last a day or two in the water. But if you’re talking about uh a real sort of, you know, pure competition, uh you know, you’re going to want vehicles that can stay in the water for, you

1:57:46 know, weeks on end and and travel across oceans. And so um in terms of the sort of like high low mix that is envisioned uh you know we’re we’re trying to provide the uh you know lowcost medium capability system uh that can complement really large exquisite assets um and actually do meaningful jobs during uh you know pure competition. Um the way that we’re doing that is rather than focusing on uh battery powered uh systems right batteries just don’t store very much energy fundamentally um and if you’re pushing through water that is a

1:58:16 very energyintensive task. Uh we have basically brought back uh conventional World War II submarine design in an autonomous form factor uh where we are able to surface uh use onboard fuel and generators to recharge our batteries and then continue uh transit underwater. And that enables you to basically 10x uh the amount of distance that you’re traveling um while maintaining a really small envelope. Uh I don’t know why but these pictures are really not loading but uh this should be a picture of the ocean uh taken directly

1:58:47 from our drone submarines uh periscope. Um but uh the bluff is that uh we have deployed these in the ocean um many times uh and demonstrated them to a number of different stakeholders and uh as of today have um um contracts in in multiple countries uh and are working with people across uh three different continents uh including you know US Navy um and you know international uh coast guard and special operations customers uh and uh this is a a pressing issue this image I’m glad actually loaded uh

1:59:19 this was us um and the president of the United States a couple of months ago um at Senator McCormack’s defense tech conference. Um you know these issues are top of mind for everybody. Uh you know the president was talking with us about you know the straight of hormuz um but there’s a number of uh different uh geographies where these types of systems really matter. Um and you know people are starting to look at uh how unmanned systems are changing uh aircraft and you know there’s a lot of focus big programs like drone dominance for example that have been really successful at ramping up US production of unmanned aircraft uh

1:59:50 and you know we anticipate that in the coming years uh things that are similar will start to happen in the undersea environment as well and we think that there’s a sort of unique opportunity to actually get ahead of the curve here and start mass-producing these systems ahead of our adversaries to be uh ahead and uh rather than behind um and sort of cement the US’s historic uh undersea advantage. Uh and our goal for the next year is we plan to build 75 uh drone submarine systems which would make us one of the probably three largest manufacturers in

2:00:21 the world. Um this year we’re on track to build 10 and deliver them to uh initial customers. Yeah, that’s all I have. Thank you.

2:00:54 Hey baby. >> [music] >> Hey everyone, I’m Russ from Nine Mothers. Uh we’re in the X26 or P26 batch. Um we do affordable autonomous defense against suicide drones. So group one suicide drones. If you’re not familiar with these, these are the ones that are in Ukraine and uh Russia using against each other repeatedly, but also

2:01:26 in other places. We’re starting to see them. So, they’re small quadcopters. You’ll see some outside from other people. And um yeah, they’re super cheap. They’re $300, $500 each and up. And they’ll kill everything from vehicles to people. Um they can be entirely RF denied, as in fly with um fiber and not be tracked. Uh and then they carry payloads from very small to to to pretty large. They’ll do up to 100 miles an hour, at least in our

2:01:57 experience, and some of the newer ones will go a little faster. And the the problem we’ve got is that there’s no real cheap defense to this. There’s no real good way of stopping them. As you’ve seen with some of the other companies, there’s expensive solutions. And so when we started looking at this problem, we realized that we needed something cost-effective. Uh because this was the same kind of problem as IDs were in Iraq, Afghanistan, right? Uh how do you protect everything? How do you protect all the vehicles? And having a million-dollar system that works is cool. Having something that’s low six

2:02:29 figures is more viable. And so, one of my uh co-founders is driving a robot uh that has our system on top of it. So, it’s a basically a 20-in tall robot that we make for low six figures that can uh with a human in the loop autonomously defeat drones going fast. So publicly we only talk about going about 30 meters a second and that’s what we’ve consistently been doing for about a year now. We can do faster but uh come find me if you want to know more details on

2:03:00 that. Um but yeah the the point of this system is to have something that’s highly effective like high PK. And so this this version is a shotgun on a robot. Uh we can track and defeat drones day or night. Um and we’ve done different weather conditions as well. And we’ve we’re pretty early on the move, but we haven’t really announced what we’re doing on that. We’re on a couple of projects. Um, uh, I’ll talk about in a sec, but yeah, this whole thing uh, excluding the robot, like the tracked vehicle, is

2:03:30 sub50. We make a manportable version of it as well that, uh, some people are using. Um, and yeah, it, as I said, low six figure price and maybe we’ll get it lower than that at some point. Um, yeah, we’re working with a whole bunch of people. Um, uh, Air Force is the newest one. Done a bunch of stuff with DevCom, Jox, SOCOM, uh, Marine Corps War Fighting Lab, DIU and Army Applications Labs as well. Um, so this ranges from creators to actual sales and delivery of

2:04:01 things and uh like R&D efforts too. Uh, we don’t tend to go into detail, but if you want to ask more questions, let me know. Um, but we are delivering a bunch of units at the moment is the short version. Um but yeah, one of the things that we do is uh try and own the entire stack, right? We want to have ownership of everything from the ammunition that we use to the gun that we use to where we source all of our stuff and we write all of the software ourselves. This means that when things inevitably don’t go as

2:04:32 planned that we have full control over the outcome and we’re we’re the ones fully responsible for it. And so far that’s been really good. Right? We tend to, as I’m going to say on the next slide, underpromise on what we do and then overd deliver, which has also served us pretty well with the relationships we’ve built in the DoD. And for us, that uh that means being very confident in what we can do and then extending ourselves to the point of failure, right? As in we will happily do things that may work and then uh see

2:05:04 what happens and then improve it. But yeah, at the moment this system’s capable of shooting down uh direct tech drones at over 30 meters a second. Uh so 65 miles an hour and then yeah uh in multiple different weather conditions and time of day. Um and if you want to learn more, come find me afterwards. Um but yeah, this is just the starting point by the way. So we have this a manportable version. We have a vehicle mounted version publicly. Here we’re on a a project called Sant Hills 2 for 20th Engineering. Um which is an autonomous

2:05:36 autonomous breaching project with a company called Forterra. Um there are some more things coming out and if you’re going to a USA you will see us at a bunch of things. Um come and ask us questions if you want. Um but yeah we’re working on a couple of other things. So one of the things we’re doing is controlling existing weapon systems using this. Um, so our control system and our algorithms can be used to control other things which is kind of an exciting time. Um, and then yeah, there’s also different munitions being added to this for extended range and

2:06:09 some different capabilities. But yeah, um, uh, help from anybody in the audience like uh, very interested in more DoD contacts. As I said, we’re connected to a whole bunch of people, but DoD is very big and very uh, different. So happy to talk to anyone afterwards. We’re in the corner. Come and ask us any questions, but we’d love to connect. So thanks [applause]

2:07:14 Hello everyone. My name is Ryan. I’m co-founder and CEO of Metalware. and we’re a cyber security company based in the Bay Area. Um, this is about defending the embedded stack, our platforms that the Department of War runs, including munitions, UAVs, and more. Uh, to start, I want to remind everyone that cyber is the first war fighting domain. When you think about Venezuela and taking out equipment and power grids so our war fighters can go and perform their operations safely. On the contrary, you look at Salt Typhoon

2:07:45 and Volt Typhoon and you have adversaries prepositioning in our infrastructure to disrupt our troops from mobilizing ahead of conflict. As you all have seen, AI capabilities in cyber are progressing very quickly. Back in the spring, Enthropic announced Mythos and its capabilities. Uh it obviously found a ton of vulnerabilities uh in code bases that were very old and affecting systems and devices that run on millions uh and billions of devices.

2:08:15 Even what’s changed in the last few weeks that many may not know is that these capabilities are commoditizing. So, Enthropic published a report literally yesterday saying that mythos level capabilities are now available for download open- source for all of our adversaries to use with similar performance to mythos. Saying that adversaries are only a couple months behind the US capabilities on the frontier. In short, adversaries have frontier

2:08:46 level hacking tools currently within the department of war across air force, army, and others that we work with. they are lacking the tools. They’re using bench rigs and checklists rather than having uh being equipped with the best tools. The type of software that we focus on is is firmware and embedded system software. So this runs uh individual components and sensors and controllers on munitions on aircraft on UAVs. Uh there’s a number of different reasons why this is a more difficult technical

2:09:17 domain from lack of updates, legacy code bases, uh lack of access, uh and obviously the dependency on having hardware in the loop. To paint the picture a little bit clearer, when you think of a munition, a drone, an aircraft, you have hundreds of small essentially embedded computers dispersed throughout that are performing mission critical functions for that system. Today’s DOW requirements around cyber security particularly for platform IT

2:09:49 embedded systems uh are typically RMF and ATO right these are standards uh adapted towards hardware systems very different in technical domain from enterprise IT systems uh so there are no login pages for uh an aircraft or a munition a lot of these standards do not apply and are outdated when it comes to actual physical hardware security. Bringing it to Metalware, where did we start? We actually started in the offensive security community uh building

2:10:20 and developing offensive capabilities for the US intelligence community. This is where we learn the tip of the spear of hacking and access and understanding how to reverse engineer and find vulnerabilities and systems that we build capabilities off of. What we’re doing now is taking these tactics, techniques, and procedures, automating that en in order to enable scale for our defenders across the armed services to be able to increase cyber resiliency of our own platforms before the adversaries begin attacking our systems.

2:10:54 So in short, we’ve developed a lot of technology and algorithms and research in partnership with a number of institutions including the Air Force Research Lab. uh and we are automating this process that is traditionally very lengthy in terms of vulnerability research for these types of hardware platforms. Couple of differentiators since we get asked this a lot. Why not just use these frontier models uh on our own systems to secure them? Number of different reasons. Uh first and foremost being if if if adversaries are using these

2:11:25 frontier models against our systems, we want to have something better, right? Uh secondly, the AI models inherently to embedded systems and platforms. They were not trained on this type of code. It’s a different technical domain. That’s where our expertise differentiates us. Uh a number of other aspects such as when you’re testing a weapon systems, you’re going to be in an air gap high-side environment. How do you get access to AI in those environments? That’s something we help our customers do. Couple of screenshots of our platform. We don’t have uh build physical devices

2:11:55 so it’s not as flashy to have a demo up here. Um but we have an automated platform that’s very easy and reduces the uh labor and expertise required for our end users which are cyber engineers and developers to secure and understand their systems. Couple of the product screenshots here. So yeah, in short, we’re taking the tip of the spear, techniques and tactics, automating that uh in order to provide scale across the department of war, uh

2:12:25 provide our tools to the prime vendors who develop these to the verification and validation teams within the armed services to be able to to verify primes work and have additional confidence uh to cyber teams uh penetration testing teams uh and so on and so forth. Quick summary and background on our company. I mentioned we work with the US intelligence community. We’re beginning to work with the air force and air force research lab. Uh we work with defense primes as well as commercial critical infrastructure companies particularly

2:12:57 around developing operational technology and industrial control systems. Bit of our background. We come from the same type of places where we built these systems and broke them. Uh and that comes with our expertise. If this resonates uh whether you’re a customer uh or a vendor on OEM, feel free to email me at ryan@metalware.com. Thank you. [applause]

2:13:44 testing. Hello, uh, Under Secretary Hurst, thank you so much for joining us today, especially on the last day of the fiscal year. I’m sure it’s a super busy time. Um, and so on behalf of all the founders in the room, I’d love to chat with you about what it takes for them to build, deploy, and scale their tech across the department. And so I guess to kick things off from your role in comproller, what do you think is the single most important thing we can do to get technology into the hands of our war fighters faster? >> Yeah. So it starts with the biggest piece, right? The macroeconomic piece, which is the defense budget itself. Because that’s where we create markets

2:14:16 for startups to be involved. And if you look at the FY27 request, we’ve got $54 billion for drone dominance. We’ve got $60 billion in there for space, 24 billion for counter UAS. And I would describe these as markets in the defense community that are not locked up. There isn’t a single provider that dominates any of these sectors. And so new interests are going to have an incredible ability to step in there. The other thing that’s going to help is acquisition reform. We’re trying to move from a program manager structure to a portfolio acquisition executive. Why is

2:14:47 that important? Program manager has a very narrow task he has to do, a narrow capability has to field. His resources go against that. His planning goes against that. And so when a startup comes in with a better idea of how to do things, this guy’s already got a plan that he’s going to execute. Again, the port portfolio acquisition executive can move money between different capabilities and create opportunities for startups to get involved. And then the final thing that’s very helpful is just funding flexibility inside the department. Right now, as the comproller, I can move less than 1% of the defense budget between

2:15:17 appropriations. It’s it’s Goss plan. It’s I’m the Soviet Union. We’re doing a five-year plan. I do it two years before we actually get the budget, right? and then they let me move 1% around to meet need. And if you think about what’s happened in the last two years, whether you’re looking at the ISR reconnaissance strike complex fight that’s going on in the Middle East or in Ukraine, if you’re thinking about frontier AI models, we need to be able to move that money around faster because the technological change is so great. >> And I guess to double click there a little bit, what advice would you have for founders looking at these new funded

2:15:49 programs who want to turn those policy demands into actual funded opportunities for their business? I think you can break your opportunities into two different types. There are growth opportunities almost like growth stocks where there’s a trend inside the defense budget where the funding line just keeps getting bigger and bigger and bigger every year. And if the funding line is getting bigger, then those markets are fresh. There’s opportunity to get in there and get involved because we haven’t marked out those capabilities for specific vendors and create a program yet, a real program. But there’s also value investment. There’s over

2:16:19 50,000 suppliers that support the F-35 alone. A lot of these are small and medium-sized businesses and there are opportunities to come in there and get into that supply chain for that system and many other systems and do things more efficiently and do them better. And as we start to look at the supply chains for a lot of our major defense acquisition programs, we find a ton of inefficiencies there where we’re literally relying on like a twoerson three person operation to provide some critical parts. And so we’re trying to inject capital in those areas. But a lot of you, you could probably foresee opportunities to utilize technology in a

2:16:51 way that would get those parts to us cheaper in a more reliable manner. >> Awesome. And founders, uh, it’s quite easy sometimes to find users who love the product you build, but unfortunately a lot of the time they don’t always have access to budget or the authority to spend it. And so what advice would you have for founders who have those users and want to align them with a program office or a budget owner to actually get their technology from a demo to a funded capability? I think having the users like your product is incredibly necessary. Um, that’s a given, but it’s the program offices and the PAEEs who

2:17:22 really have the power to make you successful. And so when you develop a technology or product that’s liked by the users, that’s great, right? But what you really need to do is look at the program office of the PAE and find where their gaps are and figure out where your product or technology can fit in. You have to provide the solution to them. again, they’ve made their budget. They made it a year and a half ago. They’ve made a plan to execute that budget. And so, anything you’re bringing in is at times disrupted to them if they have an established program. So, you’ve got to

2:17:53 figure out how you can fit in. That means being a subcontractor. That means taking on a part of the program for them. That means looking at helping them shape requirements for a future capability. It’s not as simple as just coming in with a good idea. You’ve got to fit into the cycle. And I know that’s not a lot of you want to hear. and we’re working on changing that by doing acquisition reform, but it’s still a case in a lot of programs across the department. >> Do you have any examples of simple questions that a founder can maybe ask a PA to figure out the answers to to these questions? Because a lot of times this is the founder’s first time selling the

2:18:24 department and they might not know how it works internally. >> My my best recommendation is to hire somebody who knows how it works. It should be if you want to do business with the government, get somebody who understands business with the government. That will help you immensely, right? They’ll help you find those opportunities with program offices and PAES. But you’ll also find program offices and PAEEs who are just exceptional and want to court you. And so sometimes instead of just bringing your capability, I would suggest to you, you go shop for the program offices and program managers that are really interested in change and you align

2:18:55 yourself to that person and you start to help try to solve their problems with your team. If you find the right program office or program manager, they can do immense good for you. that can open up lots of opportunities across the department. They can introduce you to people even if you’re not a good fit for their program. So part of it is also personality shopping and find somebody who really wants to make change and isn’t fixated on just compliance. Do you have any advice for how to find the right program manager if you don’t have someone on the ground on your staff that might already have a relationship with that person?

2:19:26 I think if you start asking them, you know, what opportunities are available to you and how they respond to that series of questions, right, that will kind of tell you whether or not they’re inclined to try to bring on another company into their program. Um, but it’s a complicated cycle. You know, it’s really it’s it’s your job to understand enough about that program or that capability the service or the department is trying to develop and then figure out how you fit in. You don’t want to add work to people. You want to give them solutions. that’s going to make you more successful. And I know you got another

2:19:58 question about this, but people are often like, “How is the government going to fund me?” Don’t leave that up to the government. Give the program manager of the PA options on how they can fund you or get you involved. Take it on as your problem to solve. And if you do that, you make it more likely the program office, which is already working very hard to feel the capability to comply with the plan or the PAE, is more likely to try to bring you in. Because let’s think about it. If you have employees, right, and they just bring you problems and not solutions, how helpful is that to you as the owner of a startup or a small company? You want the employee to

2:20:30 come and bring you a solution to your problems. And that’s solving how you pay for it. That’s solving uh problems on your existing program or system. You don’t want people to say, you know, uh help me solve the problem with you. You want them to do it for you. And I guess to like double click on this a bit. So if we use an example of let’s say we have a drone interceptor startup that has an amazing demo and that customer wants to buy more. What needs to happen on the funding side to take that amazing demo to a production order instead of this endless cyclical loop of demos for that company? >> Yeah. So I I think what you’re trying to

2:21:01 do there and this is more of a a new space for the department lowcost interceptors is you’re trying to make sure that your technology adheres to the existing requirements or you’re able to shape the requirements as they come out. And that’s the most important thing to becoming part of a program of record. There’s two pathways in the department. You chase the existing requirement or you try to shape it. That’s one pathway and that’s probably more reliable. But it takes time, takes dedication, it takes form relationship with those offices. And then the other pathway is you create a great product the department doesn’t have a requirement

2:21:33 for and you try to come in through DIU or AppFit and then you hope you survive the transition to a service or to a department program office. Sometimes that happens. Sometimes your product or capability aligns with the needs of the moment and we suck it up, right? Uh if you’re making autonomous vessels that do demining right now, it’s probably a pretty good time to be in business and the department’s looking very heavily at what you’re up to. Same thing for lowcost interceptors. But that’s a bet on timing against some kind of geopolitical scenario or need in the

2:22:04 department. And a safer option is try to align with requirements or shape those requirements. You can do both. You should, but those are the two avenues to get in. >> Interesting. On the on the latter, how early would you say a founder needs to ask their initial government user or government customer to find transition pathways for what they’re building? They’re not going to be a give you answers straight away on what the transition pathway is most of the time unless they’re an exceptionally talented PA area program office. U you’re going to have to prove yourself to them over time and align with their solution

2:22:35 needs. Does that make sense? Yeah. Yeah. Um I guess for small hardware companies for example like back to the interceptor example, right? Um, a lot of times winning a contract and actually getting paid are vastly different things for that archetype of company, especially if they need to hire people or stand up a facility before actually being able to deliver on that contract. And so, what ways can the department maybe get flexible on payment structures for these hardware companies on technical milestones or creative ways that don’t that don’t require them raising outside capital or waiting for investment to

2:23:06 start delivering on that contract? >> Yeah. Um, anything that’s legal under the FAR can be negotiated with a contracting officer. And when you’re having that conversation with the office, right, the program office or PAE, make your needs clear to be successful. That most of these people, if they’re trying to engage in business with you at that level, they do want you to succeed and and they’re not particularly concerned if you get paid off at, you know, milestone X or milestone Y. So just have those up conversations about what you’re going to need to be successful. The alternative for them is you don’t have those conversations. You run short of capital

2:23:36 at critical time in the program’s development and then you don’t produce for the government. So, it’s much better to talk about what kind of financing needs you might need. Also, they can link you up with entities like the Office of Strategic Capital, right, which can give you lowinterest loans to help your capitalization. Uh, we can use IBBAS dollars or DPA dollars if it’s an important effort for the department. There’s $13 billion in the FY27 request for industrial-based support. We need to get those dollars out the door to companies interested in working with the government over the long term. So, ask

2:24:07 your program managers about those. So if we can’t give you progress payments, maybe we can give you financing that’s low interest that allows you to grow your business. >> Are there any specific ways that your office can help a company if their technology is ready and maybe it’s commercially proven, but the funding pathway for transition might not exist or be active in the government? >> From a function perspective, the biggest thing that Comproller can do for you is we get our service inputs for the budget and then we shape them with the deputy secretary and cape and the joint staff uh in a series of meetings we call DMAGs. And so that’s where we get the

2:24:38 opportunity to kind of shape priorities for the budget and make sure that there are large enough markets that make it worthwhile to invest. Uh I’ll give you an example. If you look at the medium unmanned surface vessel that was awarded yesterday, right? That came out of reconciliation. We’ve created a market there for companies to start to build those autonomous vehicles for the department. So that’s the kind of things comproller can do. We make a market by putting space in the budget for new technologies or things that startups are interested in. On a personal note, if there’s a technology that you think is

2:25:09 really important, you’re having trouble getting through to a program office, we can be an advocate for you. That’s a that’s a smaller thing we do compared to shaping, you know, the the fence budget, but it’s still an avenue. >> And I think all of us in this room want our hardware companies to increase their manufacturing capacity. But I’ve heard it’s been pretty hard for founders to make the decision of opening a production line if they don’t know how many units the government might actually buy in a short period of time. And so what can the department maybe do to put out more dependable demand signals for founders to make these decisions prior to actually manufacturing the

2:25:40 capabilities? >> One big effort right now that’s in the news that you’ve seen is the idea of using more multi-year procurement contracts. um why right so we give a steady demand signal to the big prime making those things obviously but what’s probably more important and we don’t talk about enough is the steady demand signal we give to the supply base below that because if you’re Lockheed Martin if you’re north of Grumman you’re trying to hold on to your cash just the way everybody else is right and so you’re paying your suppliers at time of need

2:26:10 and you’re not giving them enough awareness of what the long-term orders are going to be so by giving a multi-year order we can cascade that demand signal down all the way to the supply chain, the very base of it, tier four or five suppliers and help them get more stability so they can seek out the financing they need to grow. Uh they can seek out investment. Um that’s one thing that we’re doing. The other thing I’d say to companies is if you’re thinking about doing something with manufacturing, see if there’s a way you can make your product line dual use. The government, we have things like

2:26:42 government shutdowns. We don’t always control the timing of when our budget comes out. And if you can make your product dual use or some of the capabilities of your production line dual use, you can insulate yourself from a lot of that flex, right? Um it’s it’s an attractive thing I think for investors. It’s attractive thing for the government because we can see you produce other things and you can showcase your manufacturing prowess. Uh so always think about that if there’s a dualuse application of your product. A lot of you are patriots. You want to be involved in national security. That’s awesome. But I’d say to you, we don’t

2:27:13 have a defense industrial based problem. We have an industrial-based problem and if we’re going to fix the defense part of it, we obviously have to have suppliers in all areas. >> I guess outside of being dual use, is there any other evidence that may give you confidence that a younger company might be ready for a larger production order? >> I think the the best thing is to show that you can make stuff. Um, a lot of startups right now, you know, they they’re sometimes they’re paper and that’s kind of a risky bet for the government. Not that we shouldn’t take it, right? Because sometimes our

2:27:44 alternative is a company that does manufacture things but doesn’t produce on the timetable we want or the capabilities we need or at the cost schedule we want. So again my best advice to you is show that you can manufacture things and if you are not manufacturing things hire people who have experience manufacturing things so you can show the government that you’re bringing on the talent necessary to scale your product and capability if they decide to buy it. And for the founders in this room that have technology that works, it’s commercially proven, but they might not have any relationships in the Pentagon,

2:28:14 what would you recommend to them to become a credible supplier to the department in the next 90 days? First thing is personal relationships. Reach out to the program offices you think have overlap with you. Talk to people in the building, of course, at Comproller, R& AS, and start to socialize things. Uh you want to attack on all fronts, right? But you’re developing a reputation with the department and then you’re trying to come up with solutions that are going to solve problems for program managers or PAE. So develop your reputation, come up with a solution that solves their

2:28:45 problems. Be creative about how you fit into an existing program or requirement. And I think that’s a good solution for success. >> Awesome. You comfortable opening up to the floor for a few questions? >> Yeah, let’s do it. >> Cool. >> This is an incredible conversation. So much alpha right here. Uh, raise your hand if you have a question. >> You’re making it run, man. [laughter] >> Hey, thank you. Marcus Weisscover with the Wall Street Journal. Um, Secretary

2:29:15 Hurst, what’s the status of omnibus reprogramming? Did you guys do one this year? Uh, can you give us a sense of the size and does it being happy physical New Year’s Eve? Um, does it have to be approved by today by Congress or is that able to be approved, you know, down the road at some point? >> Yeah. Uh, this is probably boring for most of you. We submitted omnibus reprogramming in July, just like typical timing. Uh, we’ve been going back and forth Congress on having various elements of that approved. Over time, uh, we’re in good shape for the end of the fiscal year. And when we finish with

2:29:46 the reprogramming, we post it to our website online so everybody can see what trades we made. How big is it? I don’t remember exactly how big the July one is because there’s one that falls August. Uh we’ve made good use of our general transfer authority this year. >> Yes, ma’am. Jane. >> Hi, Jane. >> Great to see you both. Uh Jane Lee, I’m the chair of the Senate Defense Modernization Caucus Alliance. Um question regarding your new task

2:30:19 force on waste, fraud, and abuse. I think one element having been on Senate Appropriations Committee is building trust with them if you’re looking for funding flexibility, it’s uh making sure that the department who’s hasn’t had a clean audit. I love that the progress that you guys are making. Um but can you tell us a little bit more about that task force? >> The task force one more time >> on waste, fraud, and abuse. It was just announced uh it was uh Secretary uh Fineberg had released a fast. >> Yeah. Yeah. Um, we learned a lot of

2:30:49 great lessons in the first year of the administration about how to find waste and every government agency, every business has it. And so we stood up the fast team which is a lot of experts who either have a finance background or worked in the department before and I’ve got people applying to this organization across the entire department and honestly across the entire US government. And the idea is right now we pay for things more than we should. Sometimes we pay the same organization four or five times for the same data services in the army will pay them, the air force will pay them, etc. And we’re

2:31:19 having them go through and find efficiencies like that. We’re also finding where we have duplicative efforts, which is probably not shocking you this exists inside the department. And we’ve been successful, I think, in finding a lot of those in the first year and a half of the administration, but there’s more to do. And so this is a dedicated team and it sits under compiller because of our budget authority. So we can take what they find and immediately implement that in our budget request. Um, do you want me to talk about transparency in Congress too? Okay. One of our big struggles to get the funding flexibility we need to support new technologies is to making

2:31:50 sure we have the trust of Congress. And so we are building a tool which is going to show obligations and expenditures in near real time that we’re going to look at sharing with Congress so they can see exactly what kind of trades we’re making through the execution year with our budget. We don’t have this now. I manage everything inside the department on a PDF. It’s insane, right? Um, but we’re slapping the table on data standardization for that. We have a visualization tool that’s in place in multiple services and we’re doing it at the OSD level and we’re going to use

2:32:20 that to help build with Congress, put ourselves in a glass house and then make the request again for more funding flexibility so we can support new technologies at a faster pace. In the case of a CR, do you see more or less funding flexibility on the horizon? >> CRS hurt us. Um, when the CR happens, we all get really stingy with how we

2:32:51 execute funds, right? We only worry about our core responsibilities and so our ability to start new programs obviously is hindered very much uh and to make large bets. Uh, so the longer the CR period goes, the more it slows technological adaptation inside the department. Cool. >> Thanks very much for having me. I really appreciate it. Um, and I’ll hang out for a few minutes. Maybe wants to ask me questions. [applause] [music]

2:33:24 [music] >> [music]

2:34:06 >> Get a low resolution. All

2:34:40 right. Good morning, ladies and gentlemen. My name is Grant Demeury. I’m the CEO of One Brief. Last time I talked to a Y cominator audience was at my own demo day in summer 21. Um that was actually five years ago. So it means a lot to be back here uh talking to a YC audience for the for the first time since. Um I want to organize this talk into two parts. Uh part one I want to tell you and actually demonstrate what we do for the department and and why. I’ve heard it’s a really bad idea to do

2:35:12 live demos on stage, but uh I’m feeling adventurous today. And then in part two, um five years since demo day is is enough to accumulate quite a few observations about defense software, and I hope I can condense them into a couple of minutes and and give you something really useful. So today, what I’m ultimately going to show you is a live demo of how AI can create military plans from the scale of how do we defend Taiwan down to what do we do tomorrow in Ukraine and can

2:35:42 validate those in a live planet scale physics-based war game and ultimately generate the kind of afteraction review that leads you to figure out to to learn uh how to make that plan better. do it again and again until you’ve achieved true superhuman decisions. Um, but before we get to that, I want to start at the beginning. I want to tell you a bit about how we got here. So, pre- one brief, I was a officer in the army. Um, during the Ebola epidemic, I was actually a second lieutenant in the

2:36:12 101st and uh was was very fortunate to be able to work out of the Liberian Ministry of Health on the Ebola data. And thankfully I was I was actually attached to the J5 in that joint task force. Uh the J5 um for those familiar with uh military staffs the section responsible for operational planning and that meant that really early on I had a chance to see some largecale operational plans. Um I ended up reaching two conclusions out there and the first conclusion was that planning drives

2:36:43 almost every major military decision. Um so fixing planning isn’t really about fixing planning. It’s about fixing decisions. Um, and the second conclusion was that the status quo is nowhere near the frontier of what’s possible. I’ll have some front reflections later on where the frontier really is, but I think wildly superhuman uh would would be of of really world historical uh consequences. Now, a few years later, I had some similar experiences as part of a couple task forces this time out of

2:37:14 Baghdad for the counter ISIS mission. I concluded okay this is a a very broad problem that matters across uh substantially all military headquarters worldwide. Um so uh got out of the army to try and fix it and had quite a few adventures with uh defense go to market and uh and defense products. Um so I want to start um before we get to the full AI plans with core one brief that is what we built first. Um, that’s what you’re looking at right now. And the purpose of core one brief is to serve

2:37:46 every military headquarters from roughly receipt of mission until roughly the issuance of their order. Um, I want to give you a couple of demonstrations here. Right now, you’re looking at a sync matrix or for the non-military folks, it’s it’s basically a militarized Gant chart. Who does what and and and when. Um, when I make a change on the sync matrix, um, you’re going to see that everything that depends on it changes too. And it actually automatically updates my slides. Let me go to those now. It updates my slides. It updates my operations order. It

2:38:17 updates everything. Um, same for the maps. Actually, there’s uh quite a lot of map in this map. Let me go uh to it now. And when I make a change in the map, it actually automatically updates the slide. Same idea for the command relationships, updating the order. Same idea for everything across the application. And the point here is how fast can your headquarters really move. Um that brings me to a really important trend in defense software that I think a

2:38:47 lot of the department should be very proud of. Um a couple years ago it was actually really hard to integrate software applications across the department. Um you would get a lot a lot of stove pipes. That’s not fixed but it’s a hell of a lot better. Um, we integrate with a pretty wide variety of applications from very new startups all the way to governmentowned uh things like what NGA serves. What you’re looking at right now is uh integration with Maven Smart System. Um, and what I want to show you is that integrations can actually be really really easy. One

2:39:19 second. Apparently less easy. That’s okay. That’s why uh live demos are lovely. Hey, here we are. Um, and what I want to show you is that, uh, integrations can be fast and very easy. Simple drag and drop. If I want to change this mapping from, uh, birectional to one way, I can do so. And and the real point here is

2:39:52 that your operational user um that is someone sitting in a military staff can just do this sort of thing without an engineer present. Um and the intent is is that what the department ends up with is instead of a number of disconnected systems, it’s basically all part of of one system. Data flows cleanly everywhere. Um I want to show you another part of of the speed up. Um, this is the sort of thing that if you served on any military headquarters, uh, you know how much time is spent on

2:40:24 routing and approvals. Uh, don’t worry, I’ll get to AI generated plans, but to be honest, I’ve spent a lot of my time on this. Um, what we’re looking at right now is a warning order that came out of this plan. And when I need to um route this warning order for approval across say a a joint task force headquarters, there’s actually a pretty large number of people that need to review ultimately generally the the J3, the chief of staff, the commander will need to to approve it after staffing and then route it to a multitude of units. Um there’s

2:40:54 enough officers who do this like literally their entire day. Um so I’m going to make a comment right now. I’m going to categorize that as a critical comment which anyone who’s who’s done uh the staffing I think is going to be quite familiar with and what you’re going to see is that the uh comment resolution matrix is actually just created for you. Um and what you’re ultimately going to see is each of these

2:41:25 things routed distributed in app exported to wherever it needs to go. Uh pretty pretty enormous time savings there. Now I said superhuman level staffs in in the beginning and merely making the staffs faster isn’t uh isn’t sufficient to get there. Um and that’s why I want to show you how we get to full AI generated. If you give me one second, here we are.

