Johannes Schaback

1. Key Themes

Production Is Deterrence: The Factory Is Part of the Weapon

Stark designs products for industrial scale from day one, rather than treating manufacturing as a downstream problem. Uwe Horstmann frames the factory itself as strategic capital: "how can we make sure that, you know, factory is part of the weapon. You know, production is deterrence." 00:01:57. He notes that visiting operational experts compared Stark's manufacturing complexity to "about as complex as a washing machine," which he took as the ideal outcome: "If I tell you, 'Hey, Pat, today, we're building 1,000 washing machines a month. We need to scale this to 10,000 a month or... to a million a year.' You would say, 'Okay. I mean, it's not easy, but it's doable somehow.'" 00:12:25. Johannes Schaback explains the discipline behind it: "we optimize for time-to-market, and second, we optimize for production readiness... we shift left on manufacturability. That's extremely important, because otherwise, you accrue a debt that you have to fix later on." 00:29:53

The Soldier Is the Customer, Not the General

Stark's core go-to-market insight is that the buyer (a high-level official) and the user (an enlisted soldier) are different people, and winning the user wins the product. Horstmann says: "the customer, oftentimes, is a relatively simple soldier, you know, on an operational, tactical level... We need their input. That's more important to us than anything else." 00:03:07. He adds: "the staff sergeants, uh the enlisted personnel, who then actually use those systems are maybe even more important to us, because in like in a true startup spirit, you want to be close to the customer." 00:05:22. The wood-versus-plastic box anecdote is the proof point: "We show up with this beautiful plastic box that is light to carry, and the soldiers look at us like, 'What do we do with this? Why don't you give us wood? Because wood, at least, we can... put in the fire, and it keeps us warm in the winter.'" 00:04:26

Venture-First Funding Instead of Government R&D Dependence

Stark funds its own product development with venture capital and treats defense like a startup portfolio. Horstmann: "be venture first, thanks to Sequoia and others, which basically means we fund the innovation, kind of the product development ourselves. We don't rely on some government R&D spending." 00:01:21. Stark runs "10 to 15 different programs, almost like a venture-like innovation portfolio" 00:03:27, and plans to invest roughly "€600 million, you know, significantly more than half a billion dollars" over 12 to 36 months, which he says is not far from the core, non-government-funded R&D of the big European primes. 00:36:23

Organizational Design: Startups-in-a-Startup Plus a Compounding Platform Layer

Schaback describes an engineering company organized into autonomous, cross-functional programs, with a shared platform layer of "capability teams." "We're organizing the company almost and running it almost like a venture fund, where we're building small startups in the startup." 00:13:28. The platform layer builds reusable capabilities: "GNSS denied navigation, or object classification, target re-identification, communication links... there are certain pieces that we then bring back as in compounding value into the family of systems, into the programs." 00:14:18. Function heads build "communities of practice" to share "tooling decisions, AI approaches, AI workflows, skills, very important, MCPs across the company." 00:21:45

Hardware-Software Co-Design as the Real Moat in Precision

Stark's claimed edge is tightly integrated hardware and software. Schaback: "just building a hardware system is good, but and just building a separate, isolated software stack that would be able to cope in the simulation... is not enough, but the actual sub-meter precision for striking a tank, or a moving target that moves at a speed of 60 km/h requires to fine-tune and get the settings and the parameterization in terms of compute, in terms of compute budget, a memory allocation, all perfectly well done." 00:15:17

Innovation Ecosystem, Not a Wonder Weapon, Is the Lesson of Ukraine

Horstmann argues the real takeaway from the Russia-Ukraine war is systemic: "the real learning from, you know, the Ukraine-Russia war is probably not drones, no drones. It is about an innovation ecosystem. That is the actual wonder weapon." 00:10:48. The ingredients: soldier feedback, talent ("our smartest minds should be willing to work in defense"), and capital. 00:10:57

Europe's Reality-Driven Procurement Change Is Durable

Horstmann argues European reform lasts precisely because it is not charisma-driven: "Changes like that in Europe, they don't happen because there is like a charismatic leader... Europe works based on, 'Well, I guess we kind of have to do it now. It's reality kicking in.' And the positive side of that would be hopefully changes are more lasting, because they are based on actual structural fundamental changes." 00:10:02. Procurement and certification moved from "years to months." 00:10:22

Vertical Integration vs. Plug-In Suppliers: Resilience and Localization

Stark's rule is customer-problem-first. Horstmann: "The governing principle for us is, how do we solve the customer issue? And the scalability of output is a big customer issue." A counterforce is selling to other countries: "we want to root the production of those systems in the countries in which they're being procured," so they must be able to "plug in their own suppliers." 00:26:58. Schaback adds a component philosophy: "we may not always buy... the greatest and like most leading-edge camera... but we buy those that are very very good, top quartile, but actually procurable in large numbers." 00:29:53