2:41:55 Everything in here uh everything that a human can do is fully operable with AI. Um meaning if you want to optimize your sync matrix, place things on a map, draw your concept sketch, build your slides, optimize your order, um everything. And I think a lot of folks in software have observed that user interfaces are converging to interact directly with the LLM. Um, in a second you’re going to see some uh analysis for our approval on the warning order one that uh this AI has uh

2:42:28 has created for us. And uh as we wait uh these cards are going to start to populate and we’re going to see an audit trail for our approval. Um, I want to move into something that I see as even more important, which is in order to get to truly superhuman, um, like we’ve seen in software engineering, what made that possible with AI was an error signal where you could tell is this code good or not through a fast feedback loop.

2:43:00 That’s generally missing in military operations. And our view is that the only way to get there is through simulation. Um, so I want to show you a full planet scale uh physics-based sim. Um, and I want to walk you through why using this uh simulation actually ends up with some pretty uh pretty wild outcomes. Um, so what you’re going to look at uh as soon as this uh briefly loads um is is a a full planet modeling all the way

2:43:33 down to the individual terrain. uh modeling all the way down to the individual vehicle uh modeling every uh missile trajectory, every radar cross-section, every terrain effect. Um and I’m just going to briefly conduct a air strike. And what you’re going to end up with

2:44:03 here, and let me start this in. This is running at uh 300x real time. Um what you’re going to end up here is a pretty detailed combat simulation. Um I care a lot about this one in part because it’s scalable to many hundreds of thousands of entities. Um and in part because it’s it’s quite extensible where most military operations that I’ve I’ve seen end up with some variable that that no simulation would have thought of in advance. I think a good recent example is not a huge number of simulations cover the uh commercial uh shipping

2:44:34 vessel behavior in the straight of Hormuz and obviously that ends up having great military relevance. Um so the the point is to make it very very easy to add things so that anything you need is is simulated. Um I want to put that all together now. Now, I want to put that all together and show you that uh in the beginning, we took a look at AI generate a plan, set up the war game, let the AIs play the AI, get an afteraction review, uh shown here, um and then use the

2:45:05 afteraction review to improve everything about your plan. Um I want to convey for a moment, um what we’re uh what we’re we’re really envisioning here. What I imagine is is a world where we meaning US and our allies uh find every possible military brilliance every time we should do the Inchan landings instead of landing at Busan. Uh we we discover that right away. Um likewise looking for a world where planning D-Day takes you

2:45:37 know seconds to to minutes. Uh where every military option that goes to the president is the perfect risk adjusted way to achieve our goals. Um we anticipate so well every logistics and people move happens before you need it and even policy changes and international coordinations and procurements happen before we need them. So when we talk about decision dominance that’s what we mean. Um that’s it for uh for one brief. I want to take the last two minutes to offer some very brief uh

2:46:09 defense software reflections that I I hope will be useful to the the folks involved in in buying or or making defense software. Um I think the the first observation is the level of progress in the last year is very very different from what came before. Um there’s been a pretty insane discontinuity here. Um I I think a lot of people in government should be very proud of that. There there’s a number of causes. Part of that is that AI got good, but part of that is that some reforms really worked. Um, I’ve watched defense software startups go from 0 to

2:46:40 12 million of ARR in one year. Um, and I I think 5 years ago, no one would have seen that coming. I actually acquired a company with a similar profile to that. Um, and I had a good friend go from 1 to 50 of ARR uh in in in this one year this year. Um meanwhile on on the government side uh one piece of evidence that reforms are are working is the speed of delivery is much much higher and the program that I think uh everyone involved in software really needs to to get familiar with is the army’s next generation command and control or NGC2.

2:47:12 Um so these guys did a lot a lot right. Um fast and easy to swap out vendors as seven major exercises quite detailed ones in in one year. Um an operational commander in charge able to pick and choose the software of of his choosing. Um very efficient and supportive contracting command uh forced integrations on site which is I think a inherent inherently governmental role is to compel thou shalt integrate. um and and very strict penalties for non-performance in the form of being

2:47:42 quite rapidly booted from the program. Um I I spent about 200 nights on the road last year. Um saw the software of substantially all the services, quite a lot of foreign militaries. I don’t think there’s any example to learn from better than than that program. Um so if if you’re in Congress or or senior leader, I I think that’s really something that needs a pretty detailed a AR. Um okay, second observation is that AI changed the game really really dramatically. Um transforming our own company into a Gent was a huge bear. Um but it also works.

2:48:15 Um we measured about a 45x increase in the rate of feature delivery to customers uh across our best teams from uh January of this year to present. And 45x is just like insane. I thought AI was going to get good. I I would not have said 45x would have been practically achieved. Um, and I I think the the lesson from that is isn’t just the models are very good, but each time you unblock something, you you earn the right to find out what the next blocker

2:48:47 is. Um, so a really good example of that is as soon as you finally gotten good enough at delivering highquality features that reliably work, um, they’re now delivered so fast that the DevSec Ops pipeline basically breaks because you’re shipping, uh, PRs that are, you know, 180,000 lines of code. Um, and you solve that and I I expect that there’s a next problem and a next problem, but the results are in fact quite good. I I think folks in government should should think very very carefully about the pace

2:49:18 of improvement in the frontier models. If it’s true that the task length that a frontier model can can do at least in software is growing super exponentially and then the doubling time is actually getting less than than 4 months then our expectations of software over the next uh 6 months should just be stratospheric. Um the the last point what I want to conclude on is that it’s it’s really a lot better to have a reasonably good guess on a very important question than a precise answer

2:49:49 to a question that’s just not that important. Um and I think the central question of defense software is how much combat power do we really get from various levels of human to superhuman uh command and control. Um, so basically is it about as valuable as a 10% increase in your force structure or is the frontier of what’s possible about as valuable as a 10x increase in force structure? Um, I think it’s closer to a 10x. Um, I I pretty carefully maintain

2:50:20 my own estimate of this and I also think it’s very important the government independently evaluate that question and that each person in defense software take a lot of time to have an inside view. U I I think the fundamental question of uh of resource allocation. All right, ladies and gentlemen, I really appreciate this. It’s good to be back with YC. Thank you. [applause] [music]

2:50:56 [music] Hello. Okay, good morning everybody. Uh, nice to meet you all. I am Matia. I’m one of the co-founders and CEO of Hop Arrow. Uh we’re building rockets that

2:51:27 can deliver cargo to contested environments. Uh and so today I want to talk about rocket cargo for contested logistics. Just introduce myself. Um I’m an engineer by by background. Previously founded a satellite propulsion startup. Um led that to orbit on the third ever Raichure mission with SpaceX back in 2022. Um I then scaled the prop the production of uh large scale loitering munitions for the Ukrainians. Um my co-founder served in the Marine Corps. uh he was based out in Okinawa working on the MV22 Ospreys, worked at SpaceX

2:51:57 launching over 20 uh Falcon 9 and Falcon Heavy rockets uh during his time and then he also led the NASA Eclipse lunar lander program at Mass and Space. So just to give a quick quick overview of our company, uh we were founded in 2024. Um just this morning we announced that we’ve raised over 11 million in in VC funding. So we’re really happy about that. Uh currently delivering on a Thank you. um currently delivering on a uh direct to phase 2 cber with AFL that we’ve been really happy with. Um also building out that relationship uh even deeper. We’re based out in in uh Orange,

2:52:29 California. Um but we do a lot of our work at the Oklahoma Spaceport um in in Oklahoma. Uh currently a team of seven and just wrapped up the summer 26 batch. So I think what we can all agree upon is that warfare has effectively already become um autonomous across all domains. Um, air is kind of the the most commonly talked about one when it comes to either drones with RPG munitions strapped to them or cruise missiles. I had the privilege of of working on that, the the cruise missile in the middle image, increasingly ballistic missiles, the

2:52:59 same thing for sea, same thing for ground, same thing for for space. And so when you start thinking about um the Taiwan scenario, which is kind of the the most commonly discussed situation right now, the problem that we face is how do you get to um one of the most contested environments in the world, especially when the adversary is about four times closer to to you than you are? You are being tracked by space domain awareness. Um there’s a potentially a blockade. You have a you know barrage of um autonomous systems

2:53:29 that are being thrown from from the adversarial side. And so these are a lot of problems that we are thinking about how do we overcome not just in in Taiwan but in any sort of pure pure conflict that may come up. And so that’s why we’re building the world’s first uh cargo delivery rocket. Uh we’re calling it Rook. Um Rook is able to launch from a standard 40ft shipping container. So what that means is that we can launch it from pretty much anywhere on on Earth in minutes. Um we reach a max speed of Mach 7. So that’s hypersonic speed. Um and we have a range of about 450 mi. we can get

2:53:59 to um to Taiwan from Okinawa in about 15 minutes using using our technology. And so that’s the the main thing is like the time to launch is about 15 minutes. So we want to be really really quick in letting uh war fighters from the moment that an order is put out to being able to launch. So this is just a general simplified concept of operations. Um we reenter as I said at hypersonic speed. Um we land on unprepared environments. So, a lot of my co-founders work was landing in um obviously with with the NASA eclipse lunar lander being able to land on dust, on gravel, on unprepared

2:54:30 surfaces. And so, we’ve been super busy during the the YC batch. Um probably the coolest thing is that we’ve got our um engine right here. Uh we we built the software that builds this engine. So, um we we built a tool called Hop OS and I’ll talk about more of that later on. Then we put that engine that we hotfired into our vehicle. So we did a halfscale prototype of Rook, hotfired the did a static fire of the entire system and then we demonstrated the deployment of drones out of our payload bay. So we can deploy uh drones, interceptors, um

2:55:01 electronic warfare payloads into a contested environment. The way we like to phrase this is that we want to deliver a fully autonomous 21st century D-Day using rocket cargo. And so our secret sauce is that we use um our intelligence layer called hopos to not just build our rocket engines but do a lot of things that were previously very manual and laborious um internally and we can do that in minutes. So not going to go through all of these tools but the the coolest one that we really want to highlight um is hopos. Hopos allows a user to put in highle inputs

2:55:32 for what an engine needs to look like and they get out a full 3D geometry that’s ready to manufacture. So that’s that engine right there. Um it worked the first time. This is the coolest thing about Hop OS is that it builds verifiable um engines. So we we did this at the Oklahoma Spaceport. Uh these are some more cool shots from up close when it was integrated into our vehicle. And so just to wrap up, what what we need help with uh when it comes to, you know, getting support from DC in order to be able to succeed. Um to those that are familiar with the TACI and Stratfi programs with the Department of the Air

2:56:03 Force, that does require matching funding. And so we’ve just submitted our application for Attackfi. Um, but this is still a m multi-month process. And so we’re ready to move fast. The department is ready to move fast. Um, we just want to make sure that we’re continuing to engage with stakeholders and make them aware of the fact that they are slightly becoming the bottleneck in in terms of getting this technology out to the war fighter. So with that, I’d like to thank you all for your attention. Um, feel free to to grab me at any point and happy to talk rockets. [applause]

2:56:43 Ladies and gentlemen, we will now take a 20-minut break for lunch. Please visit the exhibitors. Thank you.

3:21:55 Ladies and gentlemen, please take your seats. Our program is about to resume. Thank you.

3:22:49 All right, folks. Welcome back to the program. Uh our next um fireside is one I’m very excited about. Uh I I first want to kind of take a moment of uh personal privilege and just give a shout out to Eric Albborg who’s the author of Build for Defense. Eric uh left DIU uh a handful of months ago and has been working incredibly closely with all the YC defense companies uh over the last uh four four months or so and has really

3:23:21 been an incredible resource. And so we’re really excited to uh talk a bit about his upcoming book which we’ve provided advanced uh copies for for all attendees. And I believe Eric’s going to be uh doing some signings at the uh the uh afterparty this evening. Uh Eric is the author of Build for Defense, a practical guide for helping commercial companies enter the defense market. He previously served as a commercial executive at the Defense Innovation Unit, connecting emergent technologies with uh defense and intelligence needs.

3:23:52 He now advises founders, investors, and companies on navigating that market. And Olivia Morat works on pont partnerships at Y cominator helping support companies after the batch. She’s a graduated of Sworthmore College where she studied art history and economics. Please welcome Eric and my colleague Olivia. [music]

3:24:24 Right, >> let’s set this up here. >> Thank you. Yes. >> Great. Thanks for joining me today, Eric. >> Oh, thanks for having me. pleasure to work with you the last four months or so and thank you all for being here. Um so the book is called Build for Defense. So who is this book for and what does a founder walk away knowing? >> Yeah. Um well I I guess the book is really for the founders who think that uh you know winning their first contract is going to

3:24:56 be the hardest thing. Um it’s actually not. I mean, uh, the first contract, of course, is an important milestone, but getting to that second or third or fourth contract, um, it can be quite a bit harder. And it really kind of depends on the archetype uh, of the company that you’re building. If you’re building sort of a pure play defense company that’s kinetic, that’s building a missile or a drone, um, your path to market is going to be the Department of War. It’s pretty

3:25:27 pretty obvious market. If you’re dual use um and you’re a startup, it can be really hard to grow two products simultaneously um because you’re you’re trying to get commercial traction and trying to get DoD traction and and those are not the same thing. Um commercial traction looks quite a bit different. Um and then you might just be a commercial first company. You might be something like Microsoft and >> sell Microsoft Office or products like that. And if you’re that kind of company, you probably want to think

3:25:57 about subcontracting. You might not need a prime contract with the DO. Um, and so really the the mistakes or the different decisions you have to make as a founder are really going to be different depending upon what sort of archetype you’re you’re going to pick. So yeah, this this book is about sort of the gap between those three those three archetypes. Um. >> Yeah. >> Yeah. >> Nice. So you spent the last four months like Luther said immersed in the founders problems who are building for defense. So what have those founders changed your mind about from the DIU?

3:26:30 >> Yeah. Um incredible group of founders. I I was just really struck by how much risk they were willing to take. Um and not just risk in terms of a startup and putting yourself out there. I mean, some of these companies have put themselves on the front lines in Ukraine or in in sort of wartime conditions to test their products. Um, but in addition to that, I mean, if you, you know, look outside at the YC companies, you see just the incredible breadth of technology, uh,

3:27:02 things that’ll help maybe an intel analyst, uh, things around wargaming, um, of course, UAS, group one to five even, uh, meaning smaller drones to larger ones. Um and so you know I wrote the book about the this gap sort of between the different archetypes but um really what I realized is that the book is for uh helping startups actually optimize for speed because um they don’t have a runway beyond maybe 18 months at

3:27:35 any given time. And so if they’re not converting on whatever their business model is um they’re they’re not going to make it to that next milestone. And what surprised you most about what’s stopping them? >> Sorry, say that again. >> What surprised you most about what’s stopping these founders? >> Yeah. Um, you know, I I think that access is getting better. Um, we heard from under secretary Michael and Under Secretary Hurst um about changes that are happening to try to make it easier to access the department. um at DIU they

3:28:07 they’ve always sort of leaned forward into the startups into the smaller companies to help give them a way to access the department. So I don’t think it’s really that so much is it understanding that you know the the market that they might be pursuing is not probably what they think in the beginning and so it’s it’s helping them sort of configure the way they think about their go to market the way they think about their business decisions for defense versus sort of the way that most startups are uh taught which is um sort

3:28:38 of the the framework of a commercial go to market. So, want to talk about the book. Um, so it opens with two startup founders and an established contractor and they’re pitching at the same table. And so, what did you see in that room that a technically strong founder would be least likely to understand on a first visit? >> Yeah. Um, so my job at DIU was diligence. And what that meant is I would sit in rooms where decisions were made on what what contracts would be

3:29:08 awarded to what companies. And it would generally be something sort of like this. There would be a table. There would be 10 20 government uh folks around that table, active duty program managers, acquisitions folks, um federal contractors. And you’d have a startup come in and they would um you know look rather tired most times but uh be casual and do an exceptional job at pitching

3:29:38 um to the people around the table. But pitching sort of optimized for the problem to solve and a very technically focused sort of engineering focused pitch. and you’d get to the question and answers and um you know people sort of didn’t know how to engage. They sort of they they sort of it went into a black hole if you will. Um and then you know after them the defense prime would come in they were all in suits and ties looking very well-rested. Um and uh they would shake hands with

3:30:11 people in the room. In most cases, they they knew people in the room already from sometimes decades long relationships and um you know uh they’d pitch uh and they would uh get great questions and answers um and you know sort of that was this pattern that I would see over and over again and it started to make me think like what is the startup getting wrong? like what how are they not like the pitch looked great to me like why isn’t it and I think it’s because there are many people in the room that it takes to to get to yes it’s

3:30:43 not just the engineer the engineer will be a champion in many cases because they understand how your technology is solving the problem um but there’s a program uh manager in that room and they’re thinking about does this fit the mission in many cases startups don’t even discuss a mission when they’re pitching which is a huge mistake um there’s an acquisition official the acquisition official official is taught to avoid risk at all costs. And so a two person startup with $500,000 looks like a lot of risk maybe to an acquisition official. Um and then maybe there’s, you

3:31:16 know, um other members on active duty or other federal contractors who maybe just don’t understand the Silicon Valley language. And so somebody sort of has to interpret that uh for them in order for them to understand. >> Yeah. So, a little bit more about acquisition and acquisition reform because probably a lot of people in here are skeptical about acquisition reform. So, what’s changed? >> Yeah. Um, well, I it was really great to hear from under secretary Michael and under secretary Hurst. >> Um, I guess two things in acquisition have changed. Probably one uh hasn’t.

3:31:49 The first is um you know there are really good faith efforts at reorganizing the department uh to make it more responsive to founders um and you know we heard that on the stage here before uh when I was at DIU that was that was the case and I think um that’ll have long-term consequences beneficial con consequences for our national security. Um the second thing is there’s more money. Um uh there’s the budget has just been continuing to to go up.

3:32:20 Reconciliation is sort of on top of the base adding to that. And so um as under secretary hir talked about creating markets for startups that that’s you know going to put a lot more wind at their backs. The one thing that hasn’t changed um and unfortunately it won’t change probably anytime soon is the budget. >> Yeah. The budget is an annual process. It’s sort of arcane. It’s sort of mysterious to a lot of founders and um

3:32:50 you know it’s it happens once a year and we’re at the end of the fiscal year this today. Uh fiscal year 27 starts uh tomorrow on October 1st. We’ll likely be in a continuing resolution which does not favor startups quite frankly. Uh that favors the incumbents. Um, and so, you know, I think that the, um, the budget, I mean, if if we really wanted to have a budget that was responsive to technology changes, we wouldn’t do it on a year time frame. It’d be far more u

3:33:22 frequent, maybe quarterly or semiannually to actually adapt and and create um, more opportunities for startups. >> Great. So next I want to address an elephant in the room which is is defense tech a bubble? >> Yeah. Um cash is moves quickly. Cash moves very quickly. Uh defense contracts do not move quickly and founders are sort of caught somewhere in the middle.

3:33:53 >> Yeah. Um, and I think that the challenge for a founder is to raise money to the next milestone. And I think when we kind of conceptualize product market fit in the commercial sector, we think about that really in terms of customers and growth and how fast you can grow. And in the Department of War, it’s a little bit different. you you really need to fund raise to your first milestone, which is probably a demo um or maybe a commercial solutions

3:34:24 opening, a a prototype project where you’re actually trying to derisk that and optimize your product for the Department of War. Um and then you’re trying to now raise again for the next milestone, which is probably, you know, um something that comes after a prototype contract. Some internally people talk about it as a transition. it’s a program of record or a longer term uh and that’s probably where you have product market fit. And then from there, if you can get three, four or

3:34:54 five different uh Department of War uh contracts in production, uh multi-year agreements, you’re probably on your way to actually thinking about what is the scale and size of this defense market that I’m I’m trying to go after. And maybe you’ll be uh what what Under Secretary Michael called, you know, one of these new primes. you’ll you’ll start building different types of products to chase other opportunities. Um >> yeah. >> Yeah. So now that we all know how to build a defense startup after this

3:35:26 15-minute conversation, um what should founders in this room do differently come Monday? >> Yeah. Um so I think founders are always trying to do a million different things. I think one of the most important things is sequencing um the things that are ahead of you and the things that need to be done. Um first of all, I think persistence uh with the vast array of people that you need to sort of expand your surface area

3:35:57 too in order to find where you actually fit within the Department of War is sort of like the thing that you need to do yesterday. And if not yesterday, then Monday or or whatever the next day of the week is. Um that means reaching out to 20 30 maybe more people a day and consistently reaching out to them until you can say um you know I I’ve run this contract to ground or not because I think the biggest problem in the department of war is people are not very

3:36:27 good at saying no to you. Um, if you go through a actually an RFP process or a commercial solutions opening process, an actual, you know, where you submit your technology, you’ll get a yes or a no, but you generally won’t get a lot of feedback because there’s a lot of people making the decision. It’s actually quite hard for the Department of War to give you a very clear yes or no. So if you’re sort of trying to get yes or nos from individual people along uh sort of that that that first

3:36:57 engagement where you’re expanding your surface area um I think that sort of gives you a very good foundation um for having a defense tech startup. I think that’s thing number one. I think the other thing is, you know, raise money and understand your capital stack in a way um that you know understands that you’re working in the national security business. Um you can’t just take money from anywhere like you can in the um if you were in the private sector. Maybe you have to be a little bit more

3:37:27 thoughtful about who’s on your cap table, who you’re giving up equity for money to. Um, and that will get more and more important as you get more and more contracts. Um, because those um, those people on your cap table ultimately will be uh, evaluated and maybe at some point a reason for someone not to fund you in the future, >> right? It’s a good place to end. Thank you so much, Eric. And everyone, Eric’s book is available outside. You’ve probably seen it on these tables and

3:37:58 also at the happy hour. We’re going to have a signing for his book, Build for Defense. So, Eric Alberg, thank you so much. Appreciate you. [applause] [music] Uh, can we do one more voice of God reminder to the lobby to folks need to come in as the uh we’ve got next panel

3:38:29 starting. Um, it’s my [clears throat] honor to uh introduce um uh Senator Tim Shehy who represents Montana in the United States Senate and serves on the Senate Armed Services Committee. He’s a Naval Academy graduate and a former Navy Seal officer. He founded Bridger Aerospace, an aerial firefighting company bringing experience as both a military operator and entrepreneur. And Vivian Shin is a visiting partner at Y Cominator and the founder of education companies Juny Learning and Acy. She

3:39:00 previously worked at Google Mckenzie uh and the startup operator and studied computer science at Stanford. Please welcome the senator and Vivian. Thank you so much for joining us, Senator. >> Thanks for having me. >> Of course. You have had an incredible career as a Navy Seal officer, founder,

3:39:30 and now senator. Tell us a little bit about your journey. Uh sort of the lessons that you learned at each of those stints. Well, probably more applicable today, uh, Bridger Aerospace was our aviation company, but actually uh, uh, more applicable to this was my defense technology company, which now it seems like there’s a defense tech company on every street corner. But in 2015 when we started it, uh, defense tech at that point was when I was in the Bay Area trying to raise money um, from many of the folks who are now defense tech investors, they literally laughed at me. Literally like, wait, you’re you you are starting a defense hardware

3:40:02 company. >> Uh, and quite literally laughed at me. And uh so that that company was called the scent vision technologies ABT and we invented a family systems called the Mattis family systems lattis IMATIS Mattis XMattis and that was the first uh counter drone program record system uh back when counter drone was again a thing most people hadn’t heard of this is 201617 and and uh scaled that company from scratch about two guys in a barn and grew it to about 300 employees and and exited that uh to a strategic in 2020 and um and as a result I think I’m the only person on Capitol Hill who has

3:40:33 founded and scaled a defense company from scratch. And as a result, I’m passionate about defense acquisition reform. I’m passionate about uh clearing out the bureaucratic hurdles and the sclerosis that has created a dysfunctional buying uh system and a dysfunctional customer on the other side of the table. And as much as people want to criticize the primes, the Boeings, the Lockheedes, the Northrups, and blame them for where we’re at, they certainly have some blame to bear. But the reality is uh the Pentagon trained those customers and created that ecosystem over 35 years of of of bureaucracy and sclerosis. So, um, my top priority on

3:41:05 the hill, uh, is to is to reform that acquisition process, everything from specification to requirements to actually how we physically buy those items to then, of course, most importantly, how those items end up in the hands of the war fighter on the front line. And um, as a war fighter, several deployments overseas, my wife’s also a war fighter. She’s a Marine Corps officer. Uh, we take this responsibility seriously. So I’ve been leading most of the effort the last year and a half on the hill to to both uh lead the legislative process but I’m also deeply involved with the department um uh at the member level secretary level and a

3:41:37 lot of our program offices both in uniform and civilians to make sure we’re seeing the adoption of the technology um and actually fielding it because the military actually hasn’t really had an innovation problem. It’s had an innovation adoption problem. We have all these labs and all these AF works, softworks, all these DIUs, all these on-ramps and all this, you know, mumbo jumbo and it’s like how many labs and innovation commands do we have? It’s ridiculous. Like, you know, and then we innovate and then it sits in a box and no one uses it. I’m more focused on getting that innovation and and fasttracking it to the front line. And

3:42:08 um uh we’ve done that in some corners of the military, but we want to try to make that enterprisewide. And I think this leadership of the department now has really done that. Well, >> that’s amazing. Well, actually that’s a good place to start because I know a lot of founders in the audience are curious how you got your first government purchase order when you were running your company. Who took a bet on you? Um, and what mistakes did you make and what kind of things would you tell them as well? >> Yeah, I mean made a lot of mistakes. Um, and of course knowing what I know now as a senator sitting in my office with people coming through every day. Um, you

3:42:39 know, I never went to the hill as a founder. I was a Navy Seal team leader who had no business experience and we started the company and it was me and a lot of my my former teammates that started it from scratch and and we never went to the hill. We never went to the hill and advocated for contracts or for appropriations. In retrospect, if you know Jared Eisman and I are good friends and we go back many years because we each had companies in in a similar space. He of course runs NASA now and we joke that if him and I had known what we know now back when we ran our companies, we would have spent a hell of a lot more

3:43:09 time on the hill. uh w with our hat getting passed around because uh you know both of us were kind of naive tech founders who just thought the best product would win um and we didn’t realize that you have to go much further upstream um to be to have max effect but uh how we got our first customer was was our initial business was Bridger Aerospace we were fighting wildfire from the air and we had an infrared sensor system that could sense fires more accurately from the air and I got that uh experience having served several deployments overseas during GWAT uh having that airborne surveillance

3:43:39 capability was a game changer and a lifesaver for us on the ground. And about the time I got out uh the investigation was complete of an incident uh the Jarnell Mountain Fire. You may have heard of it. It was made a movie about it called Only the Brave where 19 hot shots were burned alive on the side of a mountain in Arizona. And Eric Marsh, the team leader, as I read that debrief, you know, I could feel myself in his boots having been on the side of mountain in Afghanistan on missions calling in air support. And I realized that um him and his whole team died not because he made the wrong decision but beca he he made the right

3:44:10 decision with the wrong information. Uh he did not know what the fire was doing and where it was going. Uh because he didn’t have access to the type of information that I had half a world away fighting in Afghanistan. So we took that solution and applied it to wildfire and in the process uh caught the eye of of some uh customers in the defense space who liked what we were we were uh pushing out there and and we we literally sold guerilla style at the lowest level and um the Marine Corps lad community became our first major customer who really took a leap of faith on us as they were dealing with the with

3:44:40 the drone penetration problem back in 2016 and we had no air defense capability. were still pretty soft in that regard to be honest with you. But nobody had yet fielded anything capable of detecting, intercepting, defeating, or even identifying uh small drones. So they took a bet on us and we uh from back of a napkin to field it in combat in eight months. Um we’re really proud of the systems we created. >> That’s incredible. Um and along that same vein, so it sounds like you were you sort of started as a dualuse product. I think there’s a lot of uh

3:45:10 founders in the audience who get pressured to do dual use instead of you know focusing on one side or the other. Any advice that you would give them for for that space? Focus is key and I think when you try to do everything you do nothing well and I think you have to focus on on what’s the point of your product and um you know when I was trying to raise money to scale my defense company like I said and everyone laughed me out of the rooms which then when we sold the company for $350 million and I didn’t have a whole bunch of VC taking all the gains okay like I’m glad I didn’t I’m glad I

3:45:41 couldn’t raise money but you know it was pretty frustrating at the time when I couldn’t pay my bills but um the the the the obsession with what the mission of the product actually is I think is incredibly important. And if you’re trying to create a product for everybody, you’re going to create a product for nobody. And uh if your product is a consumer product, then make it a consumer product. And I’m sure there will be or could be defense applications. If it’s a fence product, then make it for your customer. And um I think starting with a customer and the mission and working backwards always leads to the best product. Starting with

3:46:12 a product and finding a customer um generally does not. So I think figure out who your real customer is, focus on that customer and if it becomes a dulios product along the way um great but uh especially in this current environment um I think the the do now is a real customer. I can promise you 10 years ago oftentimes it was not especially for a startup but but now it is and I think you can raise money you can build a business you can create a a product pipeline um that’s real based on DOW’s

3:46:42 not just their their rhetoric but actually what they’re doing >> makes sense and then you also mentioned that you know in the appropriations acquisitions process there probably could be some changes that could be made especially for a lot of the founders that we work with they’re sort of in the prototype phase and they want to get to the scale up so what kind of changes do you think that Congress could make to make that process easier. >> I don’t want us to have a defense industrial base. I want us to have an industrial base that the defense department partakes in. >> And I think this narrative that we need

3:47:13 to have a dib, we can strengthen the dib, you know, increase the dib, whatever. Um, you know, Ford made most of our airplanes in World War II. You know, GM made, you know, they were making, you know, pizza pans, making hubcaps. you know, we have the ability, the best telecommunication company that’s making the best phones for consumers is probably also going to make the best phones for war fighters and the best communication devices. And if we bifurcate the the excellence of the American economy, um we’re not going to end up with the best players uh in in

3:47:44 both fields. And I think if you’re making the best vehicles for the American consumer, that’s probably also going to be the best vehicles for our war fighters. You’re making the best communication devices, so on and so forth. And I think that’s how we’ve seen in the last few years. I mean, if you remember, again, I keep heating going back to my founding years, but those are the years when Google and Microsoft wouldn’t talk to the DoD and they had, you know, workers walking out of their buildings saying, “You take you took $1 from the American Defense Department. That’s a genocidal evil organization, even though I’m an American and live under this umbrella of freedom that they provide. I will not work at this company

3:48:14 if you do business.” And that was a real thing. If some of you don’t remember that, that wasn’t that long ago. That was like 2015 16 time frame. Um, now the conversation has changed greatly. Thankfully it has um but we lost a lot of talent in that time period because we we had you know bifrocated economies. We want the same companies that build airplanes for civilians to build airplanes for the military because that commercial iteration is going to give us a higher quality of product and a higher capacity of volume because if and when the next big one comes we are still producing for regional low inensity

3:48:46 conflicts. if and when we have to do what Ukraine’s doing now. I was just there last week and you go in their underground factories where they’re building more drones in a day in Ukraine than China’s building in a day. And they’re building them underground in hidden factories accluded all over the country. Uh when we have to do that, hopefully we don’t, but when we do, it’s going to be really frustrating if we have to now create a a separate industrial base from nothing. So it both will breed better product excellence. It’ll breed breed better innovation within the human capital ecosystem.

3:49:16 It’ll also prepare us to have a more scalable defense production base if and when that time comes. And if we keep them separate, we’re not going to have that. And that’s what we’ve done for 30 years. That’s why China built ships 232 times faster than we do because we let all our shipyards close. That’s why they build aircraft 50 times faster than we do because our our aircraft factories have split into defense factories and civilian ones. Um the only industry left that really still exists from a manufacturing standpoint in parallel is the auto industry. >> That makes sense. Actually, that’s a great segue into my next question, which

3:49:46 is, you know, for a lot of these companies that have smaller orders, they’re sort of small suppliers. How can Congress work with them to really scale up their production, potentially help finance it or help them find ways that they could um know know a little bit more about the future before they invest >> actually purchasing and quantity? uh again the innovation contracts you know we don’t really have much work to do there between the all the different labs and DARPA and the on-ramps and everything there there’s innovation grants from SBS all across the board if

3:50:17 you want to create a product or take a product from TRL 2 to TL 5 or TL 6 to TL8 there’s plenty of programs and grants out there to do that that’s not my concern and frankly I don’t really care because we have plenty of innovation I’m I care about taking the product that we’ve made 10 of and it works okay and how do we make a 100 thousand of those quickly and economically because that’s the other side of this equation that that doesn’t get talked about enough is the attitability of our systems and the reality that at some point a conflict becomes economic in nature. And if we’re spending 10 million a shot and they’re spending $10 a shot, that’s a problem.

3:50:49 And although in a brief conflict where we have massive overmatch, it’s fine. In a sustained conflict where we’re dealing with a peer adversary, and stop using the word nearper, I get tired of that. China is not a nearper in every respect. They are a peer. Their navy’s bigger than ours. Their air force is bigger than ours. Their space assets are more than ours and more capable than ours. They we have force overmatch against them. I should say they have force overmatch over us in every single operational paradigm. It’s not an earpier they appear. Um so we have to appropriate and multi-year appropriations which we’re getting

3:51:20 there. It’s hard to get, you know, big changes through Congress, but I’m spearheading that and we are making progress so that there can be multi-year assured assuredness for companies and for investors say, you know what, there’s a 5-year appropriation here to buy medium-range drones, long range drones, ground vehicles, UGBs, USBs, whatever it is. Uh, we can invest behind that. We can build a factory with five years of certainty. We can’t build a factory with one PO, but we can build a factory if we know they’re going to be buying 10,000 a year for the next 5 years. And what I’ve explained uh given

3:51:52 my business background and and this department doesn’t need a lot of that with Fineberg there and others. They have plenty of business background but reminding them that we have never had [clears throat] this amount of private capital available to create our own defense infrastructure. It’s been the American taxpayers been paying for it out of hide for a long time. And now we’ve got 8VC and we’ve got Andre and Horowitz and we’ve got you know Kleiner Perkins. all all the best money in the world is now available to fund our defense uh procurement ecosystem. That’s not going to last forever if

3:52:22 we’re not a good customer. They’re going to take their money and go elsewhere because money follows yield. And if there’s no yield here after a couple years, and if any of you were around when the drone bubble burst in 2017 when, you know, everyone had their own consumer drone company, well, eventually the buying wasn’t there because DJI crushed everybody. that capital dried up very quickly because these these firms, many of which you know of, will very quickly move their money to where the yield is. So, if we’re not buying, that capital will go away. >> It’s very important we send a steady buying signal that is longer than just a CR, a few months in one PO. And that is

3:52:54 happening. We’re changing that. Um, but there’s some more work to go there. >> Yeah. I mean, one trend we’ve seen at Y Combinator as well is the defense companies are raising some of the largest rounds at demo day and then also sort of the downstream of their supply chain as well. those companies are really scaling up as well. >> And that’s key because, you know, the the unmanned ground vehicle or the surround drone boat or the counter UAS jammer system, all those are great, but if they can’t get the microchips and the thermal camera pads, and they can’t get the processors, they can’t get the

3:53:25 magnets, the batteries, whatever it is, because their downstream supply chain is either entirely housed in China, uh, or it simply doesn’t exist and is not resilient, we’ve got a big problem. And that’s where Ukraine is probably going to save our bacon because they have already cracked a lot of this. Um the innovation that’s going on over there is pretty incredible. We’re not going to fight a war like they’re fighting it. Um we’re not necessarily going to have the same contours of a conflict as they are. But when you look at the agility and how organic uh their their production capability is and how fast they’re

3:53:56 iterating, uh it’s fantastic because um we had something that could do that many years ago. It’s called Gaido. Jaido was not perfect, but Gaido was stood up with authority, with funding, and with responsibility to solve a very specific problem set, which is IEDs. And it was incredibly successful. And for over a decade, it saved thousands of lives, mine included. Um, and iterated technology in real time. It didn’t iterate technology in two-year cycles. When our enemy was changing the design of their IEDs overnight because they saw

3:54:26 how we disarmed it the day before, they would change the design. And in the previous paradigm, it would take us months or years of PowerPoint slides and committee meetings and RFPs and RFQS and sources sought to to feel a new solution. Jaido cut through that and it did so well. Do you know what we did with it? >> We got rid of it [laughter] because why not? Why would we take why when something’s doing so great, why would we keep it? And we didn’t. We got rid of it because the the bureaucracy was so eager to go back to how it used to do business. It just couldn’t wait to get rid of this aberration and go back to

3:54:56 the old sclerotic sclerotic ways. So when we look at, you know, uh, uh, you’ll hear an announcement today from Secretary Hegsth about something else coming down the pike that’s going to be pretty revolutionary as far as how we can take robotics and autonomous systems, remote systems, AI, and integrate that actually into our force structure and and and and protect it in statute uh, so it doesn’t just go away. Again, >> makes sense. And then um, actually at Y Comodator, we also work with a lot of founders who served in the military like yourself. What kind of advice do you have for folks in the military who might

3:55:27 want to start a company one day or who already have? >> So, I I became super active in kind of the transitioning veteran, you know, mentor space. I’m actually going to the Honor Foundations have an event tonight. I’m going to uh great organization there. My old CEO, Matt Stevens, used to run that organization. I’m sure some of you guys know I’m there as well. and um you know getting out especially if you’re getting out of an elite community, a fighter pilot community, special ops community, light infantry, you know that Aspree core, that mission, that that brotherhood or sisterhood that you have there um can be really challenging to recreate when you get out

3:55:57 and and there can be kind of that identity search where you’re like, you know, what am I going to do next? Um and for me that that purpose and that mission for what why my company started with another veteran and I two vets started in our barn. you know, we found that mission in uh protecting our communities from wildfires and protecting our war fighters uh from drones and bringing that capability to the front line and and keeping that connection to the war fighter was huge. And I think the same uh lessons that made you an effective junior leader, whether you’re an NCO or junior officer, um you know, bring bring that leadership

3:56:29 paradigm to the private sector, but also recognize your shortfalls and recognize that if you’re all a bunch of veterans fresh out, you probably have some strength. you probably also have just as many if not more blind spots. And um I usually encourage you know vets if they’re getting out um you know don’t jump right into something because you’re probably going to figure out your first job is usually not the job that you’re going to stay in because you’ve got to get out figure out what it’s like up down left and right. Um and luckily there are some great um programs now. The Skillbridge program is fantastic.

3:57:00 There are a lot of different internship programs especially if you’re coming from the soft world. They’ve done a fantastic job of really injecting that talent into the industry and saying, “Hey, this guy’s on his Twilight Tour. We want to get him embedded with this company so that it becomes a two-way, uh, knowledge transfer. Um, that’s pretty powerful. But, um, at the end of the day, at some point, uh, you know, as Cortez says, you got to burn your ships. At some point, if you’re going to be a founder, um, you’re going to walk away from the shity of a paycheck and a benefits package. You’re going to have to step off into the darkness and start a company. and and that can be a super

3:57:32 lonely thing to do because especially if you are the founder. Um you’re going to have to make decisions that can be every bit as challenging as decisions you had to make on the battlefield. Um and make no mistake, lives are in your hand. Uh the livelihood of your employees, in my case, the livelihood of my family because we had nothing other than the company I just started. So it was do or die. Um and of course never forget the livelihood of of the war fighters who were going to acquire your equipment and use it downrange. Um and for us, you know, we always took that incredibly seriously that this product was going to go forward and it was going to be used by somebody uh who who was quite

3:58:04 literally going to be fighting for their life. And and I think as we had young engineers joined the team over the years, um we did not shy away from informing them the gravity of the work they were doing that every every screw they screwed in, every line of code they wrote had to be ready for combat because that’s where it was going. >> Incredible. I think that’s a great place to stop. But thank you so much, Senator, for speaking with us. [applause]

3:58:53 feel. >> [music] >> There we go. We do have some slides here as well, but uh check check click click.

3:59:25 Hopefully we do. Otherwise, this is going to be very difficult. There we go. Perfect. So, who we are? We’re Arendelle. And what we do is we build the whole ecosystem plug-andplay for shooting down anything from fiber optic drones all the way up to Gar Fives and newer jet powered threats actively deployed in combat confirmed shootowns. And we’re a team of all combat veterans. And why is this doing this? There we go. So, I’m sure a lot of people in this room, everybody’s heard this, Patriots

3:59:55 are great. They’re very expensive. We can’t build enough of them. And current systems that are in use in place as well are very expensive. When you compare that against a $50,000 show head, $80,000 if you have the Jetson fitted, there’s an major issue with in terms of economies of scale. And additionally, as well, a lot of primes, they sell the hardware. Most buyers as well cannot absorb that. The training doesn’t necessarily go with it. A lot of kit sits in crates ends up being unused while the threat actually develops.

4:00:28 So where we play is we own the low cost of tritable layer. So what we do is we shoot down anything from fiber optic drones. In fact actually we’ve got a nice little infographic on the bottom. These are some of the things we’ve actually shot down in Ukraine right now. So fiber optic FPV drones, molasses, lancets. I particularly hate lancets because they make no noise when they come in as well. They’re very very scary. Uh Garen 2s, Gar 3s. So that’s kind of your Shahhead 1 136s and 236s as well as your Garen fives which is the newest system we’re pushing out right now.

4:01:00 We do that via our range of products. So we have our base platform which is our Aloss system. This is our base interceptor platform. We have different variants depending on where we operate in the world. So our desert variant is slightly different to our cold weather variant for example. This is fully autonomous in that it does not require a pilot at any stage of the process. We’ve reduced months of training down to three mouse clicks. Launch, pursuit, intercept. And essentially what you get is a nice live stream of this doing the work for you against the show head where if you’re active military and you want to come have a look, we have some footage we can share with you as well.

4:01:33 We also have our hollow site. Correct me if I’m wrong, but I believe this is the lightest thermal fuse holographic site in the world. Thinking let you spot people in the dark. And then we have our Orion headset as well, which we originally built as a sensor for our interceptors. So this is thermal fuse night vision with a rangefinder built in. You can look at a target, laser distance, reports bearing, heading, everything you’d need for a fires order. And then you could watch an AOS going in on that as well. But turns out we actually built a super powerful infantry tool as this is the only thing in the world that allows you to shoot and move

4:02:04 as a drone pilot. We also have our entire training track that goes with that as well. So we update our techniques every two weeks direct from the front line and we train on everything from flying to engineering navigation as like a section commander for sergeants corporals depending on where you are even down to county fortifications and interceptor drone flying as well and we will do all of that with no Chinese components at any stage of the process. So this is essentially what we do. Uh we

4:02:35 operate under three budgets with one vendor. So county soldier senting training and we can actually do the managed services wrapper that comes in on that as well. So we train to equip. So if you have need the plug-andplay capability, we spin up the whole thing from scratch. We’ll skip ahead of this and I think this is a very very good point. The guy that made the slides when we put primes a lot of companies sell a system and it’s all well and good having a system. How do you actually use this

4:03:05 system? How is this system deployed? How does this system updated and relevant as as threats go? You look at the drone marketplace today, it’s a game of cat and mouse. There’s a new hardware iteration every couple of days. You know, when you’re talking six months of authorizations to get something across the line that’s nonviable. You know, we have a new variant every two weeks on our hardware every two weeks. And this is what it takes to actually shoot down drones in contested environments. Does it work under GNSS spoofing? We do. Does it work under electronic warfare? We do. And that’s something that being combat

4:03:36 veterans, we learned the hard way. I still have big scars on my neck from where we got it wrong quite a few times as well. So, let’s talk a little bit more on the team. And we have myself, James, my co-founder Dan sat front and center over there. We’ve got quite a bit of experience between us. We’re surrounded by incredible guys. So, actually there’s uh one of the guys we have in there called Pedro. He’s our capability lead. He shot down so many shad heads by himself that the units stopped paying him the unit bonus because they’re running out of money. And between the team, we shot down more shads than half

4:04:07 of NATO combined at this point. So if we don’t know what we’re doing, nobody does. In fact, actually, we shot down more uh charads and garan drones than uh US Navy did over the recent Iran conflict as well. So pretty handy there. [snorts] But I think the key line that we’re trying to get across here built by operators for operators. So, we know full well that our equipment is going to be wrapped in sniper tape, dropped in a puddle. Every shiny bit is going to be covered up with some kind of matt tape, and it still needs to work. I really wish this would move a little

4:04:38 bit quicker. So, here is what we’re based around. We give the West the ability to outpace the threat. We bring that iterative development back and we’re directly plugged into the front line now so that countries that aren’t necessarily exposed to these contact zones still get the benefits of that frontline knowledge. We have orders going through. We’re uh actively deployed in Eastern Europe. We’re actively deployed over the Gulf. In fact, we’re one of the few interceptor companies that has confirmed hits over the Gulf and over Ukraine as

4:05:08 well. And I don’t have too much else to talk about. So that was us, Arendelle, giving the West the ability to outpace the threat. We’ve got a booth out there. If you want to see some interesting stuff, come and have a chat with us. Thank you. [applause] [music]

4:05:40 >> [music] >> All right. Well, uh, thank you to the hosts and, uh, I’m Edward Henson, the CEO of Zephr Fusion, where we are building compact, uh, megawatt scale fusion reactors for orbital applications. Uh, the team is uh, uh, comprised of a

4:06:11 number of experts uh, in uh, in the relevant fields. Uh, my co-founder and myself were lifelong fusion physicists uh, coming out of Oakidge and Lawrence Livermore National Labs. So lifelong uh, uh, thermonuclear fusion physicists. We have engineers from Turan Space, NASA, Proxima Fusion, Commonwealth Fusion, Type 1, and advisers from Relativity, Planet, Air Force Research Lab and and NASA uh including with space nuclear experience. We got started with the project because uh every major space

4:06:42 effort is power starved. Uh it’s almost something we’ve forgotten about. Uh there’s a huge fraction of the total engineering effort that just goes into to launch an asset that uh that that goes into working around this problem. In practice about 10 kilowatts is the working uh uh ceiling of what what actually operates out there uh commercially. Skylab broke that uh that ceiling in uh the 70s early 70s is basically stands today outside of uh government uh assets. Uh but whoever

4:07:12 delivers me power in orbit will become a foundation layer of the space economy. Uh this stands in stark contrast to the uh the future a lot of us grew up reading about. There are by now uh decades of well-th thoughtout applications that in manufacturing for novel materials and microgravity uh transport propulsion applications and and um uh uh many more that simply await a source of this this power. There’s also compute as a major uh uh ambitious

4:07:44 new effort also is uh is uh uh a lot of times discusses how essentially a discussion about how to get the power uh there. Uh and as ambitious as it is gigawatts have been proposed none of the individual installations exceeds about 150 kilowatts per panel. And this is because solar scales with area. So if you uh are doing something under uh 100 kilowatts that’s uh it’s probably the solution or doing it using a distributed load. But if you want to get something

4:08:14 uh powered at 10 megawatts locally uh you’re invoking a an area that’s going to be uh span the the base of the great pyramid at Giza uh in orbit. Chemical uh energy deserves mention because it powered the industrial revolution down here. uh but uh but it doesn’t and it gets us up to orbit but it doesn’t pay its own freight uh beyond that. But what does pay its own freight is nuclear. So nuclear is a million times more energy dense than something you get out of a battery. Uh and then fusion that’s

4:08:45 fision and fusion on top of that gives you another factor of 100 per unit fuel. So uh you know 1 to 10 gatt hours per per gram. So you can um yeah it’s a megawatt day basically per gram. So why [clears throat] fusion in orbit? Um the topline result of fusion research over the past few decades is essentially that you need lots of volume and you need a lots of field inside that volume in order to confine the plasma well enough to heat it to fusion temperatures. Uh this large volume then sets the uh the

4:09:18 size of the entire reactor. There’s a bunch of layers of infrastructure that have to be built around it that set the then the total cost uh in the uh uh 10 to 100 billion range. So this is eater. This reactor this is $65 billion piece of hardware that will achieve 500 megawatts thermal. Uh what you can do in space is of course leverage the natural vacuum environment and all you need the only resource you need to pro provide is is high magnetic field and that in the past 10 years with commercialization of tapes this high temperature superconducting tapes for the first time

4:09:50 uh has become possible and in the past five years it’s dropped by an order of magnitude in price. So then the uh the expected performance of this type of reactor is uh very favorable compared to to the best of best-in-class uh currently. Zephr’s uh program has uh has now initial designs for 1 to 10 uh megawatt installation. This would be something that could fit in single launch capability with room left for the payload. Um but what we are doing uh with our hands and feet now is is

4:10:20 developing Balas which this is our first demo. Uh it will uh field superconduct high temperature superconducting tape uh in orbit and it will levit levitate the uh uh the coil according to a method that we’ve developed and patented uh out in front of it. Uh so this will be a a first of its kind demonstration. This is uh we’re in the process of signing launch paperwork for this. This should be late 2028. uh the ask would be that fusion have a lane uh alongside fision. So this

4:10:52 leverages existing investments that the government has already made in this type of infrastructure because much is in common with uh held in common with with fision. So fusion has the same generally same energy conversion needs uh heat rejection uh pathways and uh and high power bus work etc. So a lot of the infrastructure can be developed in parallel. Uh however, fusion uh has a uh uh uh a non-hazardous bill of materials on the launchpad. So it has a distinct advantage there and as I said it has it

4:11:23 operates with higher energy density fuels. Uh another strategic importance or aspect of of fusion is that it can uh can be used to create uh uh uh uh sensitive uh uh materials that uh that that that would not otherwise be be available otherwise very difficult to uh to produce. Uh so with that uh I’ll yeah we’re Zephr Fusion we were founded a year ago now up to uh a team of 10 people uh and we plan our initial demonstration late uh 2028.

4:11:56 Thanks >> [music] >> In case you didn’t notice, we are entering our uh space section of our

4:12:27 programming uh which is very exciting. And uh after this amazing panel that we’ve got, we’re going to have the NASA administrator and fireside chat. So encourage folks who are out in the hallways to make their way in uh because uh this is some very exciting stuff we’ve got going on here. So uh Bram Shor is the co-founder and CEO of uh General Astronautics which builds robots for research and manufacturing in microgravity. A Caltech train aerospace engineer. He previously worked at SpaceX’s Starlink laser systems and industrial autonomous robots. Skyler

4:12:59 Chan is the founder and CEO of GR space which is developing lunar construction and infrastructure. He studied electrical engineering and computer science at Berkeley and previously worked on a vehicle software at uh Tesla and NASA funded 3D printer launched into space. Alex Lisso is a co-founder and CEO of Inflection Space Systems which is building spacecraft manufacturing capability. Uh his engineering experience spans spacecraft designs, testing operations with work at Millennium Space Systems, Rocket Lab, Planet Labs, and Interloom. Jacob Assos

4:13:32 is co-founder and CEO of Orca Aerospace, which builds AI spacecraft operations that run on board the vehicle. He spent more than a decade at Loheed Martin and guidance, navigation, and control, including starting its first AI focused team in that field. And last, but certainly not least, uh really honored to have Jacob Reese’s who is our moderator. Jacob very recently serves as the chief of staff to Vice President JD Vance, uh both in the Senate and the Vice President. He previously advised Senator Josh Holly and is a Stanford Law graduate. Please welcome to the stage

4:14:04 Bram, Skyler, Alex, uh, and Jacob in conversation with Jacob Reese’s. All right. Uh, thank you all for joining us. Really, really excited for this. Um, so for each of you, uh, what are you building, what exists today, and, uh, you know, any other sort of introductory

4:14:34 spin that you’d like to put on it? >> Go ahead. >> Yeah, happy to kick things off. Uh, I’m Bram, founder of General Astronautics. We build super intelligent robots for onorbit and lunar applications. So today everything from orbital warehousing to asset resupplies to cargo management uh is a labor problem and we have a labor shortage. So general astronautics is building the orbital fleet to fix that. Today we have robots that are 26 years old sitting on the international space station. They are uh decaying and

4:15:07 despite what they’ll put out there uh there’s numerous problems with them and they’re they’re not intelligent systems. And so, General Astronautics is applying kind of the novel approaches with uh reinforcement learning and uh robot control policies to build fully autonomous robots. Hey everyone, my name is Skylar Chan. I’m the founder and CEO of Guru Space. We’re building America’s moonbased company. Our goal is to build the first cities on the moon and Mars within our lifetime before we die. And we have a

4:15:37 mission that is going to the moon as early as next year to manufacture the world’s first product on the moon, the moon brick out of moon dust material. We’re based in Hawthorne, California, and we’re on a mission to change the course of human history once we build a foundational layer for America’s moon base. Whoever gets there first to the very few resources that are left currently unclaimed on the moon will control the moon and we have to make sure that we’re the first ones here. So very honored to be on this panel today. Excited for everyone. >> Hey everybody. Uh [clears throat] my name is Alex Lisso. I’m the co-founder and CEO of Inflection Space Systems. Uh at Inflection, we are mass manufacturing

4:16:09 spacecraft. Uh we’re targeting a production capacity of 10,000 satellites per year in the early 2030s. Uh today, the space industry, manufacturing in the space industry is sort of structurally uh high cost and low volume, you know, like it’s very normal for a single spacecraft to cost millions of dollars and uh a volume a manufacturing volume in the hundreds per year is aspirational at best. Uh we believe this fundamentally needs to change at like a fundamental level if we’re ever going to move infrastructure or industry off planet and it’s really critical um for this to happen really soon in America uh to ensure that we maintain our strategic

4:16:40 edge um as other countries sort of catch up to us and like launch capability and cadence. Um so yeah, we have our first launch booked for about a year from now. Um and we’re looking forward to getting um a flight heritage on our entire hardware and software stack. Um I’ll leave it there. >> Very cool. Hey everybody, Jacob Waski. Super excited to be here today. I am co-founder and CEO of Orca Aerospace. Uh we are building super intelligent spacecraft operators that live fully on board spacecrafts and satellites for full uh fully autonomous systems to uh

4:17:12 not rely on the fragile ground link anymore. So right now we believe that Orca is the only option for building and scaling constellations to be able to uh work with satellites and keep missions operational when ground stations are jammed, denied or fully destroyed. We have our first launch with a major aerospace prime within the next six months and we are going to be building

4:17:44 the future of onboard operators. Thank you. >> Great. Um, Alex, what has to change to make a spacecraft production repeatable so that you’re not turning every project into a sort of new engineering project that has all sorts of new cost time associated with it. >> Yeah, totally. Yeah, thanks. um a lot of things I think um you know top of mind for myself and our team right now is supply chain right so you know the the individual piece parts that constitute a spacecraft the micro electronics the screws and fasteners the adhesives even

4:18:16 um are almost entirely you know uniquely made in a boutique way for space space applications right um this means that like not only are they really expensive but they’re prone to supply chain bottlenecks and uh really rapid parts obsolescence so you know if you design a satellite and you want to launch you know build more and launch more of them um often often times you need to go back to the drawing board on some of these these uh these are very small pieces and end up redesigning entire sub assemblies or the entire spacecraft. So it ends up being like you said a new engineering project every time even when you’re notionally building the same thing over again. So the way we’re trying to design

4:18:46 around that is by using only terrestrial parts you know looking at established high volume manufacturing industries like uh consumer electronics and drones, cars and the like. Um making sure that the entire hardware stack is based only on these robust uh industries and these supply chains so we can build in a lot more scalable and robust way. Um the other thing that comes to mind is uh like the interfaces that you know we use across the space industry. Um whether that’s between spacecraft u you know that are docking and interacting whether it’s between uh spacecraft and their payloads. Um you know space for so long has been like this almost science

4:19:16 project right where we trying to see what’s repeatable what works again and again. uh but there’s a bit of organic standardization happening now across the industry where we all realize oh it’s better if we all use Ethernet where we all use the same voltage levels for the same mounting patterns right so I think it’s going to be really important for that trend to continue um and for you know partners on the private and government side to find more ways to sort of standardize interfaces and uh basically move quicker to get sort of uh new things integrated and launched really really fast >> um Jacob great name how do you get customers to trust the idea of taking

4:19:49 decisions on the ground, bringing them into space, having those decisions. What sort of validation do you need to provide customers that they’re going to get what they need? >> Yeah, it’s the same way that traditional aerospace, you know, new products like what Alex is building um goes off the ground. You go from you climb the TRL readiness level. You go with and test your operators from software in the loop, processor in the loop to hardware in the loop. You test on flat satellites and then you do on orbit tests like what we’re launching early in March of next year. And each along the way on each of

4:20:21 your tests, you have metrics in place that you compare it to traditional uh human in the loop human operators and you see if your AI operator is meeting or exceeding the traditional ways and you test that with thousands and thousands of synthetic missions and real missions that you compare it to and see can the AI operator perform as good or better. In our case, we’ve always shown that the AI operator performs significantly better than a human operator on the ground.

4:20:53 >> Um, Bram, you’re building robots for working microgravity. Who’s the first customer who’s going to needed this done often enough that you can make it a repeatable service rather than just an experiment you that they buy one off? >> Yeah, absolutely. So, there are a number of customers. The first one largely is anyone today who pays for astronaut labor. So NASA currently figures astronaut labor at about $135,000 an hour. If you ask the private space stations, that number is about three times that, to half a million dollars an hour. It’s not safe or scalable for

4:21:24 astronauts to continue to be the labor providers for space. So we look at who is currently paying for astronaut labor on the International Space Station, who has requested space on the private space stations. It’s the manufacturing of uh biologics of metamaterials of compute functionally which we don’t think as an asset that we’re manufacturing but ever increasingly so it’s a real possibility in space and so um one of these customers starcloud is uh proudly one of the companies uh part of the $30 million

4:21:54 of commercial LOIs that we’ve gathered uh and so they will be kind of the first commercial customers and then we see on the government side we have a ton of in very interesting applications all the way from Golden Dome for we have orbital warehousing. We have orbital logistics. Uh we kind of look at the way the US defense industry currently works. We have the F-35. When we’re finished with the F-35, we don’t throw it away and let it burn up uh in the ocean or whatnot. Instead, we bring it back and we resupply it and we launch it back up. And so our satellites are going to be no

4:22:25 different. Especially, I mean, we saw two days ago we had Starship launch again. As we get larger launch vehicles, suddenly the economics repairing assets in LEO make more sense. This is not to mention the higher value, higher technology ad assets that were adding to LEO to GEO to the moon. Uh the economics of repair uh make sense today and will only become more profitable. >> Uh Skyler Gru’s [clears throat] ambition is lunar infrastructure and construction. Uh obviously there’s a mature version of that ecosystem. It’s a

4:22:55 really exciting vision, but between now and full maturity of that ecosystem, uh what are the products or services that customers are going to be able to buy that are useful right now? If we’re serious about building a moon base, the first thing that we need to solve is how do we turn lunar resources into the world’s first ever construction material on the moon. So the first step is we need to make, you know, a brick on the moon before we can build a moon base. Then we can go on to build things like landing pads, roads, warehouses, barracks. Literally look around the room

4:23:26 here. Every single thing that we built on Earth has been built with concrete, has been built with construction material. Instead, right now the status quo is we have to bring all of this material from Earth to the moon, which is a huge problem. This is why people when they look at moon bases or moon construction, they think it’s crazy. If you break it down, if you can use the lunar resources on the moon locally and turn that into building material, that’s the kicker. That’s how you bring down your unit economics at scale. Not just for ourselves as a business, but for NASA, for partners, for everyone else in

4:23:56 the industry. America wasn’t built with just resources brought from outside. It was built with American resources. The same principle will be applied on the moon. And the first step is making the world’s first ever bricks on the moon as early as next year. um for each of you uh when you think about the risks to your business and you think about where we’re going in space, what has to go right for your core business thesis to be successful and whoever is most excited to answer that question go first. >> I can go first. Yeah. I mean, it’ll probably be similar to what you said too

4:24:26 because I think the standards and putting in place, you know, common standards across both hardware but also software. You mentioned a lot about the hardware standards that we’re really trying to drive naturally to some cohesive standards. But the same thing needs to be true with software. The same thing needs to be true with data sources that you can ingest different pieces of software, different types of flight software and different disparate data sources into one common synthesis engine. And that really needs to go right across the industry. We can’t keep

4:24:57 building up bespoke pieces of software with bespoke interfaces so that no new companies can go in and build up novel software that can really federate throughout the ecosystem and hardware. And I I think that’s what really has to be the next generation, the next evolution of our like typical bespoke space architecture that we’ve been predicated on for the last few decades. Yeah, I would certainly add on to that just uh we see kind of the incredible advancements in software and then we look at it. We live in just an

4:25:27 incredible age of abundance, an age of abundance intelligence, an age of abundant super intelligence. And we look at like all the domains that that’s going to apply to. We’re kind of at this point where everyone agrees that it’s an not it’s not an if but a when we get this translating to the physical world. So if we think about any sci-fi future that we imagine, we think about what any form of a dominant American space enterprise looks like. It requires the competence and the labor capabilities to own this domain. And so uh it really is not what needs to go right, it’s how

4:25:58 quickly can be accelerated. >> I think the the elephant in the room in the space industry right now is is launch, right? I mean it’s been a big year for launch with, you know, SpaceX going public and maybe winding down the ride share program that the industry has relied on for so long, right? Um I’m really really optimistic u when it comes to launch. I mean um having worked at Rocket Lab and across the space industry I’ve seen there’s I know there’s a lot of really brilliant people and and billions of dollars invested into alternative uh forms of launch other rockets right um we think about like 10 15 years ago before Falcon 9 was really launching in a regular cadence and there

4:26:29 was not a ton there wasn’t much of a new space sector right uh but when once Falcon 9 became like a repeatable reliable uh ride to orbit this entire incredible industry was sort of born and I think um you know when we see three four five more rockets come to market that are reliable um and can launch at a really high cadence. I think we’re going to see just an explosion of growth across the commercial and government space sectors. So, I’m really optimistic when it comes to that. But, um I think you know that is a risk because I mean you can’t have much of a space industry for any of us without getting to space on a rocket. >> I know what Alex is saying like the biggest problem is launch right access

4:27:00 to launch that that’s a problem for everyone here who’s trying to build in space. Um but on top of that like when we do big things like you know we’ve got ambitions to build a moon base the question is often like what is like that key technology like risk drisk that you have to do as a startup. Um I mean for us like we’re our core technology is building an autonomous robotic factory that goes out and excavates lunar regalith soil on the moon processes it compacts it cures it and then places the world’s first product on the lunar surface. Now obviously this is like a crazy thing for some folks and people

4:27:30 ask well how do you like derisk this? like how do you ensure that this is actually something that’s possible? Um when we’re building a moon base there’s a whole sort of different set of technologies that need to occur and especially launch to get this right. Our take is that okay instead of you know for example building large structures on earth and doing like huge demos we should be focused on a critical path like what is like the singular most important demonstration we need to prove to the world to increase our TL to the max to t9 and show that this technology is even possible because at least for us humanity has never built a product on

4:28:02 the moon just think of pure facts like that has not been done before. So this is why if this is successful this becomes a historic moment and then now the question is the key is how can you do this on a lowcost way which is what our first mission is. Uh and so actually this question at is really like been the thought process behind how we’ve designed our own technology missions for our road map. >> We’ve got Jared Isaacman coming up soon. He’s probably listening backstage and so it’s a great opportunity for each of you. um if he’s listening, what’s the most important thing that NASA could do right now to help each of you with your

4:28:33 major milestones? Or uh alternatively, if you’d rather talk about things that have happened, what’s NASA been doing that’s really helped you along the way? And Jacob, you can start. >> Yeah, sure. I’ll happy to start. Uh for building up super intelligent operators, we need training data for missions. And NASA’s the best place that has the greatest history of operations, both operations that went incredibly well, but also operations that had faults and anomalies. And we need to be able to ingest and have access to all that type

4:29:03 of data to build into our operators so it can learn from prior missions and understand what to do in all those types of situations. >> Yeah, I would love to see more like commercial pathways for companies like ours to work with NASA, of course. Um, you know, what I think of specifically is shortening the cycle time between uh maybe like science need identification and and launch into orbit. You know, I know I know from experience a lot of spacecraft, a lot of payloads, vehicles go to space with excess payload capacity, right? So, it’d be great to

4:29:34 find more ways to, you know, just get science instruments up there, right? Like every kilogram counts and if you have excess mass, we should be doing something with it, right? We should be doing cool science. We should be learning more about the Earth, the universe, the space environment. So, um, yeah, finding ways to really shorten that cycle time from like years to like months, um, and then basically move move way faster with commercial companies would be really cool. >> At Grrew, we have a laser focused mission, which is to build the moon base and to win the race and be the first to manufacture product on the moon. Um, NASA is building a moon base. They’re investing a lot of dollars into building this. Uh, we would love an opportunity

4:30:05 to brief the administrator about our mission, see how we can um, make America the world’s first bricks on the moon, you know, first ever made on the moon. Yeah, that’s for us. >> Yeah. So, we live in just an incredible time where we and the royal we as an industry are incredibly fortunate to have the backing of the private capital markets at a time when simply no one else has had access to. And the key benefit of that is that we get to move on disjoint timelines. And while the government’s timeline is rapidly increasing, we get to start today. We to

4:30:36 have started months ago, years ago. And so, General Astronautics, we’re building robots. We’re going on a parabolic flight. Next month we’re going to do a fully autonomous demonstration of our robotic policy swapping GPUs in microgravity. Uh we’re continuing on our p on our uh engineering pipeline completely disjoint from the numerous contracts that we’re pursuing. And so the one thing that I would ask of Jared is we have robots ready and uh there’s some pretty cool opportunities on the moon and our robots would be on those rovers to show humanity what can really be possible with American dominance,

4:31:07 American super intelligence and American labor in space. We’ve got a little bit of time uh left, so I’ll ask one more question. Um what do you think might surprise us the most about where commercial government space uh development’s going? Like do you think there’s something that’s going to happen over the next few years and we’re all going to look back and you know be on stages like this and say I can’t believe that move that fast. >> I think we could spend an entire panel on that but I think the number one thing is that humans really suck at exponentials. Uh one really bad at estimating large numbers. We’re even worse at estimating exponentials and

4:31:39 what is super intelligence if not the world’s humanity’s greatest exponential. And so uh I can sit here and I can conjecture what I imagine it might be but I simply think it’s going to be um unimaginable and uh we just got to build towards that. >> Anyone have a definite guess about the exponential and what it might look like? >> Sorry, what was that? >> Anyone have a very biased take if that’s all right. I think the exponential is it all comes down to let’s just boil it down for a sec. like remove space for a second. Every single technology revolution in human history has always

4:32:10 come down to humans learning how to make use of resources in a more intelligent way. The exponential has always been like how we figured out how to unlock a new set of resources that we haven’t tapped into. For all of human history, we’ve been on Earth. We’ve built real estate on Earth. We’ve been using resources on Earth. But for the first time in human history, we can go again bias to the moon and use those resources like around 38 million square kilometers of free real estate that we can turn into products infrastructure. It’s like the Louisiana Purchase 2.0, but we don’t have to buy anything. It’s it’s

4:32:42 incredible. And so I think like this is the exponential moment in our history where like 30 years everyone’s going to be like, “Oh my god, like this is so obvious. like we should be building infrastructure there on a new place because that’s how every single technology revolution has started where a lot of people are like this is crazy at first then we prove that’s possible and we continue building the technology needed to support this exponential growth. >> I’ll take it a step further. I think you asked kind of what will surprise people. I think people will be surprised at what we accomplish in our lifetimes the next 20 even 10 years. Uh what gets me jazzed

4:33:12 up I mean thinking about this panel right here the technologies we’re working on if you combine them. I mean, we have thousands of satellites that are autonomous that can op, you know, interact with things in space and supported by lunar infrastructure, refueling stations and comms networks. That’s what you need to like mine the asteroid belt, right? Like the crazy stuff that people think is super sci-fi is actually physically possible and enabled by the technology that people are building today uh when we combine it in the right way and uh build lots of it. So, I think people will be shocked um in in 10 years with what we’re starting to do and do in space. I fully agree with all that and the exponentials is my big bet for constellations and

4:33:43 mega constellations and the whole need for AI operators and why you can’t do things like collision avoidance on the ground anymore. But I’ll take it one step further as well. But the need also is going to come from the war fighter. I feel like a lot of what we’re seeing right now is how the war fighter and uh on orbit warfare is actually starting to occur. And I really do think that we’ll be surprised as the battlefield shifts from on the ground, what we’re seeing in Ukraine right now to on orbit and how

4:34:13 that becomes the most contested area that we as startup founders really need to shift our mindsets and focus to build the best technologies for not only our commercial partners but also for the war fighter. >> It’s incredibly exciting. Thank you for all the work that you do. >> We have time for one one or two questions if uh folks have a question. [clears throat] >> Any questions? >> Yes.

4:34:43 >> Hey, Scott Freeman from International Preparedness Associates. Um we actually work at NASA now helping to train astronauts at the neutral buoyancy lab. My question to you is as you moon as you u factor in what you’re doing on moon base are you thinking of the human interaction with your robots? >> Thanks for the question. Yeah, >> our robots are design our factories are designed to be as autonomous as possible because this decreases the risk of um having like human like it’s very costly as Bram is mentioning as well. It’s very costly to have an astronaut on the moon

4:35:14 to operate these factories. So we’ve designed these factories from the get-go to be as intelligent and autonomous as possible so that we can increase the supply capacity of output of uh not just bricks by the way. I mean this is the atomized unit of the key technology which is the construction capability to do all kinds of things. Uh so that’s how we’re approaching it right now. Eventually we’re going to want to have a vision where we have cities on the moon and Mars which implies that there’s a human element there which is exactly on our road map. eventually want to have places where humans live on the moon and

4:35:45 Mars, which I think personally this is like the most exciting thing that we can work on in our lifetime. I mean, what a beauty it is that we can actually have humans live on the moon and Mars. We could have been born a thousand years later, thousand years in the future. We’re born today in 2026. Here we are where we can make products on the moon in this decade. Thank you. >> Any other final questions? >> All right. >> Do you got one, Luther? >> I don’t. Go ahead, Jacob. All right. Well, uh, thanks so much for, uh, sharing everything that you guys are doing. It’s an amazing time to, uh, be a

4:36:17 founder. It’s an amazing time to be innovating right now. And, uh, we’re just really, really grateful for all the work that you’re doing. Excited to see what happens. So, thanks so much. >> Thank you. >> Thank you. [applause and music] >> [music]

4:36:50 >> Heat. Heat. [music] >> [music] >> I think this is one that a lot of you have been uh highly anticipating. Uh so

4:37:20 now we have a fireside with uh uh Jared Isaacman who is NASA’s 15th administrator, an entrepreneur, pilot and astronaut. He founded payments company Shift 4 and co-founded Draken International which provides military aviation training. He commanded inspiration for the first all civilian orbital mission in Polaris dawn during which he performed the first commercial spacew walk. [clears throat] Do Ashi Desyaki is co-founder and CEO of Spacium which is

4:37:51 building fully automated refueling stations in orbit. Her background spans orbital mechanics and rocket propulsion, including work in interplanetary transfer models and rocket systems. Please welcome administrator Jared Isaacman and Ashi to the stage. Thank you so much for being here. Uh

4:38:23 it’s really exciting and thank you wifey for this event. So my first question for you is for a startup building space capability that might serve both civil exploration as well as national security. What would you advise the founders to find their first mission and their first customer? >> Yeah. Well, I want to uh just since I know a lot of these uh where a lot of these questions go, opening opening disclaimer. NASA is like a really terrible buyer. D is great great buyer

4:38:53 in a lot of regards. You all you have to do is look at our our budgets to know uh to know that that that trajectory. But meaning uh especially for like the the big prime type players, we buy a handful of really exquisite things. But what we do really well is help mature technology and that should be very positive for a lot of folks in this room. We’ll probably talk about that on Moonbase and other tech demonstrations. But in terms of like, hey, I want to build something in space and I want to I want to be able to appeal to as many potential buyers as possible. I think it’d be very smart to think through right from the get-go what overlaps both national security needs uh

4:39:25 and potentially NASA’s requirements uh as well. Where to put your attention? I would say uh if well if you’re going to events like this and other u space related conferences, AFA for example or uh you know any of NASA’s events, what are we all complaining about that don’t that don’t have an owner? Just I mean right now what would you hear a lot of? Well, you’d hear a lot about launch. Uh if you want to get into that world, that’s so be it. But there is an immense amount of demand to put more payload mass into orbit. But you hear things like hypergall thrusters. So all the payload buses that NASA procures science

4:39:57 missions on uh have hypergall thrusters. Do satellites have hypergall thrust. So you hear immense supply chain issues on that. But I would say uh yeah step one is like listening to where people’s pain points are right now uh in terms of problems that exist without owners. No one is saying that uh you know oh that company’s got that under control no problem. and we deliver within our our time frames and then yes for sure look at where there’s overlap space is still extremely these are very early days right now there are not a lot of customers that out there so if you can

4:40:29 be solving a problem that has applicability to NASA the DO other civil providers that’s a that is a great place to be >> okay and so how should a 5erson company responds to NASA’s interest in Luna capabilities also where should they start and what purchasing pathways could could it be open Yeah, I I mean I would say that is an excellent place to start for a lot of the the companies uh that would be in this room is Moonbase is an incredible opportunity for you all. [snorts] I mean

4:40:59 we put the demand signal out for lots of landers, lots of rovers. So this is this is actually more akin to how we did things at NASA in the 1960s to build up to very challenging outcomes. Mercury before Gemini, an awful lot of Apollo missions before the the Eagle landed. That’s the right way to do it. We’re doing lots of landers, lots of rovers to execute on what we refer to as the science of survival. You should all love that because the the alternative approach which we were doing more in recent history is the just build the giant glass dome with the nuclear power plant next to it. It’s going to go to

4:41:30 one prime contractor with a hundred subcontractors with dozens of international partners all making their contributions. Then it just gets canceled after we throw away $20 billion or something. So we have invited industry, we’ve invited Artemis Accord partners to say, “Look, we we’re buying these landers and rovers to learn. Yes, we’d like to learn about what we should ultimately lock in for power, communication, surface mobility, logistics. Have at it. There is going to be a lot of payload mass and volume available in these. And we have lots of different avenues to interact. I would highly advise talking to Carlos uh Galan. He’s our moonbased viceroy. But

4:42:02 we have uh we have uh Cber and um Nextep and a lot of programs out there. We’ve been dropping RFIs like crazy on who’s got good ideas that you want to test out on lunar surface. Haven’t been there in a long time. That is a great way to mature uh technology versus again thinking in in this context it’s going to be like a prime dominated event. It’s not it’s going to for the next six years on the moon base there’s going to be a lot of opportunities to prove out technology of which to not to get ahead. I would highly steer you towards robotics and I know it’s an area of great interest to yours. The astronaut bouncing around on the moon thing is

4:42:32 very cool. It’s very inspirational. Believe me, it’s like the last choice. uh you know, you you want to send them outside when something broke to fix it. Uh every time we put astronauts out in that extremely harsh, dangerous, demanding environment, uh you’re taking risk and uh and risk with not a lot of uh rescue options on the lunar surface like you have on the International Space Station. So point being is we are going to need to dial in robotics. If the moon is our technical proving ground to go to Mars, we’re going to get really really good at robotics and lunar surface. It’s an air it’s a steer I’d give you if you’re looking at uh surface payload uh

4:43:04 opportunities. So that brings me to my next question. So which space programs make sense for NASA to buy commercially and which require the agency to take a different role? >> Yeah. So it’s somewhat of a actually I don’t even know if it’s an unpopular position. I assumed it would be. Um, a lot of people thought I would come into this because frankly I’m only in this position right now because I had the opportunity to go to space twice through commercial space industry and as such I’m I’m a a champion for commercial space. But I also acknowledge some of

4:43:34 the limitations associated with it which is not every problem can be solved by just slapping commercial on it and assuming that’s a solution. So space is really hard and the economy associated with is very immature. What I would say is I am uh I think NASA, I think DO, I think industry would naturally gravitate uh towards buying services, you know, commercial services when it’s very clear you are one customer of many and you are not propping up the entire industry as a as a sole customer because in that respect really you’re just acknowledging that we don’t have enough of a budget to do it. The company you’re buying the

4:44:04 service from doesn’t have enough of budget to do it. So they’re going to borrow private capital to finance it. And then NASA which we should have quite literally like lowest cost of capital is paying you know 20% hurdle rates or something launch observation and communication and all the associated agencies there’s markets there for it. So what do I mean by adjacencies? When we say launch observation communications I mean you know you’re you’re you’re doing a variety of sensing satellites that also serve DOWRO agriculture weather space weather. Great. I’m I’m very sure that you can continue to build out those platforms and instruments and

4:44:35 have a service model associated with it. juice it up extra with uh super intelligence makes lots of sense. Uh launch, do I would I put transfer stages in that? Of course, like that, you know, we’re going to have our our trans moon railroad of of sorts with transfer stages bringing payloads back and forth. Like I believe that will be a useful service. The more you get into low TRL stuff, the riskier it is to say that should be a commercial service. Um you know, we took some hits early on with moonwalking as a service, uh space suits on that. was no no fault of the company or what it just we’re the only

4:45:06 foreseeable customer for that for some time and NASA also didn’t play its part by actually helping drive an outcome. We got convinced at some point if it’s a service just sit on your hands and everything will be delivered. So we’re course correcting there but if it’s low TRL you probably want some sort of a cost plus to mature to high TRL and then you offramp to fixed firm price and if you’re truly one customer of many it probably should be a service-based model. >> So at Spacium we work on inspace refilling. So what would make NASA a buy of refueling as a service rather than sponsor of another demonstration?

4:45:38 >> Yeah. So it it uh it’s a it’s a good question and this is where I would say we’re probably not one of your your best customers today for that. I think DOW is going to use prop all the time. Uh I mean it’s a that we we’ve got we’ve got dog fighting in space these days. So, uh, they’re going to use a prop and it would probably make a far more sense to go and refuel it, service it, enhance its capabilities than just expending more dollars on another on a, uh, another, um, you know, another orbiting asset of sorts. See, with NASA, you

4:46:09 know, we nail our trajectory and and send Nancy Grace Roman to L2, uh, we’ve got 23 years of of fuel on it now. Uh, because we’re just steping. like we really don’t want to move a whole lot uh when we’re collecting imagery of the of the universe like that. So, uh what we need to do is take advantage recognize your capabilities. It’s in the same way as say like we would be crazy to build a new telescope today that didn’t take in account the fairing size of a new Glenn or a starship at some point. Like we’d be crazy not do that. As industry capabilities are mature and it’s like we

4:46:40 can count on on onorbit uh servicing. So whether that’s refueling or even on orbit assembly, which for sure, right? We’re never going to try and build the whole damn thing on one rocket anymore. So that robotic assembly is going to be is going to be critical. But um we just need to build vehicles that take better advantage of it. I want us to commission and the president’s very supportive a vision powered uh starfleet if you will where we are sending fision powered spacecraft to the outer solar system to Mars and back. I mean you could do every two-year type missions on this. You’re going to have to top off xenon and you’re going to have you’re going to

4:47:11 probably chemically augment it. So you’re going to top off your hypergalis before you send it back out. So we just have to you know start our capabilities are on our roadmap need to measure up with what your your capabilities are that are maturing but I I I think probably DO is gonna be a better first customer. So because you’re talking about capabilities, so in the future for NASA to be a customer, um is NASA looking for more proven reliability from like DOW or like compatible interfaces because there’s no standard interface in the like currently in the market or like

4:47:41 a customer committed like the DOW. Is that where the NASA is looking for? >> And you’re referring to spacebased refueling in that regard. >> Uh I suspect that you know the DOW will establish the standards. We do have a lot of um you know presidential executive orders that uh try and ensure there’s uh dialogue and cooperation for all purposes. The National Space Council is uh is chaired by um director Katzios from OSTP that kind of brings us all together in a room. We do a space summit with uh SECAF and and NRO and chief of

4:48:14 space operations to talk about these things. I suspect they’re going to be the bigger user up front. So we’ll probably you know we’ll ask for our input and we’ll help support it. Um but uh the inverse is probably we will I I suspect establish the standard with industry as we mature uh and demonstrate technologies on the lunar surface of this is what you know surface water interfaces will look like or this is what surface prop transfer interfaces or power interfaces will look like for rovers and other things. Can you also tell me where do NASA and defense

4:48:45 customers have genuinely shared requirements and purchasing plans rather than simply overlapping interests and how could a startup find those opportunities? >> Yeah, I I mean our standards are are are across all of infrastructure. So, uh from our stage zero launch complexes here, we need we need the same we we need lots of launch pads and access. Uh we need prop transfer lines that go out there. We care about launch vehicles uh and lowcost ability to accelerate mass to orbit. We care about transfer vehicles the same. We generally use a

4:49:16 lot of the same uh you know storeable propellants and cryo propellants. So there’s a lot of things that are applicable right up until the point where our missions go in different directions. But even then you you you can actually have a lot of overlap in communications. Look the the the ch uh chairman of the joint chief of staff said that this is uh from the surface of our oceans to cis lunar space is a warfighting domain. So we care a lot about oop sorry we care an awful lot about cis lunar space because uh we’re building a moon base. So we’re going to need communication uh comm re surface comms comm relay observation

4:49:47 capabilities to support our moon base out there and they’re going to need it because they’re going to need space domain awareness which is also beneficial to us in that. So every step of the way all the way out there we’re a common customer. Uh we might even be on the same bus. Uh you know what they’re doing for space domain awareness to look out for us in cis lunar space might be also being a communication relay. You don’t know. So there’s a lot of opportunities there for right up until the mission diverges. >> What can you share about the next steps for commercial low earth orbit stations and the role NASA expects to play in

4:50:17 that? >> Well, I would just say that I uh I I absolutely value our presence in low earth orbit. Uh so I think that there may come a day first of all it’s an it’s an unbelievable achievement. I’m sure many of you know this. I mean but for more than 25 you know years there hasn’t been a time that you didn’t have astronauts orbiting above you on ISS. an amazing accomplishment, great laboratory, taught us how to keep people alive in a very demanding environment. And now we’re going to extend that. We’re going to extend that to the to the lunar surface itself. But until that lunar base is operational or until such time as we have exhausted everything we

4:50:48 can learn from the unique environment of microgravity, we will have a presence in LEO. And what’s important is not to accept a gap because we have a problem doing that. We’ve done it before when we gave up the shuttle and had a dependency on the Russians for 10 years. And we’re never going to allow that to to happen again. and we’re not always on time or on budget. So, we want to be very careful in this transition. And part of that is making sure that if it is a model that is dependent on NASA being one customer of many that other customers can get there. And that is why we prioritized our discussions earlier this week with Starlininer. It is not a

4:51:20 secret. SpaceX has said it many times. They will, you know, Sunset Falcon and Dragon, which is going to give a great opportunity, by the way, to the rest of the launch industry. Uh but that means we need to have continuity to lower Earth orbit. It’s why we’re uh we have a get well plan with Starlininer. It’s why we’re extending additional missions. It’s why we’re going to help crew rate it over to Vulcan as a spacecraft so that astronauts can get to the ISS and commercial astronauts can get to future commercial space stations. Without transportation, that model’s dead on arrival, right? So, um we got to we got to work that.

4:51:50 >> What should a small robotics payload or a servicing company build into its plan now so they can support the future commercial lower Earth orbit stations? Yeah. And and maybe just uh one extension of that since this is a room of problem solvers, right? And I’m expressing uh the problem you find find the human use case human in the loop. I’m an astron. I’ve been to space twice and I love it. And I want there to be a lot of reasons why we need people living and working in space and on the moon and farther out there. That is going to be pretty paramount if you want lots of commercial space stations or commercial

4:52:22 hotels because we’ll do our best to try and drive cost down, but that’s largely going to be a factor of demand and uh it’s going to be expensive to put humans in space. It has had it has come down from shuttle so much so that it has open space to others including myself in that but it has not come down enough where you’re going to have you know a serious order book. So help solve that problem. What is what is the use case for a human in space uh necessitates their presence there and uh and that will be very welcome to the CLLD community. I promise you that and the and the crude

4:52:52 transportation. And then in terms of just robotics, like I I don’t care whether you’re on the lunar surface itself or in low Earth orbit, Mio Ho, if you’re playing around with robotics in space, there’s going there’s going to be demand. There’s going to be national security demands for it. There’s going to be on orbit assembly demand for it. Uh all of the institute resource manufacturing is the operations themselves on the lunar surface. Step one, it’s got to be robotically driven. you got to only bring the human in uh when something broke down or it’s something extremely unique because you’re just you’re you’re taking a lot

4:53:23 of risk every time you put a human outside whether it’s in uh microgravity or on the lunar surface. >> Uh how is industry feedback shaping which infrastructure NASA provides and which services it buys? >> Oh, I I would say it’s a it’s extremely informative. You know, we uh we had our ignition day earlier this year in February where this was basically the formalization of the president’s national space policy. This is president says this is what I want you to achieve and this and we had ignition day to explain how we were going to do it and

4:53:53 get everybody aligned uh around it. But we dropped I don’t know two dozen RFIs, a couple RFPs on it like all right tell us what we got wrong. We’ll course correct and tell us what we got right. We’ll we’ll try and put a little bit more energy behind it. But yeah, I think a lot of the innovations that are coming out of industry uh are very informative to our architecture. Like we’re not going to build multi-billion dollar bespoke comm relay satellites for the moon. I promise you that. Like this is a platinum age for on orbit uh comms. We’re going to leverage industry capabilities there. Um you know, if we

4:54:23 ever again build the next giant telescope, the follow on habitable worlds observatory as we call it, it’s going to be sized for a fairing of what we think is going to be extremely reliable 10 years from now. whether that’s Starship or some you know even enhanced version of of new Glenn. So we’re very responsive to the progress industry is making. We’re also communicating our intent and I would say again please please listen to it too because there was a time period where NASA was had all these external impositions of everybody’s good idea that they get a you know a congress me member to force upon the agency and

4:54:55 spread our resources everywhere and those days are those days are over. Those are what you do when you don’t have a space race. Let’s try and make uh everyone happy. We have a real space race right now and the implications are very real. So we’re very focused on things that matter. It’s why I started. We’re going to the moon. We’re building a moon base there. It’s a great it’s a great area to mature your technology. Um uh so we’re trying to send demand signal of what we care about that’s out there. So people are generally trying to solve problems that are in in that direction. Can you also share a little bit about what is the critical path for Artimis 3 today and which paths depend on the

4:55:27 government decisions versus commercial providers and is there a meaningful role for new suppliers on that path? >> I mean all you know critical path runs through the landers. Uh but I will just say that it was not nothing else was a slam dunk. I say that because there was a uh you know a lot of pe a lot of people tend to beat up on our our commercial landing providers blue and SpaceX and their their readiness in terms of returning to the moon. Let me just tell you having you know been in the job now since uh mid December extremely focused on returning

4:55:57 astronauts to the moon every step along the way has challenges. We just had to do you know substantial pivot and I’ll tell you in every step along the way it was designed for the dream state jumping right to the most ideal outcome. Let’s fly around the moon on Artemis 2, wait three years, and then land on the moon. What are we talking about? Nobody I’ve used the analogy all the time, but nobody plays golf well if you play every three years. It’s terrible idea. Our suits, I mean, the the the design for the initial suits, like I mean, it’d be great if the suit held up throughout the the requirements, but the astronaut

4:56:27 inside of it would have probably had a stroke long before it was basically running a marathon every single day in the suit. Like, hey, it’s hard to get back to the moon. Let’s get to the moon. Let’s go out. Let’s get some rocks. Let’s learn. Then we’ll go out and we’ll make it a little longer the next day.$2 two billion dollar like uh you know uh transformer robot for the service uh you know people be pretty jazzed just seeing the Apollo rover going and then build on that get back to doing things in kind of an iterative motion so let me just say that throughout the whole architecture we’ve made changes standardize try and get back into a rhythm of going to the moon learn improvements into subsequent

4:56:59 models the the the landers are still an area of great interest we’re putting a lot of resources driving that outcome but every step along the way had to be pulled in standardized to get back into the uh you know kind of the right evolutionary motion, not turn everything into a work of art and such. And we’re we’re getting there. Like we will launch in 27. I put out a memo to the workforce 55,000. So I knew it was going to get leaked, which is fine. That basically said we didn’t have an Artemis 3 8 months ago. We had no you know we had half the hardware that we needed for it. We had no astronauts in training for it. Astronauts are in training now. Hardware is converging. We’re 90 days broken to

4:57:31 schedule on a mission that didn’t exist six months ago with lots of opportunity to improve. That’s a confidence inspiring message to be clear and I probably lands with a lot of you guys. Um and we’ll get that back and then uh 2028 we’re going to still protect for up to two shots on the surface. >> We have time for one last question. You have been a commercial customer and an operating space. So what did that teach you to look for uh for a new supplier that is easy to miss from a procurement office? Um what did what did uh that well look a

4:58:01 couple things just first the uh having that unique perspective in space and I know that doesn’t go directly to your question just reinforced why the hell we have to do it I mean what a great adventure that we’re on out there and really within our time frame right now to return to the moon build a moon base commission a fleet of vision power spacecraft go and explore you know Saturn’s moon of uh of Enceladus Europa may you know we know there’s complex chemistry there you know be able to have a path to seeing astronauts on Mars in our lifetime. I mean, this is all just so exciting. That’s what going to space,

4:58:31 I mean, really did for me. It was just an appreciation of how small we are in the grand scheme of things. And it’s our duty, it’s our obligation and destiny to get out there and unlock the secrets of the universe. But where it was maybe applicable on the procurement, the the ingredients that need to come together to see it happen is all the development work that goes into it. I mean, three years of development before we launched Polaris, seeing a suit go from an IBA suit to something that was qualified to go out in the, you know, the vacuum of space and see all the parties that have to come together to solve those kind of really complex problems. Everything that comes together to, you know, again, all

4:59:02 the ingredients uh that are necessary to do the near impossible and that’s what we’re that’s what we’re doing now. And we’ve put a lot of acquisition reform directives in uh where it makes sense, you know, leverage every step along the way. I described before low TRL here’s the right approach high TRL here’s the right approach you know is there a high probability NASA will be one customer of many help industry mature or uh procure as a service but in all cases embed the talent that exists within NASA across the supply chain to drive outcomes the

4:59:33 kind of outcomes that everybody’s excited to see and NASA’s responding well I think industry is responding really well too >> um that’s all the time we had administrator thank you so much for being here >> thank you thanks >> [applause] [music]

5:00:12 >> Heat. >> [music]

5:00:52 >> Hey folks, uh my name is Andrew Peterson uh from a company Array Labs. Uh here to talk to you about what we’re building in uh space-based radar systems. Um thank you all for for joining. Uh YC is an amazing place. we go to talk about the future and where things are going. Uh there’s been a number of revolutions that YC has been a part of and maybe space is uh is part of that transformation. So if we think about space and where space has been and and where space is going uh space has traditionally been

5:01:24 kind of a strategic asset, right? We have satellites in space. This takes images of of various locations around the world periodically episodically uh to support IC uh as a as a primary user base. The quality metric here is the highest resolution image you can possibly collect. And this is where systems have been built. Very very large systems, very exquisite systems, very capable systems. But this is no longer going to be the uh the case going forward or or there’s going to be an augment to this as space becomes more

5:01:55 tactically relevant. uh we’re going to see systems where uh we need to have uh much uh higher frequency, much higher uh operation to support uh real-time approaches. So, this will require greater performance than what we have right now. We’re going to need more satellites. We’re going to need persistent coverage. It’s not good enough to have something uh every couple hours or every couple days. You need to have something within seconds. Otherwise, it’s not particularly useful. Or if you can have something within seconds, it’s much much more useful. uh we’d like this to be

5:02:26 multi-m missission all domain uh and real time and this is something that we’ve not really uh seen before. Now there are some uh tailwinds that are helping us along in terms of this uh transition. We have cheaper access to launch. Uh we have more mature space infrastructure. We have new satellite buses that are much lower cost and easier to manufacture. Uh we have uh you know communications networks that can deliver data in uh much faster ways than ever before. And this is all helpful.

5:02:56 Uh, but we’re still missing kind of a a mission or or radar option here to support these mission areas. Radar is important. We need to be able to see things through clouds. We need to operate at night. We need to operate in all of these resilient conditions. This is hard to do with an optical system that can’t see through clouds and and can’t see at night. So, we need radar, but current options are very expensive. uh they’re built with this exquisite performance and exquisite price tag uh and uh and and not as manufacturable as we’d like. Uh so what are we going to do

5:03:29 about this? Uh at Array, we’re building the first uh scalable economic multi-m missission radar system uh that exists. This can be used for a variety of different mission sets including imagery, but also moving object detection and tracking to support uh maritime domain awareness, air support uh and all of these other missions. We have 12 contracts to date uh with various members of the US armed services including uh space force uh uh air force

5:04:00 uh DARPA and uh on a variety of different mission areas uh supporting air domain awareness, maritime awareness, uh ground imaging and all these other uh mission areas. So what does our performance unlock and and what are the uh benefits for our system compared to traditional systems and that that we’re launching in the near future. You can see here uh some traditional systems on the left side uh imaging at 15 meter resolution uh 3

5:04:32 meter resolution. These are systems that currently exist. Uh our approach and lower cost higher power uh systems can produce much much greater coverage. This is a single collection area, a single collection pass wide uh you know area collection modes with a four satellite uh cluster and a 16 satellite cluster. And here’s our AMTI uh GMTI and MTI performance for these systems. Um so

5:05:02 thank you so much for uh coming out here and and looking at the future uh with some of us and uh I’ll give you the your uh next speaker here. Thanks everyone.

5:05:40 All right. It is my honor to welcome uh Congressman Pat Ryan. Thank you so much for being here. >> You know, we’ve had uh we’ve had senators here today, secretaries, NASA administrators, but I’m glad the people’s house is >> you’re stuck with a lowly >> finally we have some representation here from the people. No. Um but thank seriously, thanks so much for being here. >> Good to be with you all. >> Um maybe just to get us started, folks in the room might not be all as familiar with your background. you get to Congress, of all the things you could

5:06:10 focus on, why was defense modernization the thing you chose or one of the things? >> Uh, well, I see some some army folks in the audience. I’m a former Army intelligence officer and uh I went went to West Point and when I got out of the army, um, I really had no idea what I wanted to do, but I knew I had a lot of friends that were still deployed in Iraq and Afghanistan at the time. Um, I knew frankly the garbage tech that we were given when I was deployed and I knew that I wanted to be part of trying to make that better. So I ended up co-founding a company with some fellow

5:06:41 veterans and we were really like a services implementation firm for for Palunteer and other data analytics software in their early days. Um, and so that was incredibly rewarding and also incredibly scarring uh to go through the battles of for the for the nerds in the room, the DIGs a battles uh and and all these kind of fights to actually get the best commercial tech to war fighters. Um, and so we’ve come a really long way. We obviously have a lot more to go, but

5:07:11 to to get to now sort of be in Congress trying to write better laws to fix that um is pretty cool. and uh kind of a cool full full circle moment. So, I’ve really dove in on uh the Armed Services Committee work and my colleague and I, Rob Whitman, who’s from Virginia, a lot of you probably know him. Uh we co-started a bipartisan group called the Defense Modernization Caucus and it’s um House and there’s now a Senate compatriate Dems and Republicans um authorizers and appropriators all

5:07:42 working together to get critical mass behind everything that you all are doing and try to really put fuel on to the to the sort of innovation ecosystem and also where needed to yell at people that aren’t doing what needs to be done. Um, so we want to be of service to to you all, whether you’re mission mission owners or builders. Um, and uh, yeah, it’s one of the things that gives me joy in this job, in this moment of absolutely crazy politics, to get to do this work. You know, this conversation is convincing me that we need more founders in Congress who have

5:08:13 experienced what it’s like on the other side and then, you know, get into places of power where they can actually help fix things. And and >> is the house a place of power? I don’t know. Is that >> Well, actually, yeah. you know, um, when you’re talking about all the other groups, >> I think the House, a lot of the founders I work with, the their representatives are the first people who are willing to take their meetings or help them, you know, start to navigate uh Washington DC. So, yeah, I think it’s very important. Um, okay. Well, you mentioned the Defense Modernization Caucus. This is bipartisan. I think you’re one of the

5:08:45 only Congress people I’ve been able to find on record supporting some of the recent, you know, innovation efforts in the Department of War and um some of the announcements there. Can you just break it down for folks in the room? Like where’s their where’s their bipartisan agreement on, okay, this is stuff we need to fix. Let’s go do it. And where is it still kind of not as bipartisan or, you know? >> Yeah. Yeah. I think I am the token Democrat here today, so I will own that. Uh, please don’t throw anything at me. Um, I think none of this should be this

5:09:16 is as patriotic as it gets. It shouldn’t be partisan. So, I think it’s important to be here. Um, >> yeah. I mean, the good news is there’s a lot the country’s divided on. This is an area where everybody’s aligned. The administration, the Congress, um, both branches, the American people, obviously, the people doing the work on all sides. And it’s so different from what I experienced, you know, 1015 years ago where we were just like battling in every direction against the old the old forces. So, um I’m very optimistic about

5:09:47 the space to see uh this administration come in and very aggressively, you know, all the right kind of focus and prioritization on speed, on urgency, on breaking things where they’re not working and taking big swings like that should be championed by everybody. Um I think the the concern points or question points I really have two dimensions. One is like love the briefings, love the memos, love the media coverage, but is it happening? Like are founders and

5:10:18 builders actually feeling this flow to them? Are tactical and operational mission owners actually seeing the right stuff get through? And I’m he I continue to hear like pretty mixed reviews from folks that while senior leaders might get it, the sort of bureaucratic middle continues to just like kill us um when when we know the speed and urgency of what we need to do. So figuring out how to address that um is important, but but having the right focus obviously and the right alignment is great and I’m

5:10:48 encouraged by that. Um, you know, the other thing that I think I I frankly I have to say is like I’m I’m I’m worried from some of the reports I’m hearing from folks in industry that all the good could be diluted by some of the trust concerns that I continue to hear come up where, you know, a lot more questions of, well, is the best company necessarily getting a contract or is it >> political? Is it who you know or you know like are people coming on to cap

5:11:19 tables that we might not have expected to be on cap tables? Like that’s unprecedented and it’s deeply wrong potentially illegal. And I just want to be clear like that shouldn’t be a partisan thing. That should again be an American thing that if any corruption or greed or self gain gets in the way of our war fighters getting what they need, that is going to be there’s going to be oversight. There’s going to be investigations. but only in in service of getting this right and honoring that I think very sacred commitment that that

5:11:50 I know everybody feels. Um so so doing that the right way I think is going to be a big challenge for the next uh you know two years if if things go the way we expect in 34 days. >> Yeah. Yeah. You know it’s funny you mentioned is it actually happening? Like is the stuff uh do the people on the other side feel like it’s moving faster beyond the announcements? >> I saw a lot of head nods on that. >> Yeah. Yeah. It it does feel like it’s improving from what it’s improving. But uh it’s funny. I was speaking to someone in the Pentagon. They said, “We don’t just buy from primes anymore. We also buy from Andre.”

5:12:21 >> And you know, for a lot of us in Silicon Valley, Andre might as well be, you know, the the fifth prime. Like, you know, there’s there’s uh a lot beyond that in in the world of innovative startups. And uh I think there may be more work to be done. >> Well, look, I mean, the numbers don’t lie. like 1% of the defense discretionary budget goes to ventureback non-traditional companies to you all. That’s just insane and counter to again everything that we all say we’re we’re doing here and that everybody says. Um, and so

5:12:53 the the fact that you look at this year’s proposed budget and all of the old expensive overengineered stuff is in the base budget and very little of the new disruptive commercial truly commercial innovative stuff is in the base budget. It’s in reconciliation land which sorry to burst people’s bubble but it’s not happening. And so I I don’t know if the comp controller talked about this earlier, but how are we going to make sure that we actually move the good disruptive lowcost attraable munitions

5:13:25 and interceptors for example from never neverland to the budget so it actually gets done and you guys can actually do the things you need to do across all these dimensions. That’s a constructive opportunity. But right now that you know I don’t I I’m concerned of where that’s going to land. uh for example. >> So how do you separate the noise from the signal? I mean it what do you look at to what would convince you or what could you see that would convince you like okay you know we are trending in the right direction on this outside of the announcements and the and the press.

5:13:57 >> Yeah. No I mean again like each of these gatherings all that does matter. I I I’m again net very optimistic that we’re going in the right direction. you know that indicator like if I were to look at KPIs like what is a percentage of our budget actually going to the places we say need to be our priority let’s get that from one to 10 20 30 you know let’s look at that five primes like that’s crazy that should be a that’s an obvious KPI like I’m not

5:14:27 saying we need to kill them off we could have that debate over drinks um but we need more actual competition crazy idea in a capitalist system we want actual full competition. So that’s that’s a KPI. Um and then I think like this is where I would invite um totally, you know, off the record like hearing from you all the exact situation you’re in. What is going well, what isn’t. We set up this Defense Modernization Caucus and we do meetings, forums, we do happy hours. We do. The goal is to hear un

5:14:59 unvarnished ground truth and then let us help you either solve that painoint or at least factor that into writing better legislation. Like for example, I’m going to wonk out super hard here, but um this year we’re able to get into the base NDAA language that lets uh a PAE take up to 10% of RDT&E money and move it to production, which is a big deal. like it helps us move the right things forward without having to get stuck in all these other sort of process wickets. So those

5:15:30 are things that the the caucus worked in a bipartisan way to get into the bill and we need more ideas like that from you all of what is a pain point how how might we fix it. >> Yeah. All right everyone hear that? That’s a very generous offer. Uh Congressman Pat Ryan wants to hear from you. He wants to hear where things are getting stuck. He wants to hear where the processes >> we take every meeting we possibly can in our office with with great companies. I mean, we really invite you. This is again, I think it’s our patriotic duty to do this, right? And so, Olivia from my team, who’s now going to like kill me

5:16:00 uh in the back, can help uh help set that up, too. Sorry, Olivia. [laughter] So, this is really interesting because you’ve been on both sides. You’ve been the the founder trying to sell to the government and now you’re on the other side trying to fix this. Is there anything now that you’re kind of on the other side, I won’t use position of power, but but uh being on that other side that’s really surprised you about seeing how it actually works. Um it’s a great question.

5:16:32 I mean I I don’t know if this has surprised me, but what I continue to come back to is you know we we’ve made all these changes. We’ve given so many new authorities. We’ve created entire new frameworks, the OTAA framework that didn’t even exist when I was originally doing this. Like, we’ve created all this stuff, but we continue to come back to culture, culture, culture, and mindset as the reason that we either succeed or we don’t. So, I’m not really necessarily surprised by that. But I do think we often, you know, overly focus on we have

5:17:05 to get the process right, obviously, we got to get the funding right and the money. But if the culture continues to be what it is and you know the right things are being said at the top but not making its way um to sort of the people that actually need to action it that’s a I would particularly welcome ideas on how to affect that. Like one of the things we’re talking about is often people come in front of Congress for not good reasons, which only ups the risk aversion, right? What what could we do to highlight the people that are doing the right things that battle like took

5:17:35 risk, succeeded, and let’s hold that up as a a positive thing for the American people to see and hopefully try to change the culture to incentivize more risk-taking by by all parties involved. Yeah, >> I think that’s like >> uh if if I I’m not exactly answering your question, but if we can get closer to that culture piece, then then I think we’re really in a in a good direction. >> Yeah. Yeah. Um, one of the sources of friction I have heard from founders is around just literally testing the things they’re building and like finding test

5:18:06 sites. I think a I’m told a large fraction of the defense tech startups in Texas all test at like this one guy’s ranch. Um but like that’s kind of the state-of-the-art for how you can actually test things that are you know um very regulated like could we be doing better here? Is that one of the areas you guys are looking into? >> Yeah, definitely. And again would invite like specific level detailed conversations if you all have ideas or creative solutions here. I don’t think

5:18:37 we focus on this enough. It’s it’s a great uh point. It becomes a huge bandwidth constraint. Um like I know uh my my parochial pitch is that uh one of my classmates is here, one of my West Point classmates, Tim. Uh at our old alma mater, we’re actually trying to expand the footprint at West Point to think of itself as a proving and testing grounds. There’s a whole plan that we’ve been fighting to get money and resources to to help build um at at Buckner actually. Uh imagine like cadets testing stuff in their summer training and

5:19:08 giving uh real-time feedback and and then researchers around that. So just being much more creative. I think that’s something we could look at legislatively is how do we give more authority and and resources. Um, I also have like an old Air National Guard base in my district, for example, that is used for almost nothing right now. And we’ve been trying to push for drone testing, for counter UAS, um, for, um, for other things. So, um, just leveraging the fullness of the country and each of our districts, I think, is something we should all take,

5:19:40 uh, as homework. But again, I would welcome any any that’s something I think we could tangibly drive in like next year’s bill, for example, if people have ideas on that. >> Yeah. Yeah. Okay. Um, so you’re saying we might not just all be relying on this one guy’s ranch in Texas for uh >> I gotta meet this guy, but uh [clears throat] I mean I’m just saying there’s a lot of space in upstate New York. That would be great. >> Which is my district. And as an example of success, um there is like a there’s been a 10 to 15 year plan in in central

5:20:10 New York around Syracuse area to build uh some some uh testing and um um evaluation capacity for um air in general, particularly UAS and counter UAS. Um but that took like 10 years to to build and it’s still got a ways to go. So we we need to speed up. >> Yeah. So think about the defense modernization caucus and kind of your baby. Do you is there a vision of like end state what you think you know a

5:20:42 founder comes up with an idea and has a prototype of something that could really you know they think could really help America. Do you have a vision of like what what should their experience be like what is the endstate ideal startup government procurement flow look like? So, we should design government thinking about the people, our customers. >> I mean, let’s not get crazy here, but >> that’s [laughter] that’s crazy. Um, it’s a great question. I mean, so we were talking before we we came out and I think what you shared about your experience at YC is really interesting

5:21:12 to me. Built a company through it, wildly successful. Congratulations again and kudos on everything you built. And then you come back as a partner. Like, that’s a really cool virtuous circle. big credit to to you guys for that being so central to what you’ve built. How how do we do that in some way? Um I think that that’s a part of it. Um I continue to look at some of our adversaries, particularly PRC, and how they are able to understand um moving talent in and

5:21:43 out in a way that’s much more fluid. Obviously, we we are Americans and have different values, but I think there’s a way to get closer to the sort of civil military fusion that that they’ve had while still honoring our our values and understanding that um we need to radically think about recruiting. for example, like I’ve been pushing for several years to build a a cyber auxiliary uh force as a really a way to just get more cyber talent like identified and ready almost the way we have the civil air patrol from back in the day where

5:22:14 you just have people that might not probably put on a uniform ever, but we know them, we can count them, we can mobilize them if they need, we can put their skills to use. Um so so I think more of that um happening. But from like the the the company perspective what you’re directly asking um you know I think it’s obvious like as much as possible what’s the single point of information entry and delivery which we’re far from but getting better at I think DIU’s an example of doing this uh

5:22:45 well certainly not the only one um but less innovation theater more actual who are the people that will actually move stuff and do stuff for you and who won’t Um, and then obviously the biggest thing is like these are your companies. You need path to revenue that’s multi-year and that you can count on. So that’s why I like this idea of um understanding that for capabilities or even great companies that we’ve already identified. We should be working to map out like if you hit your metrics, what what are those next steps for you

5:23:16 moving from prototype to production to larger scale so that you can tell your board like hey it’s not guaranteed but if we’re doing our part we can continue to do what we do. We know we have as as you know the those responsible for the public good enough checks but we’re not leaving great companies hanging out in the wind um because we haven’t thought more than one year out. So, I think that’s all obvious, but those are some of the the things I think about. >> Yeah. Wonderful. Well, on behalf of YC and um >> want to do one question.

5:23:46 >> We have time for a question. >> One question. >> Sure. All right. >> Any >> any questions for the congressman? >> Harlo, you already got one earlier. Let’s see. >> Any any questions? >> Yeah. So, Commissioner Ryan, uh, a couple questions for you as far as, uh, you spoke a little bit about, I guess, what does it take for the companies to be successful, uh, to enter those spaces and to reduce the risk so they can actually, you know, finally deliver a product to the war fighter. And I think in those confines uh right having I

5:24:19 think those uh innovative uh labs like modeling and simulation to help reduce that risk give them an on-ramp >> uh and uh you know potential partnerships to not only take uh that company’s technology but marry up with other technologies that are out there at various TRL levels uh would get us to our end state uh faster deliver those capabilities. I wanted to hear your comments on that. Yeah, great point that I missed. Um, I think you’re absolutely right. We also think of these things in all these sort of silos and don’t think

5:24:52 about leveraging other investments we’ve made to then, you know, lower cost or speed speed delivery on others. And I think models and simulation is a great example of that. Um, it’s an area um that I know relatively or I know very little about all this as you can tell in my remarks here, but it’s an area I know less less about and so would love to work with you or anybody you think to to get get smarter on that. But that that makes a ton of sense to me. >> Wonderful. Well, Congressman, thank you for taking an interest in modernization

5:25:23 in figuring out how we bridge the gap with startups in making it bipartisan. Um, on behalf of all of YC and all the entrepreneurs, we really appreciate it. I am gonna I am gonna upset people and say um go Army, beat Air Force, beat Navy. Sorry guys. >> Thank you very much.

5:26:04 Hey folks, my name is Jay. I run a company called Railing. Let’s kind of talk about what I view as a crucial part of the next 250 years. I think the American manufacturing um ecosystem needs a billion more robots. Let’s talk about why. Right now, we’re at a point where I’m sure you guys have heard this a million times today. Uh the US manufacturing base can’t keep up with current production. Uh can’t keep up with current demands. The US Army needs 100,000 motors uh by October of last year. We currently from the last couple

5:26:35 of months ago can only do 36,000. Uh we’re at a point where our weapon stockpiles and you know our production across the country just can’t keep up with where the company the country needs us to be. Why is that? We’ve spoken a lot and you know the DoD loves to to go to the primes and give them big contracts, but the actual bottleneck isn’t the primes. It’s one abstraction layer under the primes. We’re talking about mom and pop manufacturing shots uh shops that said in Texas, Illinois, in Indiana who are building small fasteners that are currently bottlenecking all of

5:27:05 production. We are talking about classical automation within these shops and and one of the crucial things that we think a lot about is what is the current bottleneck in doing this manufacturing. The bottleneck is the fact that these these traditional automation mechanisms were not built for defense. They were built for automotive. They were built for other industries where the work doesn’t require as big of a throughput and as big of a ramp up over the last couple of years. Traditional automation is insanely expensive. you’re spending, you know, on average a couple of

5:27:36 millions of million dollars a machine, it takes an insane amount of time to ramp up, usually about a year and a half to two years. And every time you change the part ever so slightly, you have to go back and re-engineer the entire system from day zero. And so essentially the problem that we’re talking about is how can we allow the the DIB or these small mom and pop manaction shops to increase throughput without really increasing headcount to the rate where they they are, you know, completely underwater. And so effectively what’s what’s happening is that these small mom

5:28:06 and pop defense shops are at a point where Boeing is coming to them and demanding 30,000 50,000 parts a month and these guys have been manufacturing 5,000 parts and you know we’ll continue building that capacity. And so how do we allow these shops to transition from being at a point where they’re doing 5,000 parts a month to 50,000 parts a month? Um the way we think about this is that a lot of the underlying infrastructure required to uh to to do this innovation already exists, right? You can go on to Amazon and buy a state-of-the-art camera. You can buy vision systems. You can buy robotic

5:28:38 arms. The problem isn’t buying the raw materials. It’s being able to take that raw material, take that robotic arm and actually deploy that into a factory. Um, the way we think about our company is essentially being that deployment layer. So we as a company are essentially responsible for bringing up these massive robots which are able to help produce U-jints, help produce high tension ball bearings, springs, aircraft landing gear, pneumatic systems. Um we are thinking a lot about how can we actually increase this without an insane

5:29:09 ramp up and as a function of that we end up thinking about what are variables which the government demands that these shops change that they can’t they can’t currently change. We think about, you know, when a when a shop building high tension ball bearings now have to go from building, you know, ball bearings for the 737 to now the C5, right? Very, very different form factor. The underline, the design is the same. And so essentially for a traditional shop, they would have to go back and reverse engineer this system. The way we think about this is can we make these systems

5:29:41 completely flexible? Can we allow these small mom and pop manufacturers sitting in Texas or in Illinois or in Indiana to basically modify their production line at the speed of modifying software? And I think that is a very very powerful thing. Um let’s instead of talking about what what I do and what the company does, let’s talk about one of the shops we work with. Anthony has anony’s grandfather started this shop. Um they do pneumatic systems for aircrafts. Um, and these guys are they’re they’re underperforming

5:30:11 by, you know, 5x from where their their customers, the nottop Gmans, the Loheed Martins, the Boeings of the world want them to be. For for Anthony, manufacturing more parts means that he will have to take on, you know, $10 million of debt and invest money into a 18 to 24 month cycle to buy more machines. And you know hopefully the math closes the books close four five seven years and he will finally make an ROI on his on his on his investment. The way we think about this is how do we

5:30:42 allow Anthony to 10x production. This is an Anthony has worked with us for the last few months. Um he you know he’s at a point where right now deburring is done manually. So he has a CNC machine that’s running through the night. Uh but when a part comes off the CNC machine, it requires a manual technician to come and take a knife and fix all the small birds or all the small problems in the part. We’ve allowed Anthony with a few what $150,000 to effectively reduce that bottleneck. Anthony is now perform making 5x more parts than he was at the

5:31:13 starting of of of the year. Anthony is not a one-off situation. Anthony currently supports 29 plat platforms across six different classes. Um he currently has uh parts which fly with the army, uh with the navy, with the air force. Name your poison. Uh Anthony has an active machine out there. Um and Anthony is just not one of one. Anthony is one out of a million manufacturers in the US who just can’t keep up with production. Boeing wants more demand. Boeing wants more parts,

5:31:45 but Anthony is the actual reason why Boeing is not able to build more aircraft. why Loheed isn’t able to deliver um more aircrafts. So let’s talk about why these billion robots are important. The billion robots are important because we are actually going to be able to get tier 2, tier three suppliers, the guy who’s the guys who set one or two abstraction layers under the primes to actually be able to increase throughput to be able to build more parts, build more machinery and reshore back to America. The way everyone in this room can help, if you know factories, we’d love to talk to

5:32:16 them. Um we want to talk to more uh more of the shops who sit under the primes. We don’t you know the primes are great. They’ve gotten a lot of money. The guys who are manufacturing things for the last three generations across their family. They are the guys who need our help. Let’s go and help them. Thank you.

5:33:13 There we go. Hey guys, my name is Johnny. My company is called Proven Metal. And the short answer here is or you know to cut to the chase the who who here knows what a PCB is? Okay, a good amount. So PCB stands for printed circuit board. It is the single most critical bottleneck across aerospace defense and robotics applications. And the existing American supply chain is really really bad. And so we are on a mission to fix that. And it’s a really really tough problem to solve. So market

5:33:44 context, the US in the year 2000 manufactured 30% of the world’s circuit boards. Now that number is 4%. This is not good because it is really really important. And so there’s something weird happening in this industry where when you when you talk to us manufacturers, they say there is not enough demand. We are dramatically underutilized. And then you go to the customers, these hardware companies, and they say there’s too much demand and not enough supply. And so there’s something weird going on in this industry. And so when you break it down into what actually attributes the majority of the

5:34:15 lead time for these circuit boards, the vast majority of it is not manufacturing. It is everything going on around the manufacturing. This is quoting, designer review, uh sourcing of all the components and the kitting processes or preparation for all those components before they before they go into a assembled circuit board. And so what we found is that if we partner with these existing manufacturers and automate these processes, we can immediately cut leons in half through automation. And so the first thing that we’re doing is essentially automating the supply chain by by integrating with

5:34:46 existing manufacturers. You can think of this like a network where every manufacturer has a different level of price capability or turnaround time or speed. And so we have a huge amount of data on a a wide range of manufacturers in our within our network such that when customers give us an order, we can really really efficiently route those orders through existing manufacturers and also increase the utilization of idle American capacity. However, as you scale this routing layer, which we’re doing really, really fast right now, you start to identify where the bottlenecks

5:35:16 are in the industry, where the manufacturing itself needs to be vertically integrated. And so, you in this network diagram, you’ve you fab 2 and assembly 3, which needs to be vertically integrated. And so, what we do then is we vertically integrate. And the cool thing about this is that the biggest problem in manufacturing uh is not the manufacturing itself. It is the effective deployment of capital. And so by scaling this routing layer, we’re aggregating demand density and picking up data on where it makes most sense to vertically integrate. Um, and so from using this simple mechanism, this is how

5:35:47 you build the the biggest electronics manufacturing company in the US. I’m a manufacturing engineer. My co-founder will an electrical engineer. Uh, this is a problem that we’ve we’ve both lived and experienced a lot and we’re on a mission to fix this problem. Uh my one ask is that if you are building hardware or you are related to somebody building hardware uh tell them to to build their circuit boards with proven metal and fix the domestic supply chain. Thank you.

5:36:47 Okay. Good afternoon. I’m Steven, one of the co-founders of Greypoint Industries. We build drone swarms that hunt hidden enemies. This means the operators, the jammers, the command nodes, everything that drives modern-day drone warfare and communication systems. Ukrainian military units have signed over $30 million of LOIs for our system. And we already have a phase 2 R&D contract in Canada. The modern battlefield is filled with autonomous systems controlled by humans that we can’t see.

5:37:18 They’re under camouflage, hidden from cameras, and difficult for radar to spot. We keep deploying interceptor drones that cost a few hundred just for the enemy operator to pick up the next one and launch it right away. But most of these notes have one fatal weakness. They have to transmit. That is what Legion exploits. Legion is a swarm of lowcost autonomous drones that pass across the frequency spectrum and geoloccates targets in real time. If they fly a drone, if they key a radio,

5:37:49 we can find them. Uh, wrong. Each aircraft takes a bearing uh on an enemy signal. The swarm then crosses those bearings, geoloccates it, and pushes it up to TAC or any other battle management system like Lattice. This is all done fully on the edge, so we don’t emit anything that can be sensed by our enemies. And our mesh is designed to be difficult to detect. This means we can find enemies while operating in contested territories, but they can’t find us. Our sensors can see many things. First,

5:38:20 the drone operators located from their control uplink. Second, interference units, GPS jammers, and other electronic attack capabilities. Fourth, comms equipment, enemy personnel with handheld radios on the ground. And finally, air defense and counter battery radars, allowing us to integrate directly into suppression of enemy air defense tactics. Here’s how it works. Legion doesn’t need an electronic warfare specialist to operate. Any soldier can designate an area, launch as few as two aircraft, and

5:38:50 get live tracks of emitting targets. The Swarm makes it decisions autonomously based on real-time sensor data, allowing for persistent surveillance of hostile signals. Aircraft rotate in and out to recharge so coverage never drops. Finally, we classify and store those signals with our onboard AI to find exactly what that threat was. There are three main things that set Legion apart. First, we target the signal, not the airframe. That’s the cheapest way to take a drone team fully out of the fight. Secondly, it’s

5:39:20 passive. It keeps working when GPS is jammed and links are degraded. In fact, we can detect those jammers and other enemy EW equipment. Finally, it’s a leaderless swarm, so we can lose an aircraft at any time, and the swarm can heal itself. There’s no master known to be neutralized, and everything operates in real time on the edge. Legion finds and geollocates the emitter. Our second product, Hound, is the cuteector that can be deployed afterwards with a similar RF sensor. Mount it onto any small drone and give it the ability to autonomously fly

5:39:52 towards emitting targets. An operator authorizes the target and Hound flies itself towards the signal before handing off guidance to its camera for final approach. We want to saturate the field with thousands of these small inexpensive sensors mounted on everything. This means drones, UGVs, static entrenches, and existing infrastructure. We want to build a full map of the electromagnetic spectrum of the battlefield. This is one of our fixed nodes. In a recent military field test, we verified that we can find rogue drone operators

5:40:22 at airports and transport hubs, counter narcotics and counter trafficking for border security, and stop contraband drones flying over prisons. We can protect power plants, data centers, and any other critical infrastructure or sensit sensitive sites. Currently, we’ve proved the tech on the ground, flown it, and demonstrated at TRL 6 to the Canadian Armed Forces. In the next few months, we’ll be deploying to Ukraine our and our targeting to be at TRL 8 by the end of this year. Field testing will also give us real world data to keep improving our classification models.

5:40:54 Our team comes from an extensive defense background, including the Canadian infantry, general dynamics, and Bosch quantum sensing. We’ve already shipped products in defense, and we build our own radios, airframes, and autonomy all inhouse. Allied militaries have already committed. We were selected for a phase 2 Canadian R&D contract within 6 months of founding. And Ukrainian units, the people who are fighting this problem every day, have signed $32 million in letters of intent. We’re a US company with Canadian roots, NDAA compliance,

5:41:25 and eligible for OTAA contracts. What we learn in combat testing this fall goes into every system we will ever deliver. Now, let me show you how this works. In this room right now, there’s a drone operator. Has anyone seen him? We can. So, currently, as you see, um, we’re live in the Gallup right now, and my co-founder here is holding a drone controller, and we’re live tracking the unknown signal as he walks up to the front of the stage. Um,

5:41:56 yeah. Uh, we want to demo our tech, so if you’re interested, come find us at our booth outside. Thank you. [applause] Yeah. Heat. >> [music]

5:42:27 [music] >> I hope this shows

5:43:01 Um, okay. Let me see. I have an HDMI. >> Yeah, should be presenter view. >> Oh, is it set up as a track? >> Maybe it’s got nothing.

5:43:34 This is like a dongle or something or go to your settings. Is there another way?

5:44:07 Sorry, I think this is Sorry about this guys. Technical glitch.

5:44:56 Yes, I think that works. Do you want a mirror? Yes. Let’s do an entire screen. That’s this. >> Okay. Sorry about that. Um, well, hello everyone. My name is Sarth. I am the co-founder and CEO of 64. So, you know, we are heading into a world where AI agents and AI is going to make really critical decisions on people and

5:45:28 associated entities and vice versa. And at 64, we are building the intelligence infrastructure for this. So the core of our technology is what we call investigation agents. Um to answer what an investigation agent is, it’s like a research agent that works like your best analyst. So it combines open source intelligence, the dark web, clear web, tracking different pieces of information across the board, um, and coming up with

5:46:00 claims that are source backed and scored for really high confidence because we are playing in a space where you cannot afford to be wrong. So on the right is an example of a 64 investigation. So our agents actually have applications across the board. I can give you a few examples. supply chain due diligence like who are our contractors who actually own them. Alison networks who are the people at the forefront and um we’ve been really involved in DPRK scanning as well. So catching bad actors

5:46:32 who are trying to infiltrate our companies and corporations. Um and there’s defensive use cases like you know our personal like what information are they putting out there that could be potentially abused by adversaries um from public information. So these are all things that an analyst would take weeks but an agent can do within minutes and all of this is possible at a scale and at a price that was never possible before. Um so across the board entities and individuals actually look very

5:47:02 similar. They have digital footprints. They have um entity networks that across the board whether it is a hospice network or a defense supplier would look very similar. And because of you know we just launched the NDFC earlier uh like a couple months ago. So I think this is a very timely thing to talk about fraud because fraud has been a big federal priority this year. Um and one example in particular I’d like to go over is the hospice fraud example. This is something which was

5:47:33 really popular in in social on social media over the last uh few months. So let’s go to Monterey Park in LA County. Seven people were charged for from three hospices and uh for taking over almost two $3.2 million over eight years. Um and what they were doing is they were enrolling patients who weren’t dying and rotating them between three hospices every 6 months to avoid u Medicare’s review trigger. So on the right is an investigation that we ran with our

5:48:04 agents. So all of this was public for years. This thing began in 2016 when the head nurse arrived from a previously prosecuted fraud network. Um 2017 people started complaining about bill uh billings for visits that never actually happened. 2021 2022 the directors started suing each other. Um so there was a bunch of things that happened but in 2024 is when this came to um some attention and 2026 is when people were prosecuted for it. Um now this was all public information

5:48:37 that you could probably read in in in the case filings but what we found was um that there were seven hospices not three there were 1,900 hospice certifications that were signed by a pathologist and theospices involved here were top percent 1% when it comes to spend. So they were spending a lot of money but they were bottom 3% when it comes to care. So they were billing for care that was never happening. And all of this was traced by our agent. So the way it works is it will look at one

5:49:09 piece of evidence and then go deeper and deeper and deeper until it can connect the pieces of information together. Um and the best part about all these things that I brought up here is none of this is actually in the news or in the case filing. So this is all brand new information that you guys are seeing for the first time. Let’s take a step back and look at it as like a clearance file. A lot of the things here whether that is UBO or personal conduct or continuous vetting or people suing each other each other for financial pressure all of these are things that you know are required in like a clearance file for

5:49:39 example and self-reporting usually misses all of this information but public records and the internet and the dark web has this information at all times. So um that was a case that happened in the past. Now let’s go at like a real case that is happening today. I anonymized information just just to not make claims, but I ran our agents on an address in Semi Valley in California a couple days ago. And what we found was that four Medicare certified hospices which were each

5:50:11 presenting themselves as an independent well they all traced down to three clusters. Two of them shared the same phone number and email domain. Two of them were incorporated on the same day at adjacent suites and then one of them was dissolved in 2024 but is still an active federal provider. Now I’m not saying that any of this is fraud but that’s the whole point of 64. We will build a map that an analyst would spend weeks building within minutes and every source is uh like every claim is source

5:50:42 back and is scored for confidence. And at 64, our approach is you should know before the damage is done. So today, investigations across the board would happen when a tip comes in. But with 64, every supplier, every entity, every filing can be continuously checked for compliance and red flags um that might not show up in in just straight up public records. And um with this I would like to end and say um if you have a hard case that involves entities or individuals feel

5:51:13 free to reach out to me or if you know there’s a team that um might be interested in what we’re building we would love to chat. My email is uh on the bottom right. Thank you so much. This is hard from 64. [applause] 10-minute break and then we’re going to come back and fill the room for um an

5:51:47 amazing discussion between Taylor Budawich and Thomas who Thomas Salvo who’s the creator of Reaper and you’re going to want to see this 10-minute

6:05:50 Heat up [music] here. [music] Heat

6:06:21 up [music] here. >> [music] >> Heat.

6:06:53 Heat. [music] >> [music] >> All right. Uh, welcome back from the afternoon break and uh, I’m going to uh,

6:07:25 introduce our next fireside, but we’re going to kick it off with a video after I uh, give their bios. Uh, first I’m really excited uh, to host uh, Thomas Salvo, who is the aerospace builder behind Reaper while studying aerospace engineering at Georgia Tech. He’s designed, built, and flight tested jet powered remotec controlled aircraft, documenting the process publicly. Uh some of you have probably seen some of his viral tweets about the project. His engineering background includes propulsion, launch vehicle structures,

6:07:56 and autonomous aircraft. Um last but not least, uh is uh Taylor Budawitch who’s going to be leading this fireside. Taylor uh is the president of Sovereign Adviserss, a public affairs and strategic communications firm. He previously served as the White House Deputy Chief of Staff for President Trump, overseeing communications and personnel uh work across cabinet affairs, comm speech writing, and public liaison. Um we’re going to welcome them to the stage and and watch the Reaper video to kick

6:08:27 us off.

6:09:00 I see you. You’re really low. Yeah. Pretty

6:09:38 cool. Pretty cool. All right. So, most founders getting involved in aerospace maybe have a couple decades of experience at a prime, maybe a PhD. I get the sense that’s not your story. Well, I mean I my aerospace story is I’d say like 10 11 years long. I mean I started airspace when I was very very young. Um it was really like my thing. Um I grew up in Chicago in Singapore. And from a very young age my parents

6:10:09 they’d like get me remote control helicopters, planes, that type of stuff. And I was a stupid kid so I break them like really really fast. And the first step for me was like actually just fixing those planes and helicopters. Um, so I take the motor from one, put it in the other, and eventually I started building my own aircraft. So I would cut up like foam board uh from like the hobby stores, and I would ask my parents for just like motors, servos. I’m really grateful because they would always supply and just let me just kind of explore. Um, and the hobby evolved more and more and more, and now I’m in

6:10:40 defense. So, >> well, congratulations. So, tell us about this thing. How did Reaper come about? >> Yeah, so Reaper, I brought it over from Atlanta. um shipped it or put on the an airplane. Um so it’s to my knowledge, watch out for the wing tip. To my knowledge, it’s the world’s fastest uh like small air breathing RC plane. Um and it started in February. It was a while ago. Um I was in this lab with my friend and we found

6:11:12 this broken turbine. Uh it was about this big and it was owned by this club and it had been broken for like five years sitting on this shelf. So I ended up like reaching out to the club asked the club like hey like this this engine I I had I had an experience like working on these small engines uh from one of my internships and I said look I mean this engine looks okay maybe I can fix it like could I put on a plane and they were like yeah sure go for it like do do it like it’s super super cool idea. Um, so I spent like two weeks fixing this engine and it worked. So, um, then I had

6:11:44 this working engine. I designed the plane or this plane, um, and 3D printed this small like subscale demonstrator aircraft and flew it to 150 mph in Petemont Park in Atlanta. Um, and posted about on LinkedIn and it went viral. Uh, the club then saw the post that they, you know, didn’t like it and they uh, ended up taking my engine away. So they sat me down in this room with like it was this 15 foot long table um with a bunch of people and they were like oh we don’t like your project. Um we think that you know there’s a lot of

6:12:16 liability risk for the club and they ended up taking the engine away. Um so then I reached out to like a bunch of companies like Loheed Andreal. Um on the maiden flight I actually had an Andre sticker on the tail. Um and they ended up donating money to sponsor the project. So I um and the idea there was that like I had this audience on LinkedIn that I could advertise, you know, their companies and they wanted to hire people in aerospace. Uh so the the deal was that I’d put logos on the tail uh for anyone who sponsored and so I I had the money. I bought a new engine

6:12:47 bigger than the other one actually. Um and yeah, then it started and it was making the airplane. So you know, the the composite molds are just like handlaid carbon fiber. So, you know, I spent like 24 hours in front of a CNC mold or a CNC mold um milling out the molds and then I like the way you do it is you like you lay up you lay carbon fiber sheets, paint it with epoxy, and then vacuum bag it. Um it’s very manual, but you know, it worked. Um and then you like assemble the frame like Lego pieces. So, you just cut these like

6:13:18 carbon fiber ribs and just assemble them like Lego pieces. Um and yeah, it was a really really cool project. Um, and I did most of it from my dorm room. So, in college. >> So, it’s Reaper is the first drone influencer selling sponsorships to get built. >> Sure. Yeah. Reaper is the first drone influencer. Exactly. >> Um, so now what’s the goals for it now? >> You have proved that you can build it. >> Yes. >> You can you proved that you can fly it fast. Now what? >> Yeah. Yeah, I mean, so, um, yeah, I mean, the once I had the plane, it was

6:13:49 the I’d say the hardest process was actually finding a place to fly the airplane. Um, so I I spent, I’d like to say, like six hours on Google Maps just scrolling through farms in Georgia. Um, looking for a farm big enough that could, you know, I could fly this plane um, and not fly over people. Um, and I ended up in Arkansas. So there was a guy who uh had this like duck farm and he put me in touch with the guy in Georgia. So it was 3 hours down from my house in Atlanta to the farm. Um and

6:14:21 this was Saturday, like the weekend before finals week. So I had a full semester of exams and the weekend before I was like testing. Um on Saturday we we went down and the moment we had the launch rail set up the the clouds rolled in and it started raining. So we went back Sunday came back and again three hours there. Um, I realized that I had left the controller at home. So, I like called my friend desperately saying, “Look, dude, I know you’re studying for finals. I know this is a really big ask, but like could you meet me halfway?” Uh,

6:14:52 so we met halfway. I went down and we flew the plane on its maiden like 10 minutes before sunset. So, and that’s the video you posted and there’s this great video of you what, 10 years ago, flying your first plane and then flying this plane side by side. It’s pretty neat. Um, so offensive, defensive, what’s the what’s the goal, the purpose for this? How are you going to utilize it? >> Yeah, I mean, Reaper, the idea is that Reaper is like a initial like demonstrator for our flight pedigree. Um, it’s an aircraft that works. It

6:15:22 flies fast. I’ve shown that. Um, we’re we we’re going to use this airplane to test. You know, I mean, the army has the US government has great programs like Army has the fuse program. Um, it’s great for testing. It’s great for initial production. just we can make I can literally we can just do the the Lego pieces ourselves. Um testing in Ukraine as well. It’s a really really there’s a lot of opportunity there for for demonstrating the capability. Um but that’s as far as it goes. It’s just a demonstrator. >> So So you said that it flies fast. Now pretend like no one’s here. How fast has

6:15:53 it gone? >> I’m going to be really really careful with the wording here just in case there’s anyone from the FAA in the room. But the plane is capable of flying 500 miles an hour. >> But of course you would never do that without approval. >> No. No. Croissant. Good. >> Um, so now you’ve made a couple of them. You’re Where are you at in your pro process um of building this business? >> Yeah, I mean, so we’re we’ve we’re just testing this aircraft. Um, learning more about how it flies, um, getting the controls right, uh, the autonomy. Um the

6:16:27 next step is obviously to make we’re making a new aircraft that you know the entire airframe is completely designed from zero to be manufactured in mass. So tens of thousands of units a year. Um you know we’re looking for large carbon molds. We’re working with some suppliers and vendors. Um to find a way we think we found a a pretty decent idea or way to make the entire airframe in one mold. So instead of just making doing the Lego pieces that I made like doing laying the carbon fiber sheets by hand, you’re stamping these out in mass. So that’s a

6:16:57 completely different like it’s a game changer. And so next steps, what I mean you have this room full of people from different parts of uh the industry um different parts of government. What do you think you need to succeed? What’s the next thing? Is it is it capital? Is it talent? Um, what’s those next steps to make sure that you you succeed? >> Well, I mean, being here is great. I mean, I think we’ve been I’ve been very lucky. I mean, I’m a college kid. Like, it’s not every day that you got an opportunity to just like chase your dreams with with something you like

6:17:28 you’ve liked since you were a kid. Um, and I’m very grateful to YC to be giving me funding to explore this. Um, that’s, you know, part of it. Um, I think that’s very I’m very lucky for that. Um, and the next step I think is just demonstrating. So taking this product which flies it’s it it flies well uh and showing what it can do on the battlefield. So testing in Ukraine, testing in you know these extec fuse uh programs like the like DARPA and programs as well. So it’s it’s really really exciting. >> All right, last question. Uh under

6:17:59 $50,000. How fast do you think you can get a drone to fly? >> What would your goal be? >> Our goal um I mean the purpose of the company is speed at low cost or high speed, low cost, high volume. Um, we’re currently developing a supersonic aircraft as well. So, that would also completely change the game. Having a really, really lowcost supersonic aircraft, you’d be flying more than or close to twice as fast as the adversar’s fastest drone at this cost. So, that would be a complete game changer. >> Awesome. >> Congratulations. >> Thank you.

6:18:39 [music] Hello, my name is Kira Rashinach and I run the public sector business at Reality Defender. Reality Defender

6:19:11 detects deep fakes. Today, anyone can fake a face with a LinkedIn photo, a voice, with a few seconds of audio using one of hundreds of thousands of commodity tools or open-source models that are out there. Over the years, AI has become absolutely indistinguishable from human. And because of decreased compute costs

6:19:42 and the fact that AI doesn’t get tired, um this material is proliferating across the internet. As of March of this year, we pass 51% of internet traffic being AI generated. And so because all of our lives we’ve depended on the idea that the face and the voice in the communication channel whether your phone or on social media is seeing is believing. Um we invite you to to try and see if you can detect a deep

6:20:13 fake. So deep fakes affect all parts of digital communication. Um given that the problems we’re seeing are across phone, your video conferencing tool, um fraud channels in a call center, and most importantly within all of these social media channels that we see every day.

6:20:44 So to solve this problem, Reality Defender, we’ve we’ve built a very flexible API. This API allows us to with just two lines of code, you can implement deep fake detection models into any of your existing platforms. So, nobody wants to reinvent the wheel with whatever system that they’re using across biometric authentication, call center security,

6:21:14 your t tactical communication at the edge. Um, what we’ve done is allow you to utilize whatever tool you have and plug in our models. So, across all of these different tools, we’ve integrated reality defender models, right? And so with low latency, high accuracy, customizable models, you’re able to use the same open-source

6:21:44 analysis system, call center security system that you’ve been using for a long time. And so just like antivirus software, deep fake detection can be natively integrated into your uh into your existing tool set. So we are a true dualuse company um across the federal civilian space in the intelligence community. We’ve leveraged our models for primarily open-source

6:22:16 intelligence use cases. Now the place that we see the strongest application of our models is with communication security. We want proactive defense, proactive identification of deep fakes. And so though the barrier to entry is higher, we’ve seen so much success with ad hoc intelligence analysis. And so now we’ve just penned a partnership um with Samsung to work on device directly to bring our models to proactively defend

6:22:47 warfighters in communications environments. Um so thank you so much. um will will be here uh in the back today. [applause]

6:23:32 Hi everyone, thanks for being here. Uh I’m Leo Kungun from Finland and I was part of YC Batch in winter 26. And today I would like to talk about a little bit different topic which is that why in future we’re going to most likely breathe from water instead of pressurized cylinders and why we should. So a little bit about myself I’m electrical engineer. I’ve served in the military and also I’m uh a certified diver. Uh now

6:24:03 just on a side note, if anyone of you knows anybody who works in CBRN or breathing technologies or has problems in this space, we would highly appreciate people that could be our testing evaluation clinical partners in future. And for policy makers, please pay attention to the CBRN topic because it’s something that we don’t see every day, but when it happens, it happens really violently as we saw in COVID. So [clears throat] water is a really really good source of

6:24:34 oxygen. Why? Because essentially every single water molecule is an oxygen plus hydrogen. Now we can actually access this massive potential of oxygen just by applying DC electricity and electrolying the water. Now, to show you more tangibly how this works, I have actually a demo video that we filmed a year ago. And it is me diving in a Finnish lake with a with exactly the device that electrolyes the water that I was diving

6:25:04 in around me. And the pipe that’s going into my mouth. Uh, sorry, there’s an issue. One second. Sorry about that. So yeah, in the video uh there’s a pipe going into my mouth which supplies the

6:25:35 oxygen and near my hip on the right side is the pipe that uh vents out the hydrogen. Now this was done on the lake setting uh with the electrol electrolysis gear. Uh so it’s been proven that it can work. Now why is it then important to develop this technology in future? Well for a few reasons. Uh current cylinders only supply oxygen for maximum two hours. Usually it’s about 30 minutes to an hour

6:26:07 and after that they’re just a dead weight. In a combat situation, if you’re wounded, you need 100% medical grade oxygen usually. And the window in which you can save a person’s life is 72 hours. So you have no time to wait for a truck shipment of new cylinders to come to you. And there’s actually zero fielded systems that could supply 100% medical medically pure oxygen in such situations that also work in uh contaminated air uh spaces where the air is not pure enough. For example, oxygen

6:26:37 concentrators cannot work in a air in a air environment where the air has some kind of contaminants in it. Otherwise, it will just pull push the contaminants into your lungs. Uh also on a side note, the current devices are about 30 lbs. Our device is 11 pounds. And I would say the simple math to understand the scale of this is that our device can provide 15 times more the breathing hours per pound carried. So the amount that the person in uh any kind of commercial

6:27:08 setting or the war fighters have to carry is now much less and this technology is not new at all. Actually ISS and submarines are using this techn technology even now and have been using it for past 50 years. But the only constraint that has been there is that both of these settings have an unlimited uh relatively unlimited energy source in submarines. It’s nuclear and in ISS it’s solar to facilitate the electrolysis reaction. Now the new frontiers where you’re probably going to see this technology be

6:27:38 field being fielded in the near future is uh different CBRN situations such as uh toxic chemical attacks especially delivered by drones then subterranean environments where the air has nowhere to move. Uh and then also at altitudes. So uh we don’t know much about it yet but there’s definitely a lot of talks about uh stuff happening at high altitudes where you cannot really use any other device as uh except pressurized cylinders or water

6:28:10 water-based systems. So come and see our device on the stand uh in the in the room behind you. And just uh talking about the earlier side, we’ve have letters of support from US Army special operations force and naval special warfare development group. But if you know anybody who has this issue uh with logistics or uh medical access to medical grade oxygen, we’ll be more than happy to talk to and connect with such people. So thank you very much and

6:28:41 remember the CBRN is a acute threat tomorrow and in future as much as every other threat. Thank you.

6:29:30 So this next panel uh we want to give you a snapshot shot of uh what’s going on midbatch. And just maybe to take a step back for those of you who might not be super familiar with Y Cominator. Uh the way it works is if you know you and a friend have an idea for a cool startup, you apply to Y Cominator, there’s a less than 1% chance to get in. You get in uh you receive about half a million dollars in funding. you give up about 7% in equity and then you’re required to move to San Francisco for three months where you go through sort

6:30:01 of an intensive mentoring process uh with some of the kind of sharpest minds in Silicon Valley folks who have built and exited uh tech companies um and we’ve been hearing from some of those people like Tyler and Vivian uh and oh and we’re going to have uh Chris Goldar a moderator a general partner uh moderate this today he co-founded for example Backtype a YC company that was acquired by uh Twitter Twitter and then today he’s going to talk to some of the current batch. So they’re in the middle of this three-month uh process. Um and

6:30:31 they’re going to and all of them are in the uh sort of defense uh domain. So uh Kush Argawal is a co-founder and CEO of Avieran which builds unmanned aircraft for geospatial intelligence. He studied aerospace engineering at UCLA where he taught drone engineering, led the autonomous aerial systems club. Charlie Jess is co-founder COO of Streamline Systems, which develops navigation technology for unmanned aircraft uh operating without GPS. He’s a Cambridge graduate and Royal Air Force reserveist. Mo Akbar is the founder and CEO of Foreshadow Industries, which builds AI

6:31:03 world models to help special operators and autonomous systems plan and reason about their surroundings. He previously worked at Anderil, uh, Vanavar Labs, P1, and SpaceX. And Hendris uh Donsburg is the founder and CEO of Lockberg Industries, which is developing AI native ballistics for ground combat. He previously co-founded Flint, a software company serving uh electrical engineers. So we’re going to hear about folks at the beginning of their journey uh in the fall 26 batch. So coming on to stage.

6:31:48 All right. Thank you uh for the introduction, Luther. Um, so yeah, I’ve been excited for this panel because as Luther mentioned, um, these are our most recent companies that we’ve just funded literally, I mean, I guess three weeks ago. And so you’re really getting a peak into what founders are thinking about right now. Um, just going to kick things off. Let’s go down the line. Um, why don’t you expand on your backgrounds and what you’re working on? >> Hi everyone, my name is Kush. I’m the co-founder and CEO at Avier. I studied

6:32:19 aerospace engineering at UCLA. I like graduated two months ago straight into IC working on avion. Uh we are building autonomous aircraft and the autonomous systems that power it. Right now we’re working on small aircraft that do infrastructure inspection but the plan is to go to much bigger planes that can carry thousands of pounds of cargo both for commercial applications as well as contested contested logistics in denied environments. >> Hi everyone, I’m Charlie um co-founder of Streamline Systems. We’re building very lowcost positioning modules that allow drones to navigate entirely autonomously without GPS. The modules

6:32:51 are drop in replacements on existing flight controllers. Uh and the kind of the core problem we’re trying to solve is producing positioning systems that scale for this mass autonomy. Every Western government is talking about a traceful mass. The problem is whether whether that’s you know counter counter US drones, loitering munitions, oneways. The problem is is that existing solutions don’t really scale for this. It makes no sense to equip platforms with thousands of pounds worth of kit if they’re going to get blown up on the other side. That’s the problem we’re solving for. All right, cool. My name is Bo. I’m CEO and founder of Foreshadow Industries. We’re building a world model that we baked onto a chip and creating

6:33:23 what I call the JDAM for autonomy. So, we have a single piece of silicon. You put that onto your existing vehicles, whether that’s a regular boat, regular ground vehicle, regular aircraft, or an existing autonomous system, and connect up your sensors, and now you have an autonomous uh autonomous system. >> Hello everyone. Uh my name is H. I’m the founder and CEO of Log Industries. We build both surfacetoair and surface surface ballistic missiles alongside with portable launching platforms allowing for um deployable and fast systems alongside with customal systems

6:33:54 of launching missiles and not having to modify this infrastructure meaning that we can decrease cost of uh termination and also deploy more terminal guidance systems in our missiles. >> Cool. Um let’s dig in a bit more. So Kush um Avier autonomous flight systems for inspection aircraft. How do you see that evolving? Like what are all the different use cases? >> Um so I feel autonomous systems should be there on every single plane out there. You’re seeing that with cars, we have VMOS, Tesla, Zuk. So much happening in the autonomous space in the

6:34:25 automotive industry. I believe every single plane can also be autonomous. Right now we’re working on small aircraft because it’s possible to test it on them without a lot of regulatory hurdles. But the idea is to expand it to much larger cargo aircraft. We are seeing a couple of people do that. And this space is going to open up a lot as cargo needs increase both on the commercial side as well as as on the defense side because in contested environments you need to transport cargo really quickly and effectively without putting war fighters at risk and having an autonomous system that’s able to do that over long distances at a low cost

6:34:56 uh is super important. So working towards that by starting small but eventually building up to large scale autonomous systems. So you’re starting small, but like what kind of cargo, what kinds of locations, and like why is the autonomy actually valuable for those missions? >> So if you look on the commercial side, uh a lot of the cargo fleet is aging. Um last year after the Kentucky crash that UPS had, all the MD11s were taken off by a few carriers. So there needs to be new improvements, new airplanes that are able to move cargo automatically. And if you’re making new airplanes, might as

6:35:27 well make them autonomous. There’s a lot of legislation already being done. There’s part 108 by the FAA for aircraft up to 1320 pounds and that’s only going to expand for much larger airplanes and similarly as I mentioned on the defense side as well specifically in the Indoacific region along the island chains uh things need to be moved across pretty quickly between the islands and those are pretty harsh conditions and that also would benefit a lot from large scale autonomous systems that are able to do that pretty effectively. I mean, if you’re designing your own airplane, I mean, what does that mean in terms of

6:35:57 like cost or or time to your first deployment, are you able to like retrofit existing aircrafts today or is there any shortcut? >> So, we’re exploring both options. Retrofitting is definitely a possibility. A lot of companies are kind of experimenting with that. Uh, I myself, I’m a pilot. I fly small planes and they’re autonomous for about 80%. The landing and the takeoff is the most critical part and it’s also possible to automate that. So, that’s really good starting point. And then clean sheet aircraft design definitely is uh on the cards as well. It’s a challenge. It’s

6:36:28 it’s basically rocket science as well. Um but getting there is a bit easier if it’s unmanned because there’s not that much safety that you have to take care of if there’s not a person inside the plane. Obviously not saying that it’s not going to be super safe and well tested, but there’s a bit less of regulation and that just makes it more possible in today’s environment where rules are going to be opening up for such systems. >> Okay. Um, Charlie, um, for navigation where GPS is unavailable, that sounds amazing, but like what kind of evidence

6:37:00 can you show people today or a skeptical customer? Um, other than just showing like a video of it presumably working. >> Yeah, absolutely. So, with drones, it’s quite hard to make a convincing video. It’s they operate a long way away, usually quite high up. It’s quite hard to uh to film it convincingly. So, we’re booking trials at the moment. We’re working um working with some British partners in October and then we’re booking some with American um American partners in early November. We’ll demo the tech in person, integrate it onto their platforms, and kind of work out any teething issues in the process. But to be honest, it’s going to be a

6:37:30 collaborative process working with kind of early adopters and commercial partners to work out where the best missions are, what the requirements need to be, and you know, any teething issues around the around the integration of the technology. >> Okay, got it. Um let’s let’s chat with Mo. So, how are you deciding which problems actually deserve your team’s attention? I mean, there seems to be like a lot of different use cases and scenarios where you could be useful. I mean, there’s satellites, there’s robots, there’s human teams, they all have different requirements. So, I’m wondering what are you most focused on to start?

6:38:00 >> Yeah, it’s a good question. So you know when we started the company um a lot of our funders um and founders suggested we go after the largest addressable market which would be the commercial side using it for manufacturing it using it for humanoid robots but I think I want to go after the most high value problem first and I see the most high value problem as the one that exists in the Indoacific um and working with special operations command making it so that we have the ability to get mass into theater and you know there’s a world in which we have a huge amount of production of autonomous

6:38:30 systems and they arrive in time. Um, but let’s say there is a conflict in next month and the Taiwan Strait is blocked off. If we send in human personnel there, we could potentially create a war. But if you send in, you know, take your existing boats and ships, turn them into autonomous systems and allow us to resupply. Um, that gives the US more strategic options. And I think that’s probably the most valuable first problem we solve. And after that, the solving of that problem will create um technological advantage that we can then go and commercialize. Um so solving how

6:39:02 you get highly reliable autonomous systems for the most important national security problems will actually benefit the commercial sector when we want to do the most highv value work with autonomous systems as well such as like repairing nuclear facilities critical maintenance of infrastructure if you solve this for defense where you know failure is not an option first you can go after the most high value work in the commercial industry after that. >> Got it. I mean uh maybe with one of those examples and I don’t know how much you can share but um even if a it’s a hypothetical one like what kind of information goes into your world model

6:39:32 what comes out like what does the operator do differently can you walk us through the technology a little bit sure so there’s two scenarios in which an operator will use this one is you’re as an assistant to a human operator the second is purely as an autonomous system inside of that the ingestion of the data usually comes in as tracks so telemetry of these surrounding system. So, you know, a regular military unit might use cursor on target. Uh, and so you ingest that geospatial data. You also ingest sensor data. Sometimes those come in as ATR chips um or other sensor models. You

6:40:04 ingest that into the world model. You take in the commander’s intent or a description of what their objective is. You combine that. The system generates multiple plans and hypotheses of how to proceed faster than real time. Determines based on the physical constraints of the environment which of those are actually physically achievable uh based on our world model. And then it actually goes and simulates this and then you just do this in a loop continuously. And you know that’s a new type of intelligence that allows your autonomous systems to reason and understand the consequences of their action before they act. Uh it’s the kind of intelligence we need to scale uh in

6:40:35 order to solve these problems. >> Got it. Um for Luck I mean what are the design or manufacturing choices that matter most to making your product actually like affordable or you know your ability to build build it quickly and repeatedly? >> Yeah, absolutely. So uh when it comes to such a I guess complicated issue as building ballistic missiles, there’s a lot of components that are often times sourced from aerospace factories, right? And when it comes to the assembly process of these components, I think the biggest thing you want to shy away from is having a person that has a PhD be needed in the room to finally finish the

6:41:06 assembly process, right? That means that first of all, you’re limiting self on how many employees you can have, right? How fast a person can go and for example, how many states you can place it in, right? And um how universal really is, right? So that’s one thing I guess, right? So the most important thing being just optimize your design for mass manufacturing meaning that it shouldn’t be a lot of components. It should be as simple as possible and it should be as fast as possible. And I think one other thing that I like to talk about is definitely the scalability of using existing infrastructure when it comes to factories to equip your design for mass manufacturing. Um, as Mo said, you know, we never know if there could

6:41:37 be an issue that means that we might have to introduce high-scale manufacturing for ballistics, right? Uh, and if this need of careers just for tomorrow, that means that our design should be capable of being inputed into existing factories, right? I think um, Andrew has done a phenomenal job when they made their missiles adaptable to car manufacturing factories, right? The assembly lines where now their designs are so simple that they can be assembled with robotic arms, right? And uh this the last thing you want to forget is having people in the room that have too much experience for building this, right? You should be with the PhDs in the design lab and then people that have

6:42:07 actual experience assembling hardware and robotics and automation of assembly in this mass operation of assisting infrastructure. >> Got it. U one more followup. I think I have to ask about AI or SI or whatever, but what part of the ballistics actually benefits from AI? Like how would you validate that? >> Of course. Yeah. So most importantly I would say what we’re working is is we’re researching on vision models right uh there’s a new way how we can take ballistics especially compact ballistics allow them to guide themselves right I think um if we ask how we validate AI I

6:42:38 think it’s very simple it’s just the amount of hours that you run an AI model whether it be synthetic data real data uh in our case we get a lot of data that we can train on in Ukraine so the models are run locally mostly on inference and we train them from parameters that are given from Ukraine in images such as delta um so for example in American scenario you might have to use synthetic data to train your models uh in some cases where Ukraine has had existing infrastructure of this training data they can source to us. um as far as I guess yeah validating the data um and how I guess we ensure if it’s safe is mostly having a in loop operator right

6:43:10 so I said very sinly that we’re using uh adaptable systems so they’re compact and they can moved and assembled what this means is that uh you don’t have to modify this infrastructure so when we see I guess legacy systems being modified on top of trucks this means that everything cost of termination decreases the fact you have to massroduce this modification right and if you can create a system that is easily deployable or easily assembbleable uh it can be the frontier of how we see ballistics from mid to low range and uh we also have have an lloop operator. So you can have an autonomous system for example for ISR

6:43:40 reconnaissance and then collaborate with our system for launching for surface to surface. But then again you also need to have an operator held liable responsible. So having an existing area where you can pretty much input your coordinates and then allow vision models to run and understand what exactly is the necessary target based on the preliminary preliminary data. So whether it be ground targets, just tanks, they’re all within a confound of existing area, and then that’s kind of how we moderate the safety of it. And yeah, cool. Uh let’s bring it back. Um I know there’s varying amounts of what you can share, but if you can, um can each

6:44:13 of you tell us like what you’ve actually been able to demonstrate so far, your biggest sort of proof point, and if there is any feedback that you can share on that? Um Kosh. Um so we have demonstrated full stack autonomy on our small airplanes. So even the airplanes have been built completely in house. Uh we have a ceiling of £55 for what we building right now to comply with 107 regulations. That’s what we are working under and it’s able to taxi take off land fully autonomously. And that’s something that’s scalable. Software is always scalable to bigger airframes with

6:44:44 more testing. And I think the feedback that we got um people we showed it to in our own testing was um testing is never enough. You got to keep doing it. Flight testing is going to be a very important part of aerospace, every single aerospace company, not just us. And it just gets better over time. And other thing is safety. We have to make sure that our failure rates are really low. And we kind of forced into that anyways by the FAA because if we have any crashes in our past 250 hours of flight, u the clock resets for certification. Uh that varies again in what regulation

6:45:15 you’re working under. But safety is a very important part and just what we have been doing is critical to making sure we are always within that framework and making sure our systems are reliable and that just scales up to bigger air frames as well. >> I mean how do you balance that because you have people like me working with you telling you to move move and go fast and faster and faster. So are you ignoring us like what are you doing? So I think what’s important is to like kind of uh understand that this is still a lot less intense than manned aviation or manned aircraft which is where a lot of these

6:45:45 um rules are coming in from like right now if I want to fly an aircraft fully autonomously it’s I have to file a waiver uh with the FAA if I want to fly it beyond visual line of sight it takes a long time mainly because it’s borrowed from a lot of rules in man aviation that’s opening up so we would be able to test faster it would be a lot quicker so post timelines are going to get compressed and in the environments that we are working with that’s over deserts it over out at sea um any crashes or anything else which could impact a lot of people over let’s say cities is

6:46:16 that’s not the case over there. So when we go into testing uh over cities or over people which is about probably a year or two out from now we would have a really good track record by then and that’s something that’s well understood by investors as well because again safety is the most important thing in aviation even though it is the most uh safe form of travel or transport. >> Yep. Yep. Uh Charlie I think you can’t share much because you’re quite stealthy but >> can share a little bit of the same regulation obligations as Kush. Um, but we can de we can demonstrate entirely

6:46:46 autonomous flight without GPS and then GPS handover as well. If anyone wants to find me after find out more details, I’ll be I’ll be hanging about >> there’s very little I can share in this in this public forum, but uh yeah, call me afterwards and I can bring you down. We can show you a live test. >> Okay. Okay. Um >> maybe a little more. >> So, we’ve been able to pretty much proof our system for autonomous guidance in the simulation. Um, what I can tell you is we’re about we’re currently in stage of R&D. So we’re about two weeks out from physical assembly and um current biggest I guess learnings have been or proof of concept has been the mass I

6:47:17 guess attraction for the Ukrainian government. Uh we understood the system was so necessary but I think one point that is really really important for all of us consider and we talked about this in the very start um was the mass iteration that we’re going through especially in modern warfare. Um I guess we’ve had a lot of technological superiority in the past and now when the war is moving to something a cheaper faster and massproduced especially when it comes to like quadcopters for example um Aaron mentioned this in a speech before but uh they pump out a new design every two weeks right so it is very very important for these companies like ours to look in the future so uh we kind of put our product basis where we want it

6:47:48 to be also said we want to reach hypersonics we want to reach you know uh large scale systems we want to see dep deployable systems right and this was something we unders that the Ukrainians really understood the market is changing very quickly whether be shot head drones, counter UAV interception, uh you have to think about the future, right? And uh we thought about it and we told them that the validation we got was immense and we understood we kind of have to build towards this, right? The market’s changing so quickly that you have to stay ahead and um currently we’re on the way to there. Yeah, I’d love to catch up more if anybody has any questions. >> Yeah, I mean shifting more to the

6:48:19 customer, I mean what what’s preventing each of you from getting to a purchase or like what kind of evaluation would you want to run next? um maybe with someone in this room but uh what kind of results could you commit to in something like that? >> Um so I think the most important thing for us on the defense side which I think is more relevant here is understanding more technical requirements from the folks working at the battlefield or where these systems are required. So stuff like what’s the range you want, what’s the runway available, what’s the condition of the environments that you’re in. That’s super important so

6:48:49 that we can demonstrate our capabilities. So right now again it’s for the next six six months it’s going to be subscale models running our entire autonomy stack. So it would be super important to understand what exact what are the conditions we will be working in and if you’re able to demonstrate our subscale models there we will have a lot of data for making our full scale aircraft as well and the horizon for that is about a year year and a half which is still pretty uh quick for uh compared to man aviation. So just understanding more where we would be working um and more conversations is what’s going to help push us forward.

6:49:21 Okay. Um, yeah, engagement’s been pretty positive for us. There’s a real need for GPS live navigation, particularly on drones. Um, the big constraint for us is time. We’ve only got eight weeks until the end of the batch. So, there’s limited uh slots available for how many pilots we can book in. We’re booking those now. So, if you know any drone dominance companies that want to book the last two remaining slots, get them to reach out to us. They’re they’re going quickly. >> Yeah, we’re doing daily uh testing out in Cornado. uh and we’re doing a two four week sprints and uh yeah so we’re

6:49:52 we’re in a good place I think in terms of the first set of customers that we’re working with always uh interested in finding more customers who are interested in this space as well but I think we we want to get as many reps and sets in operational environments as possible and then expand out uh to more customers >> um I guess in our case um it’s simply one thing which is testing sites um it’s sometimes very hard to find testing sites where the government can approve of something that you’re doing um so if there’s anybody with a contact like they’ll be interested in helping us with um testing our system for initial R&D.

6:50:23 It’ be phenomenal. Love to speak after. >> Got it. Um if you haven’t covered it already like what’s some useful help you could get from government whether it relates to suppliers or access to testing is there anything you can share? >> Uh absolutely. So I think we are trying to be as vertically integrated as possible making the plane and autonomy stack in house. M >> so supply chain is something we are taking care of from day one but having better access to NDAA compliant equipment would definitely help us shrink that even uh faster and um we’re

6:50:55 also ready to uh do pilots and tests on um in those environments themselves. So if there’s anyone who is working in such conditions and you want a solution that can move around things quickly which can carry a lot more payload than current EV tall drones or small UAVs I would love to meet you and kind of show you and walk you through what we can do and possibly then demonstrate something as well. Yeah, the big thing for us both my co-founder I Britz um it’s quite difficult to engage with the DO when you’re not a US citizen maybe a recalibration of the the metrics by

6:51:26 which you quantify whether you’re operating in the US rather than it being based on citiz based on where you’re manufacturing operating producing selling to um would be at least for us bit more preferable but yeah I can understand why it’s a why there’s why is the concern >> yeah I think if you have a serious problem that you care about solving in a short period of time. I’d love to hear with you if you care about, you know, DDL environments or like denied degrade comms. You care about like full autonomy without connected um without connections to like Starlink or radios or things of

6:51:57 those scenarios and you actually want to be able to operate in, you know, realtime environments, reach out to me and we want to work quickly to solve your problems. >> Yeah, absolutely. Um I think I agree with that definitely. Uh one place is the foreign control structure of a company when it comes to defense. Um, as Sh said, I’m also Latian, so dealing with the United States government has not been always easy. And recalibration of the structure of perhaps a 31%, you know, operational structure of the United States would help us immensely because we do want to service the United States as the government and we truly believe in technological superiority.

6:52:28 And um, yeah, also authorization to test autonomous systems. Um, it hasn’t always been very easy. So also relays testing sites. Yeah, that would be great. So if anybody knows any that would be great. >> Great. Um I mean I know we heard we want more primes like what is if you want to share what is the big vision for your company. Um feel free to share otherwise I think we can open it to questions. >> I think the big vision is to make every single plane autonomous. Um I think we’re on track factor that both on the technical side and the regulatory side

6:53:00 and uh that’s the big optimistic dream and very hopeful about it that every single plane and every single car out there will be autonomous maybe still with a pilot in the loop for large scale operations but it’s still being driven by software at the end of the day. >> Uh we want to replace lowcost um we want to place GPS for lowc cost platforms. Um, ultimately we want to scale it to kind of every platform, UGVs, UXVS, USVS. Um, it’s about expanding the kind of portfolio that we’re we’re integrating the technology on.

6:53:30 >> Yeah, we’re planning on taking our technology and moving down the stack. So, our AI model, baking it onto the chip, and then fabricating it here in the United States. So, the long-term vision is have our own fab here in the United States where we manufacture the chips as well as the models and rapidly develop those so that we can do this in the United States. Yeah. So in our case uh we want to maintain technical superiority when it comes to ballistic missiles while also optimizing our designs for mass scalability. So meaning that we can be employers a company who mass manufacture in a state of conflict through existing

6:54:02 infrastructure of manufacturing. And um that would be the great goal for our company. And if we can do that that’d be great. >> Great. >> Awesome. Thanks guys. Heat. Heat. [music]

6:54:34 [music] Heat. Heat. >> [music]

6:55:14 >> That’s not good. My name is Robert Thelen and I’m the CEO and co-founder of Rounds. Today I’m going to be talking about outiterating the threats. This is something I unfortunately have a lot of experience with. When I was in Afghanistan, we were uh doing comet advising for the Afghan

6:55:44 National Army and the Hakiri network was the uh the local terrorist group and they found the most clever ways to outiterate us. we’d have a TPP and they would find a new way to get someone in there. All culminating on December 6, 2010 when someone made it on our base and uh blew themselves up, killing two friends. The ringing in my ear that I still have today reminds me that our enemies never stop iterating. They never stop learning. They never stop finding new ways to get around what we’ve built.

6:56:16 And it’s not it’s not any better. Uh, one of our customers, uh, a user, a Ford air controller, he showed me a warehouse full of drones that have been delivered to him that are no longer usable. They’re too expensive. The radios are bad, and they don’t do the mission that they needed to do now. They did a mission six months ago. So, today I’m going to talk about what we need, and that is a new type of weapon system. It’s a system that’s uh that has mass. It’s cheap and expendable. a system that has collaborative autonomy. So, uh they

6:56:47 can talk to each other and they can do a mission without talking back to a person and they have to iterate quickly. So, what does it mean to have a triple mass? That is a drone that is cheap, not expensive, that you can throw away if you need to, that can go off on a mission like a bullet and you can leave it behind. This is not a rifle. It’s not something that you protect. is something that could be lost in battle and you have to have a system in place to replace it. The the country that can do

6:57:17 that is a country that’s going to win the next war. Uh this is just an example what we build. We build a fixed wing. We build this type of drone. You tell us uh what the what requirements are and our AI will build build it out and print it out in a few hours. Collaborative autonomy. What this means is that it understands the mission. You talk to the drone. There there are no controls. There is no cool headset you put on. You talk to the drone say go off over the hill and tell me if you see any white trucks and it does it. This is critical because if you want to actually

6:57:49 uh scale in an interesting way, if you want to have one person with a 100 or 200 or a thousand drones, they cannot be controlling all of them. We do not scale, but technology does. And all the models need to run on the actual platform because our enemies know where our weaknesses are, communications, and they will hit them as hard as they can. So an example of this is what we call round sentry. It’s a perch and stair sensing system. So it uh you you land at a landing base. Uh we’re doing this actually with the Marine Corps. So you land and you say uh set up a cord on 15

6:58:22 km out. They take off, they do that, they find a safe place to land autonomously. They put down, they put up sensors and listen. They’re listening for people people approaching. They’re looking listening through RF signals. They’re trying to find an opportunity or someone that’s trying to sneak up on you and give enough heads up in the systems already exist to survive. They they uh they listen and act. And when they get low on battery, guess what? They come back and swap swap themselves out autonomously. It needs to be disconnected. You can never trust you’re going to be able to talk to a drone again because again, our

6:58:53 enemies know where we where we’re weak. They know and where we’re strong. So, they’re going to go after communications as much as they can, especially in a peer-to-peer war. And we have to learn. We have to have a loop that is so fast that it can outloop the enemy. Ukraine gets a new drone in the air every two weeks. Russia does it every month. Who’s winning that drone war? We need to increase that speed. And that means increase the models. So when one of our drones lands, uh, a edge box is right there. It downloads all the videos from the day. It retrains the

6:59:24 models. It retrains the TPPs. Puts it back on the drone for the next mission. And if we’re losing and treating too many, we need to reach back to our facilities to start putting out new drones with new qualifications, new radios, and new motors. MSL. So, so I I worked at IBM and I was in Air Force. So, we need to have a cool equation. We call it muscle. It’s mass times silicon over loop duration. So, silicon is like all the autonomy and stuff. So, this gives you a score. Russia, you know, has this big score.

6:59:55 They can produce about 2 million. Autonomy is this uh scaling thing that we we look at as a force multiplier. If you’re not if you don’t have autonomy, it’s not going to work. And then you have um you have Ukraine over here. As you can see, they’re doing better. And we see that in battle. You have China over here that’s just rocking everything. And and that’s not an exaggeration of where we are right now. It’s not an exaggeration that that if we have a pure war against China that we are not prepared. There’s a lot of things happening in in the DO try to fix this. But as of this moment, we’re not prepared for that war.

7:00:27 But if you focus on autonomy, mass, and and that loop and getting down as low as possible, not only do we win a war, but we prevent future wars. If the enemy sees that we are faster, stronger, and better than them, they won’t attack. If they see the first island chain and second island train are impenetrable with wave after wave of drones that we can recreate over and over again into infinity, wars are prevented. We have real progress. We have prime contracts. We have subcontracts. We are

7:00:58 delivering these drones to the field right now, primarily through special forces, ABSOC, the Air Force. And we are going to be going to uh to Talisman Saber, which is one of the coolest exercises down in Australia. And we get to be on a on a ship with Marines uh trying to forge a beach. So, it’s going to be a lot of fun. But the AS from DC, so the capital markets has the SNL side of this, the silicon, all the AI stuff. It’s cool. Capital markets like that kind of stuff. the loop, the iteration loop. The AI training AI is right up the alley of EC. Easy peasy. But the math is

7:01:30 hard. Uh it’s hard to go to a uh to a bank and say, “Do you know what I’m going to do? Make a bunch of cheap products.” That’s uh cheap drones with no margin for the next 10 years. So that is where we need the most help is in mass trying to set up a production line. So any warm intros you might have to DIU non-strike drones. Drone dominance is awesome. We we are going to be participating in the next round of drone dominance. But we have a lot of cool stuff that has nothing to do with drone dominance like putting a sensor down range is is not on that same same list of things or the office of capital. I

7:02:00 really appreciate it. Our ability to win the next war is dependent upon our ability to iterate. That ringing in my ear I never want to happen to another person. I want to ensure that we have a standoff distance that never puts another American within harm’s reach. Thank you very much. I’ll see youall later.

7:02:36 >> [music]

7:03:11 >> Yeah, welcome everybody. So this talk is about how to turn risk into prime contracts and capital and it’s also a bit of the summary of everything what we’ve heard today all the great presentations from all the very innovative companies. So I am um from Munich, Germany where at the moment the October Fest um takes place and I was a PhD researcher at the University of Munich in risk management and insurance and I worked at Munich Re the largest reinsurance company in the world.

7:03:44 So this talk is about when when startups innovate there’s mostly no historical claims data. So when we go to the moon the moon landing is one of the greatest examples. The moon landing today is not insurable and it’s very very difficult that startups are getting these contracts where they can really be the first ones when there’s not insurance capacity available. And there are mainly three use cases that startups face when they raise capital when they open a new

7:04:16 facility or when they want to win a government contract. It doesn’t matter with what agency with the navy with um even with NATO and um with other departments they need to get insurance and I can promise you that when they have it everything becomes extremely easier for them. So we have seen that when the insurance market is insuring these companies everybody’s interested to take the risk on them. So more agencies are um are interested and there

7:04:46 are two clauses the F52285. This is the work on government installations and it’s mandatory. So and this is becoming very very urgent at the moment because everybody’s innovating but nobody understands anymore the risk behind it. So we have like one example for raising capital. There’s around 50 billion invested in defense tech in 25. It nearly doubles then 2024. When there is historical claims data, insurance works

7:05:18 great. When there is no claims data, what we s see today in AI liability, then insurance is unfortunately moving too slow. It takes 12 to 24 months today to bring a new product to market. So we don’t have time for that. And here we see some of the investors that invests in defense tech or space insurance. And all these investors require at series A and series B phase that you have institutional coverage in place. When startups or businesses open new

7:05:51 facilities, the landlord requires coverage in place. Here we saw like general liability for example is a huge issue at the moment for defense companies. they cannot even open the production side without having it in place. Here’s one example from our software that we are building in in-house is so we are testing multiple facilities and when you open for example um a site in in Oregon or in in Louisiana we test the facility if you are allowed to get

7:06:22 insurance at some point if you open a site where it’s extremely risky nobody will cover you even if you paid and even if the lease is approved so this is this is very difficult so here we can see like a heat app from one to five. Five means the facility is very risky regarding specific natural disasters. One is low risk and we want to identify the the place where you can open your facility where you where you really can get the insurance products in place. Yeah. So these are the two contract

7:06:54 requirements. So if you are a startup and you sell to the government, these are the contract requirements that you have when you when you sell to the department of defense. this is something these are the products that you need to have in place to be able to sell and unfortunately it’s just not possible. So we saw contracts, everything was approved. The Department of Defense was interested, the military were interested, but unfortunately there was no cyber security coverage in place. They lost a hundred million or $80 million contract. So we saw this in the

7:07:25 last couple of months more and more coming. Yeah. So this is the vision that we have. So we so what Risk goal is to unlock the insurance products the new economy needs. We see here AI liability coverage, drone insurance, Luna landing insurance. All these things today barely get capacity. And also the US doesn’t really have a great reinsurance company. The biggest reinsurance companies in the world are in Europe. China has China re China has the possibility to immediately

7:07:57 react to circumstances through the insurance market. The US unfortunately doesn’t have it today. We are highly dependent from foreign capacity. So all these products somehow I think we all will agree that with AI the world becomes riskier more very risky in the future. We need to react to specific circumstances very fast. So this is the vision and yeah so how Washington or all of you can help us is we would really like to get more into the federal

7:08:29 government to open the discussion and bring awareness to the topic how these startups can what should the startups do to get these contracts done. So we saw that there is interest from both sides to work together but unfortunately it’s always getting stuck when it comes to institutional coverage and so on and this is what we want to achieve. Yeah. Here are my contact details. At the moment we are licensed in a couple of states. So we already transact and we have um companies like refineries for

7:09:01 example as customers that are producing for the military chemicals. So we already insure a kind of companies that wouldn’t get insurance without us and wouldn’t close contracts. When you have any questions, let me know after the talk. I will be here. Thank you very much. [applause] [music]

7:09:33 >> [music] >> Wow. Oh, it’s a disgusting desktop, isn’t it? [clears throat and snorts] Hey, uh I’m Alex. I’m the uh CEO and

7:10:03 co-founder of Seeing Systems. Uh thank you for having me. Um it’s been really cool to see some of the technology. Um but today I want to try and pitch things in a different way. We are another drone company. Uh but we build modular autonomous drones. And uh hopefully I can frame it in a way that you guys will understand why. Here’s [snorts] something we don’t really say enough. uh a system can still do everything it was built to do, but over time it’ll become useless. Uh we see this with sort of expedited innovation cycles. Nothing’s broken, but the world around it has changed, the

7:10:34 demands have changed, and the problem has moved on. Uh there’s a great uh quote here that I want to take from something called the Red Queen hypothesis. Uh and and it goes, it takes all the running you can do to keep in the same place. And it’s from Through the Looking Glass. Uh Alice is running next to the Red Queen. uh when they stop, she hasn’t moved. And now [snorts] you might be thinking, how is this relevant? Well, when we look at sort of modern conflicts today, uh this is what a war of innovation looks like. We build something new, they study

7:11:06 it, they have an answer to it, and we’re back to square one. And this happens on both sides. And so the Red Queen hypothesis dictates that as [snorts] long as you’re stuck in this loop of innovation, you will never make any progress in any direction. Hence the front lines that we see in Ukraine and Russia. round and round, everyone’s running, nobody’s getting ahead. Now, there’s only [snorts] two things you can do to break from a war of innovation or the Red Queen hypothesis. And one of them is a paradigm shift. So, you change the set of rules that you’re playing at. Uh the closest thing we saw in the war in Ukraine was probably fiber

7:11:38 optic drones. It wasn’t quite enough to be considered a paradigm shift, but it was a significant step change. And the other is is that you significantly out innovate your opponents. Um, so the reason that we build modular autonomous drones is that we answer both questions. We’re making a play in the long term to do a paradigm shift which is our autonomy and in the short term we’re keeping pace with the adversaries by having a modular arc of hardware architecture that can be adjusted. An example of this here is that we actually build bricks for one airframe.

7:12:08 What this means is that the subcomponents of the drones can be removed as they become useless and replaced with stuff as technology improves. But the airframe itself, every drone currently, four motors, uh some arms and a body will remain the same. Um the other advantages of this is that it means you can have distributed manufacturing. If you’ve got different modules that you make for a drone, then you can manufacture them in different places, which is very interesting for sort of wartime uh manufacturing. If everything comes from one factory, then that factory is a target. And if that touch is hit, then you’re stepped back in manufacturing. That’s an example

7:12:40 here. And this is what we see in Ukraine today uh from various partners where these large factories eventually the locations are found out and they’re a priority target for Russia. [snorts] Adapting on your own keeps you in the race. Uh but to have a paradigm shift, you have to making a bet or a play for what the future is. Uh and at Seeing Systems, the bet that we make is that true autonomous systems is what will be a paradigm shift to to make a difference on the battlefield. >> [snorts]

7:13:10 >> When we talk about true autonomy, we don’t just mean sort of basic terminal guidance. Uh we mean a level of autonomy that is safe enough to efficiently take humans out of the loop in the long term. But the time to develop that isn’t when your adversaries are doing human out the loop systems. The time to develop that is now and to keep humans in the loop so that when that eventuality comes, and it most likely will, then you can remove a human from the loop knowing that you’ve developed safe and rugged autonomy over time. >> [snorts] >> To round up, if anyone’s interested, I would love to sort of leave today with some direction. So, if modular drones,

7:13:42 autonomy or uh distributed manufacturing is interesting, then please let me know. Thank you. Can we get a voice of God to nudge folks

7:14:14 uh uh to the audience, please? [clears throat] very excited about our uh final fireside chat of the day before uh a couple more um lightning talks and wrapping up the event. Uh we’re delighted [clears throat] to have uh Chris Murphy from TIU in conversation uh with Pete Kumman. Uh I’m just going to go ahead

7:14:45 and invite him up to stage and uh let so we can have as much time as possible with them. Come on up. >> [music] >> Thank you. >> Thanks for joining us here. >> Thanks, Pete. >> Um, maybe just to get started, >> what is the DIU? What is your mission? >> So, it it’s interesting that we’re talking about this at the end of the day, right? Yes. um because it’s

7:15:17 probably one of the most relevant conversations for the the cohort of YC companies and really, you know, startups and commercial companies in general who want to do business with the department. DIU’s mission is to accelerate the integration of commercial and nontraditional vendors into sort of the defense industrial base. And we do that through a number of ways. DIU’s been around for about 10 years. We’ve continued to sort of reinvent ourselves. It’s been one of the, you know, I think, [clears throat] you know, core um characteristics of DIU is that evolution

7:15:48 from, you know, getting tech bros in the valley to open their doors and sit down with us to finding faster ways to innovate contracting at speed on existing authorities to impacting speed and scale across the department. You know, I if nothing else highlights how far DIU has come is in the last hour, Secretary Hegsth spent an hour on stage addressing the war fighters of the nation and he explicitly talked in

7:16:18 reference to DIU several times and the leadership role we are playing in transforming the department, integrating unmanned systems, bringing new technologies, capabilities and scaling them into the services. And so we do that through a few different ways. Um, traditionally we have our commercial solutions opening. Um, it’s the traditional mechanism by which we solicit openly. Um, du.m mill is the website to go to for that. And we have three solicitations open right now that if you go on, you can see it’s the only

7:16:48 place we post it. You can ask questions that get in front of the PMs who run the project. If you are, you know, wondering, hey, I need some more information about this or am I eligible or well suited for that? And that process is the competitive pathway to a prototype OTAA which is sort of the currency of how we do rapid acquisition that also if and when successful can lead to a non-competitive award of a production OTAA which is really the gold

7:17:20 standard for everybody in here if you want to you know scale into a program of record. So there’s that pathway. We’re continuing to innovate in other ways too. prize challenges. Traditionally, maybe you go, you build a small product, the reward is 20,000, $50,000. In January, we ran a prize challenge that was worth $und00 million for autonomous vehicle orchestration, command and control of unmanned systems at scale. And the whole purpose was to bring in more vendors upfront,

7:17:50 >> have time to essentially date, prove that you can do something. Let me pay you along the way. And at the end, then we go into a long-term relationship of, you know, prototype OTAA and let’s really, you know, you know, fit the form to the end user at that point for scaling and transition. So that’s a new method that we’re using right there. And then the drone dominance program as we talked about is is another example of sort of innovation of you know there are technologies that we see available. So

7:18:21 if you think you’ve got a product that works today you’ve proven it in combat in Ukraine you are working with it in sentcom you are implementing your brand new technology into a platform that you think will outperform anybody else. Come bring it go through a gauntlet. Prove it on a public leaderboard. You can Google it. Everybody’s scores for the top performers are listed there. There’s no secrets and at the end there’s an order and then we publish who’s delivering on

7:18:51 time. Not only did your prototype perform, but are you actually scaling and delivering production. So these are examples for how DIU does acquisition and how we’re trying to really meet the department’s needs for scale. I >> I love this approach to acquisition that starts with these contests. just sounds like a very merit-based way to figure out which technology wins. >> Yeah. So everything DIU is competitive and so you know we often have people come to say okay um so and so said they really want us. Can DIU buy it? And the

7:19:22 answer is um what we’d love to hear from the customer within the department, the service, the combatant command, the people who are ultimately going to have to do the man training, equipping, footing the bill to come and say, “Hey D, we’d like to work with you to solve this operational problem.” And if that is really something they want, you know, come compete. Show that you can, you know, outperform others or you are going to bring a technology that is going to incite technical surprise

7:19:53 and then allow us to run that process, do the rapid contracting best in the department and scale it into sustained programs of record. >> And ultimately your your job is to hand these off. That’s right. Um maybe actually before we get more into the process here, I want to learn a bit more about your role. You you wear several hats there, I understand. >> Yeah. Um as the CTO for DIU, um really what we do is, you know, my shop sets

7:20:24 the technical vision for where we’re going. We work closely with our portfolios. We reorganized and that was announced in the summer time frame into uh autonomous warfare into kill webs into a 10x portfolio tagline is faster better cheaper some permutation thereof to realize this sort of step change level improvement and then recognizing that AI continues to be a cross cutter it’s AI enabled for everything we really need that core of expertise to matrix

7:20:56 out and support everything we do. And so I also run our 10X portfolio which which meshes really well with those CTO duties where we can take those big bets look into the future look into what are what are the things that services or com commands aren’t asking for. We listen to the problems and we say, you know what, we see potential for, you know, quantum or physical AI or, you know, we got word that freeze-dried red blood cells, for

7:21:26 example, are just now a thing and that can actually change how we manage our cold chain within our um, you know, combat casualty care organization. So we can take and we can provide those technologies as potential options to solve an existing problem set. It’s the technical surprise piece of it that interests me most. >> The 10X is those are those are like your venture bets. Those the the the really big swings. >> I I I see that bend into sort of two categories. One is you know things that are going to provide really big step level change. They’re the big bets.

7:21:57 They’re the risk that we are willing to take and we able to take that risk over time. If it takes two, three, four years to mature, I’m okay with that. An example there is small modular reactors. We’re running all of the, you know, small modular reactor acquisition for both the Air Force and the Army under our Janice and AMI programs and we don’t expect really first la till 27 28 and that started over a year ago. So those are longer term bets that DIU makes space projects everything takes longer in space. Most of my graduate work was

7:22:27 in space systems and the tagline was space is like everywhere else but harder. And so like it takes a little bit more. And then the other end of the spectrum are how do we how do we spin capabilities more rapidly into the war fighter hands. It may not be that 10x improvement. It may be a 1x.5, 2, 3x, but I might be able to do it very quickly. And that rate of fire helps to create more volume of change and impact

7:22:58 there as well. If uh if I’m a founder in the room, uh let’s say I have an autonomy startup, where should my first conversation happen? Should it be with the DIU, with one of the departments? I’ll pretty much say yes to talk to anybody. And um I I the answer is all the above, right? Um my my recent tagline is uh you know, I can give you a lot of specific answers, but a $20 subscription to chat GPT will get you really close. I build this thing I want to sell to the Department of War. um who

7:23:29 are the program offices and by name who are the 06 you know PMS peeos paees that are responsible for those that will get you a hit list of exactly who to call that could be potential customers in the department it’s a really big department call all of them >> fair enough >> and us >> and you and that’s important real fast because when we do curation on projects when customers come to us and say we want to solve this problem it’s helpful for us to know what you’re working on because that helps me know what where is

7:24:01 the cutting edge, what should we be asking for, what’s realistic or what is actually, you know, four or five years out and that helps us ask for realistic things. So that that communication and that regular touch point is really important >> and you’re and you’re for the founder who’s who’s looking to get in touch, you’re based in Silicon Valley. You’re amongst the tech bros. >> Yep. We’re we are out there and this was exactly why Ash Carter established DIU’s headquarters in Mountain View, California. We’re actually coll-located with an Army National Guard unit uh right off Moffett Field in sight of the

7:24:33 Google headquarters and um [clears throat] you know that that’s deliberate. It’s meant to get us out of the Pentagon, out of the bureaucracy there and sort of keep a culture more aligned with the founders in this room than that in the Pentagon. If I’m a if I’m a founder trying to get in touch with the DIU, what should that first message say to increase the chance that I’ll be successful? >> I would I would say who who should I talk to? You know, you know our four

7:25:04 portfolios. Which one do you think you work in? I’d like to talk to someone from this portfolio. They have a really interesting option there. We also have a commercial engagement team who specializes in m managing and maintaining these contacts. Um, if you’re a YC company and you have any questions on exactly how to do that, I’m going to point out Eric Alborg whose book is here in the lobby as an insider who knows exactly how to and who to call for that. So, your reference is sitting there in the third row. >> Go talk to Eric. Yeah, >> sounds good. Um what have what have you learned at

7:25:37 DIOU uh looking back at the last 10 years about the most successful ways to take uh technology from a kind of prototype stage to on a program of record and and ultimately getting uh you know the department as a customer. Yeah, probably more of a personalized than a DIU answer, but I think the most important thing is that you when you have your product that you consider it within the broader ecosystem or the larger sister system of systems that you have to operate within. What I often see is, you know, companies have an innovative solution for a widget and

7:26:09 it’s cheap, it’s performant, I can make a million of them, it’s going to solve all my problems for my one more fighter use case. But that thing has a whole host of dependencies. You know, it has to be deployed from a UAV, which is to launch off of a USV, which has to have a ship there, which needs all these new commits that aren’t bought yet. Or it requires a balloon that the army is not actually going to buy. Maybe they will, but they probably haven’t. And so when you you tie the success of your thing to all these other dependencies, magically

7:26:39 shifting the department around your thing, you try to balance this whole system like an upside down triangle. It wedges itself into the ground. The project fails, the program fails, the product doesn’t transition, disappears, and there’s this chasm there. And we scratch our heads and we say the valley of death, right? And it’s not that anybody had it out for you. it’s that a lot of these things that are known known were not holistically or realistically considered upfront. And having those realistic viewpoints um and having that

7:27:12 that devil’s advocate there is absolutely critical. [snorts] Um don’t expect anybody to want your thing more than you. Your challenge is to get that champion within the department who is willing to carve out just enough space for you, but you have to realize the billions of investments that have happened around what you’re trying to do as well. >> If um if my product is really promising but doesn’t fit a current priority, what’s your advice? >> Um

7:27:42 there are enough problems within the department that somebody cares about it. Um the challenge is when we overindex on point solutions when there are existing solutions that are good enough or just a little bit better. There’s a switching cost. Yes. >> Every time we buy a new thing and we don’t account for that. If I’m 5% better, but the switching cost is astronomical. >> It may be a better product, but it’s not going to get adopted. The other piece

7:28:13 especially in software is you know you know point solutions for applications um even with AI integrations and interoperability remains one of the main issues and so if the solution is I buy a 100 postage stamp sized applications for realistically not a big set of users 100 thousand users um that’s very hard to manage and maintain um and so that’s why you see consolidation around you know larger programs, people taking off

7:28:43 bigger slices of that pie. So figure out how you can articulate your pie in a way that makes the adoption, the switching cost, the integrations worth it. Um, you know, we have one vendor for the Aegis combat system on every single ship. >> That is fractured and that still is where interoperability is one of the main issues that we are constantly chasing our tail on. So I would say you know think broadly about how you integrate into the ecosystem. What is the switching cost? How do you minimize

7:29:13 that or make the cost of it worth it? >> Let’s talk a bit more about AI. You before your current role I think you you oversaw a IML. >> Um I’m curious just to hear your perspective on on what you’ve seen as far as adoption goes. >> Yeah. Um I prior to this role I was stationed at at Paycom and the Paycom commander Al Paparo um has been one of the most forward-leading adopters of AI within the department. We’ve run several AI um uh summits out on Aahu and really

7:29:45 convened a lot of the leading thinkers in AI and the the lessons that we’ve taken for what it means to be an AI enabled command is that that there are we have to be explicit and specific about what we mean with we say AI. Um I think that the first premise is um power is always going to be our limiting constraint. So buy the best compute you can, the most inference efficient compute you can. Um you know applications and models will expand to fill that vacuum. Um you know without

7:30:16 getting into numbers I know that you know pre enduring freedom post enduring freedom the inferencing demand loads that we projected based on you know command post exercise versus what real conflict looks like is significantly different. And so this will continue to grow. I would say, you know, if we aren’t respecting the trend lines we’re seeing in the commercial sector, we’re we’re probably off base. >> So buy all the compute you can. Um and we are working very closely with the department’s AI arsenal in order to do that through a program called Titan

7:30:46 Core. Um next big is to treat that compute as an aggregate resource. distributed compute orchestration, implementing quality of service standards for how we um service the prompts that are coming out of our AI enabled software, whatever that is through whatever vendor. And as that gets implemented into more and more situations and decisions of consequence, it’s more important than ever to make sure that we are establishing, you know,

7:31:16 methods to implement AI assurance frameworks and benchmarking to those command critical workflows because today we build qualitative confidence in the outputs of these systems and that might be backed up to a cloud-based AI model. In the event that we experience a disruption in service and it must fail over to a local model or a different model or a different cloud provider are the outputs of that system still good and the answer might be I don’t know

7:31:47 unless I’ve managed a very deliberate development to deployment cycle where I can snapshot that and understand you know it are the recommendations trustworthy and that matters because the worst case scenario is that we have a war fighter >> that is receiving information believing it is good when in reality it it it may be that the model was not sufficient to provide the level of reasoning or insight or didn’t have the context when

7:32:17 required in order to answer the question that you got used to. So um that that’s a varsity level um you know thought process that we will get to. Um, I know this is a big problem in the commercial space because every company I talk to is struggling with the same thing because we’re living inside the exponential. So we we we will get there with everybody else. This is increasingly important why DIU remains so tightly coupled to those both on the frontier and leading special

7:32:49 a specialty AI development so that we can fast follow or co-ollow with you. >> Yes. Do do I mean talking about this kind of failing over from the cloud to uh wherever you happen to be locally are you using openweight models do are you are you guys thinking about these absolutely right there’s been a big push about sovereign owning your AI owning your model having confidence in that we’re working with the frontier labs we’re also working with people building special specialty models you know tuning

7:33:21 models specifically for the government or specific you know types of applications people who are distilling models for use cases on the edge and that only reinforces the you know need to understand what these models are appropriate and capable of. We took a uh well-known US uh open weights model um working with a um you know prime um not a prime vendor but working with a uh you know customer who’s delivering a lot of software to the department. A weekend of

7:33:54 tuning improved you know track detection from that model from something like 40 something% to 78%. And so that was just one weekend just trying it out. So that there is value there, but we also need to be very clear about what we lose when we do that. >> Yes. I mean, to be honest, it’s kind of hard to [clears throat] keep up with with how quickly the frontier is changing. And when I think about that and I and I put it in the context of large rollouts, you know, with with huge

7:34:24 departments, uh, where safety is is super important and, you know, the risk is the risk of hallucination, uh, it can be catastrophic, I imagine. H how do you maintain your sort of set of evals on every successive model in order to actually adopt what is what is available? So, DIU is um running a program called Mystic Depot. And Mystic Depot was meant to go after exactly this is how do you look at models, test them, understand, you know, what what do the

7:34:55 rejections look like? What what are the guard rails that are in place, you know, what might need to change? How do we tailor them for the specific solutions? Rails free is not the answer. That that creates an insider threat risk in the same way that it may enable on the same side, right? That’s a very >> sharp double-edged sword. >> And so, um, >> you don’t want a hugging face inside of the military. >> We would not want a hugging face inside of the military, right? And so that this brings in companies on the commercial side who are doing this for, you know, finance, healthcare, other regulated

7:35:27 industries where that also have consequence of the outcomes have data that needs to be protected as well. And so we we are working closely with RNA and others um in the intelligence community to make sure that we are you know at the forefront of where the commercial side is for the department. >> Last question um I guess along those lines when you go when you consider working with startups especially early stage startups how do you en like how do you enable them to work with the

7:35:58 military while maintaining your security standards your credibility standards your trust standards? >> Yeah. Um, quiz for the audience. Um, I gave out du.mill earlier. Did anybody uh see which solicitations are open right now? >> There’s a Yeah. What was one? >> The long duration ISR. >> Yeah, the long long duration ISR one. There’s another one which is maybe you flies maybe the radar a little bit. It’s called class act. This is a first of its

7:36:30 kind. It is, you know, another example of DIU innovating on existing authorities in the same way as we did with the cso, the prize challenge. What class act is is it is essentially soliciting in this case for companies that think that they’re doing something in counter unmanned systems. And the intention is to find companies that do not have a facility clearance, don’t have cleared folks, that have these hurdles in front of them to doing business with the department and to

7:37:00 evaluate you and say, “Look, I think that you’ve got a great technology. You want to do business with the department. We are going to bring you on board, fasttrack you, clear you, clear your staff, and knock down those bureaucratic barriers so that you now can compete equally shoulder-to-shoulder with the primes and the other defense tech startups that have 6, 12, 18 months head start on you. It’s an example of lowering those barriers in order to um you know create an ecosystem that is more set up for you

7:37:33 to compete on a level of technology and not on a level of bureaucracy compliance. >> Awesome. >> Yeah. Thank you so much. >> We appreciate this conversation. Thanks. [applause]

7:38:12 Good afternoon, ladies and gentlemen. My name is Sebastian Fischer. I am the CEO of Wardstone. We build modular interceptors. So, a little bit about us. We did YC. We graduated last fall. We’re exactly a year old. We raised $5 million after YC. We moved down to Los Angeles and we currently have eight people. Okay. A little bit about us. My younger brother is my co-founder. We have a retired twostar general as an adviser.

7:38:43 and Gary Tan, if he’s still here, is our YC partner. Okay, so what are we doing? We build interceptors, but I want to be very specific about the type of interceptors that we build. And so let’s take a quick look at this chart. So it’s a spectrum and on the left side of the spectrum, we have the easy to intercept targets. These are what are traditionally known as like group one to three UAS. Think DJI, think shaheds. These are things that, you know, they’re small, they’re easier to hit. Then as we move our way

7:39:15 towards the right, we start seeing things like cruise missiles, short-range ballistic missiles, and then all the way on the right, we have the buzzwords, right? Short-range hypersonics, ICBMs, and long-range hypersonics. Over the past about four years, this kind of left side of the spectrum, the small UAVs, has exploded in threat volume. If you look at Ukraine, if you look at Iran, this is really where there are an incredibly high number of threats. Industry has been very responsive. There are over 500 startups

7:39:47 that are building in this realm, kinetic solutions, non-kinetic solutions. Um, and they’re doing pretty well. There’s a lot of competitions out there to help these companies succeed. All the way on the right, you have the, like I said, the buzzwords. But in this arena, it’s extremely low volume. There are very few real world examples of hypersonics actually being used of ICBMs actually being used and the answer is because there’s existential reasons not to use these types of weapons. But what we

7:40:18 found is that in this what we’re calling the missing middle, there has also been an explosion of threats, right? Think Group 45 UAS cruise missiles, medium-range ballistic missiles, and industry has not yet responded. And so this is kind of where we saw an opportunity. If we’re going to start a company, we want to build an interceptor solution for what we’re calling the missing middle of missile defense. Okay. I am obsessed with talking to my customer. I have talked with very very many people in the department of war. I

7:40:50 have a counter. I tell my team where this counter is at. I tell my investors where this number is at. And so this is right. This is Missile Defense Agency, Army, Space Force. I asked them, “What are your unmet needs in the interceptor world? What keeps you up at night?” And so these are kind of the key trends and takeaways across these across these discussions. One, the aerial threat diversity is expanding. So the types of aerial threats that are being shot at US bases, at civilian infrastructure, it’s not

7:41:20 just one type of weapon anymore. The names of the things that are being shot at us is increasing every single day. The next is that not only is there more threat diversity, but these threats are also changing. If it’s in range, maneuverability, uh detectability, the threats that we need to shoot down are becoming increasingly complex to hit. And then lastly, this is almost like a theme of the day, the kind of the wars of today consume an incredibly large number of munitions and the US stockpile

7:41:52 of munitions is increasingly dwindling. And so I might be biased here when I say this, but I believe this is the number one heron fire issue that the military is facing today. Okay. So we are currently right, we have an active war in Iran and the US is supplying many many assets to Ukraine. We entered this conflict with the interceptors that we had available. These interceptors, right, there’s a saying in the army, no plan survives first contact. And so we’re using these

7:42:24 interceptors today in theater. And these interceptors were built with certain assumptions in mind. They were built with certain uh right threat assumptions, threat diversity assumptions, and also threat quantity assumptions. A lot of these assumptions are turning out to no longer be true. The threat diversity that these interceptors were built for was much more narrow than what we’re up against. The ability for these threats to change and our and our ability to adapt to the changes of these threats is another assumption that fell flat. And also the

7:42:55 quantity our stockpiles of interceptors and our production capacity of these interceptors is also not suited for the conflicts that we run into today. Okay. So what are we doing about it at Wstone? So again these come from conversations that I’ve had with my customer specifically the army. So we are introducing what we’re calling a modular interceptor platform that seeks to address these issues that I just raised one right adaptability uh threat diversity and also quantity. So this is

7:43:28 a 5 in diameter sorry 10-in diameter 5 m long interceptor and it has different modules that are swappable to in order to achieve different types of performance characteristics. So for example we can put a differentiz motor in there. We can put a different camera, a different seeker in the front. We can have different configurations of aeros surfaces to change our maneuverability and essentially to select different performance and price points. And so what does this modularity enable us to do? It enables us to align the cost and

7:44:00 performance of the interceptor that we’re building against the threats that we’re shooting down. We’re no longer we will no longer need to shoot multi-million dollar interceptors at tens of thousands of dollars threats. we can essentially size down the interceptor depending on what we’re up against. Two, we can insert uh technology faster. So, this solves the problem of if our enemy adapts, if they build something new, we can adapt as well. We don’t need to go through a whole whatever it is 5 to 10 year whole interceptor redesign, we can insert in

7:44:32 place our technology much much faster. And then lastly, this comes to the production capacity side. Um, we have kind So, we’re publishing our interfaces. So, we are a vertically integrated company, we build the motors, we are integrating our avionics, we are doing our own flight controls. But if there’s a supplier out there that builds a better product, we want to make sure that we make those interfaces public to those to those suppliers so that we can swap in and increase our production capacity when needed.

7:45:03 Okay, so this is, you know, this we’re not just a slide deck company. I want to show off some of the hardware that we build as well. So on the left here, this is our um aero aeromechanical system that we’ve built. This is our solid rocket motor that you can see outside as well. And we also build our own um guidance, navigation, and control software as well as a simulation environment to prove that the software works. I’m big on proving things in software. We’re never going to launch, we’re never going to push the launch button in real life until we’ve actually done the intercept in software.

7:45:34 This was our first interceptor that we flew this past April. So, this was a 5- in diameter, about 8 foot long interceptor. Uh we went to about 40,000 ft at Mach 2.5. Uh completely open loop, though. So, we had a Seeker taking photos. We had what’s called a divert and attitude control system, also operating open loop. And kind of what the next step is for us to do as a company is to close the loop, right? Uh we want to do something useful. We want to actually hit a target. So, these are kind of our our big milestones for the rest of the year,

7:46:05 right? One of them is due tomorrow, October 1st. Hardware Hills Intercept. Check. We did it a week ago. I feel good. Next one coming up, we have two hot fires, one in October, one in November out in the Mojave um right next to LA where we are. And then we’re actually doing an intercept of a live target in December. In terms of customer, I mentioned the Army. So, Army XTech, we were selected for what was called the lowcost interceptor competition. Uh, uh, there were a couple companies that were selected for this and we’re going out to

7:46:35 White Sands next week actually to kind of demo demo our technology to the Army in a live fire environment. Lastly, my ask of you is if you want to help solve this problem, if you’re a war fighter, if you are an acquisition officer, if you are an engineer, if you are an investor, these types of problems are solved alone by teams of eight people in a 5,000 foot lab in Los Angeles. We need to grow kind of across the board, across the technology stack,

7:47:06 across the, you know, capital capital situation as well. So, please do not hesitate to reach out. Thank you. [applause] [music]

7:47:37 [music] All right. Can you guys hear me? >> We can. >> Awesome. Yeah. So, um I’m the last one today, so hopefully you guys don’t fall asleep on me. I’m also super tired. So, if I look like I’m falling asleep, somebody over here throw a ball at my left shoulder. I dislocated it on Saturday. So, so today I’m going to be talking about two areas of application of no labs. uh

7:48:08 oneliner on us. What we do right now is we provide companies with data to train more lethal AI systems. The second idea and and kind of the more esoteric opportunity I see is basically applying these same mechanisms of the training data that we create uh to how we evaluate and test autonomous systems and AI systems. because I think there needs to be a lot of reform that happens there and a lot of the YC companies in the

7:48:39 audience I think are going to be very interested um just to figure out how to break into this space without having a huge name. Essentially like what you need to know about this slide is that uh my co-founder and I dropped out of Stanford um and worked at Google Demon previously. So machines have been uh finding targets for decades. In 2017, Project Maven standardized it, essentially putting computer vision on every single drone to identify targets.

7:49:11 We have it in geospatial satellites now. It’s the standard. Now, the problem is our adversaries are trying to counter this. And we’re seeing this a lot in Ukraine today. And I’ve got a really interesting example here where Russia has literally put like mirrors on one of their vans almost as like a dazzling effect so that their the Ukrainians drones in their AI is no longer able to detect uh valuable targets and be able to defeat them.

7:49:44 And so we’re kind of seeing this thing where deployed models that Palanteer are putting out or you know other contractors are essentially drifting and decreasing in performance over time as adversaries are adapting their own valuable assets to essentially not be able to be detected by our AI systems. So what happens? Uh the enemy changes its look next week. We’ve got zero images for us to train our new AI system

7:50:16 on. So, we’ve got two options here. Let’s collect data, which is like pretty much impossible to do. Like, how are we going to get our hands on the adversar’s target? And then we’ve got to maybe figure out how to book a range or flight time. Uh, wait for different times of day, weather, the right angles, and then fly that sensor. And then not even to talk about the labeling that we have to do afterwards, which I’m sure some of you know is actually very expensive to do. With null labs, what we can do is

7:50:47 essentially scratch all of that. We generate the exact data that you need, the exact target, your conops, the domain of interest, and the sensor of interest. And all the uh images are labeled automatically. To make this obvious, one 10,000 image data set, we generate that in 45 minutes. To collect that, that’s 2 to four weeks or even longer. This is just a collage to kind of

7:51:17 illustrate we do this across multiple domains. Uh I like to call some of the missions we apply to as almost zero tolerance. And so if you’re doing counter UAS base defense, why not exhaust every single possible data source, right? Synthetic data has a risk. You can overtrain on it. You you can, you know, over generalize onto synthetic data, maybe learn synthetic artifacts, but I think the bigger risk is deploying AI systems that are

7:51:48 failing, are not working, are not detecting our adversaries. And that’s currently happening in Iran right now. And so why not exhaust every single possible data source uh that we have access to. And you can see this is the realism that we’re working with across all domains from surface to space. And this is what I wanted to illustrate earlier with models drift. And so the performance of a model is going to dip

7:52:19 when the adversary adapts their target. And how do we get back up to speed? It’s definitely not with real data. We’re going to get that bottle model accuracy back up using no labs, using synthetic data. Now, I’ve been having some interesting conversations over the past few days with the other Defense Tech founders here, and it’s it’s been interesting. I mean, everybody loves to complain. So, everybody’s like saying, you know, procurement is broken, but why? and how

7:52:50 do we actually act on it? So right now there’s not really a common standard for testing AI and autonomous systems. So how are we supposed to you know on the other side I feel bad for these guys because how are they supposed to pick the right one? So oftent times they’re just going to go with the trusted vendor. Right now basically the same mechanism that we use to generate training data allows us to generate a lot of testing data as well and I’ll explain that a little bit

7:53:21 more but some of these images are um from current uh customer engagements with SOCOM as well as paycom. So what I’ve noticed is that you know any of us cold applying to the sibers you know to uh you know some of the challenges it’s always nada we’re not going to get anything back. Um and the real contracts that we’ve gotten are from knowing someone on the inside.

7:53:51 Right now, what I think is interesting is that we’re kind of using the same almost strategies and procurement methods that we used back in the 1900s of procuring like an aircraft carrier. But when it comes to an AI system, you you can’t really treat it in the same way. An aircraft carrier at the end of the day, you still have a human that is operating that system. Maybe a better example is like a tank. Uh you have a human who’s operating that system. So when it comes

7:54:22 to certain edge cases, uh it’s up to the human’s judgment of how they actually utilize the system. But when an AI system is in those edge cases, you don’t know how it’s going to perform and you don’t have that human input that is going to control like making the right decision. And so what happens and what I think is interesting here with the status quo is that we’re you know flying 20 drones in the air to evaluate counter US systems.

7:54:53 Um and then we we’ve down selected from the white papers to like 20 systems uh to basically go shoot them down and let’s see who’s who’s able to shoot down the most drones. The problem is what about all those different edge cases? You’ve tested it at one time of day. It’s clear skies, whatever it is. Um, 20 drones on in like the Mojave Desert, I don’t know, something like that. And essentially, we’ve only got we’ve only covered one aspect of the entire spectrum of the operational environment.

7:55:27 And so, how are we supposed to know h how is it going to perform when it’s a rainy day? How how’s it going to perform, you know, in Gaza or something like that when we tested it in Alabama? And oftent times, you know, some of these guys like, let’s say Jason Cornelius at at Perseus Defense, they don’t make it to that down select stage where they’re invited out to the field to show that their system actually performs. And you know, from the other side, we often

7:55:58 hear how there’s a contractor that’s selected for 24 months period of performance and they don’t deliver. And who else? There could have been somebody else that could deliver but we never gave them the chance. I think the digital also tests a lot what the range cannot. We can you know like do a ton of swarms at scale. uh like I mentioned we can validate in the Middle East in a different location and

7:56:28 we can simulate a lot of adversarial threats uh of tomorrow’s world and it’s more secure I think if um in the day of surveillance uh everything we do in the real world is under satellite surveillance and so moving towards a digital environment is kind of that first step and so no labs kind of is deployed across all the different steps of a

7:56:59 lifetime of a an AI system and a model like I mentioned generating, training it, testing and evaluating the system and then finally deploying it to the war fighter. My ask is simply just uh I’d love to connect with folks who are kind of interested as well as you know reforming uh some of these uh I I guess like maybe controversial reforms I think but making big changes in how we’re procuring

7:57:30 autonomous systems and how we’re training them. So that’s uh I want to feel the I want to help field the best systems and stay ahead of the threat. So, thank you guys so much for listening. [applause] All right, thank you everybody. Gary opened this morning with a practical question. What would it take for a small company to become a dependable supplier to the American military? We got some

7:58:02 really specific answers today. Emil talked about competitions with clear standards and actual orders. At the end, Secretary Duffy asked founders to bring bring him a s a solution to the problem and not just an interesting piece of technology. Under Secretary Hurst reminded us that getting a user excited and finding someone who can actually buy your product are two different jobs. And those are useful distinctions. They can save a small team like the ones that you’ve heard from today months. Before we wrap, a few thank yous. First,

7:58:34 thank you so much to the founders uh who came out from all over the world really. Um some coming in from Ukraine, Europe, uh many from San Francisco and and beyond. So, thank you for taking your time away from your companies and coming to Washington, especially the ones midbatch and showing people what uh what you’re building and what you need to succeed. A big thank you to Maven Advocacy and C2 Strategies who were sort of our quasi sponsors of this event and hosted some really uh generously hosted some dinners. If you’re a founder that’s thinking about engaging in Washington,

7:59:05 you should have conversations with both of those teams because they’re phenomenal. Um, uh, all the feedback I heard from the dinners last night was that they were a lot of fun. BGR, uh, especially Sean and Sid, they and and, uh, Alex and Tristan, they organized a, uh, Capitol Hill flyin where we had 50 founders, uh, going all over, uh, the Senate and House side, uh, yesterday, uh, meeting with, uh, policy makers. Uh probably the biggest thank you of all to the Ridgely Walsh team, especially um uh Jennifer Felton and uh

7:59:38 Lily Juliana Glover. Uh they did all the logistics for this event. Please give them a hand. Um [applause] and uh thank you to Eric, Olivia, Gary, the partners, the YC team that came out. Um, and to our uh catering staff, the security team, the AV folks, everybody at Gallup who participated in hosting this, we really uh appreciate it. Please do one more round of applause. Um, [applause] we’re going to keep the conversation going at uh Dirty Habit uh which I understand is just about a block

8:00:09 away uh at our our happy hour reception co-hosted by uh Nia and YC company called Lock. And uh just want to say thank you again for coming to Washington and let’s follow through on what we started today. Thanks. [applause]