Cultural Shift: Tech Talent Is Moving Into Defense

Horstmann: "A few years ago, it would have been unimaginable for people to go into defense... we observe all over Europe... people understand, this is a generational challenge." Stark positions itself as a "good first touchpoint" because it is "as much a defense company as we are high-growth tech company." 00:22:43. Schaback says Stark recruits from "classical industry, but also at universities... startups, we also looking at governmental institutions." 00:20:54

Capability-Gap-Driven Multi-Product Expansion Across Domains

Stark expands by tracking militaries' self-declared capability gaps, then reusing technology across air and sea. Horstmann: "there is a capability gap, it is evolving, we try to contribute to the definition of that gap." 00:27:57. Reuse is real: "mission planning... doesn't matter if... the asset is flying or swimming. There is navigation in... denied environments where there is jamming... object detection, recognition. Again, the same." 00:28:59

2. Contrarian Perspectives

Factories and Contracts Deter More Than Capabilities Do

Most people think deterrence comes from superior weapons. Horstmann argues the real deterrent is pre-structured surge capacity: "what's really interesting is, can we come to a contractual design between the government and us... that already kind of like has everything in place for the massive surge in scaling that we need if there's ever hot conflict... maybe we think about this more in terms of like cloud contracts, you know, with surges... That is probably where the real deterrence comes from. It's not about the individual capabilities of one product here or there." 00:39:50. He sketches the details: prepaid manufacturing capacity, storage, and software that "probably should be updated every week with new training data." 00:40:53

The Wonder Weapon Is Not the Drone

Against the breathless internet coverage of new weapon systems, Horstmann says: "We always like to read on the internet like what new weapon system, how awesome is it. That doesn't really matter. The whole point is, how do we, as Western society, set up ourselves so that we can continuously innovate in defense." 00:10:57

Reliance on Government R&D Funding Is a Weakness

The established defense model runs on government-funded development. Stark does the opposite and says that being venture-funded is why they can move fast and control the roadmap: "We fund the innovation... ourselves. We don't rely on some government R&D spending." 00:01:21. Combined with the roughly €600 million planned spend, he claims they are "not way off" the core self-funded R&D of the big primes. 00:36:23

Cheap, Mass-Producible, "Good Enough" Components Beat Leading-Edge Ones

Counter to the exquisite-systems tradition, Schaback deliberately avoids the best components: "to the analogy of the washing machine, the greatest and like most leading-edge camera, for example, but we buy those that are very very good, top quartile, but actually procurable in large numbers." 00:29:53. He also uses the Formula 1 analogy: "you only build one... you hardly ever produce them in large quantities." 00:29:53

Procurement Is Not a Blueprint Problem, and No One Has Solved It

Many assume the West has defense procurement figured out for modern systems. Horstmann is blunt: "I don't think so necessarily. There's not a blueprint... There's not like hundreds of people who have done this and now they can just teach us. I think we all learn together." 00:41:28

3. Companies Identified

Stark

A European defense "neo-prime" building AI- and software-driven strike drones, maritime drones, and related systems, with about 700 employees after two years. Described as landing "billions of dollars of contracts within 18 months." Pat Grady: "you guys landed billions of dollars of contracts within 18 months of starting the company, which I believe is unheard of in the defense business." 00:00:48. Schaback on growth: "we went from 0 to 700 people in 2 years" (Luciana Lixandru, 00:37:37).

Sequoia Capital

Venture investor and backer of Stark. Why mentioned: it enabled the venture-first model. Horstmann: "be venture first, thanks to Sequoia and others, which basically means we fund the innovation, kind of the product development ourselves." 00:01:21

Bundeswehr (German Armed Forces)

A customer that pushed Stark in a high-stakes public test. Schaback: "the Bundeswehr, like the German Bundes army, asked us to participate in a very important experiment. And so, we couldn't say no to that." 00:16:43. Horstmann also credits the Germans and other governments for procurement change: "we can give a lot of credits to the Germans and other governments who also understood... we also need to change the process of how we procure, because you don't procure a drone like you procure a battleship." 00:08:42

NATO

A standards body and customer base. Why mentioned: usability and certification targets, and responsible AI practices. Horstmann: "really usable for NATO forces." 00:01:39. Schaback: "Stark has, since the beginning, implemented NATO's principle of responsible use of AI." 00:33:06

Ukrainian Armed Forces / Ukrainian Units

The primary source of real-world feedback. Why mentioned: access to the battlefield. Horstmann: "We do that with a large presence in Ukraine... we're integrated in those units quite extensively." 00:02:29. Schaback: "the war in Ukraine is changing daily." 00:26:13

Established European Defense Primes (unnamed)

Incumbent defense contractors. Why mentioned: they excel at exquisite, expensive systems, and Stark welcomes competition with them. Horstmann: "They're really good at building those exquisite, very important, very expensive systems. And I think they still will play a big role." 00:12:29. On their R&D: "if you subtract all the R&D spending that is government-funded... we're not way off." 00:36:23

Starlink

Satellite connectivity used in Stark's improvised field test base. Schaback: "we had like at the time, several containers set up. We had even mobile generators, Starlink." 00:16:43

Hugging Face

Mentioned by Pat Grady in the context of an AI agent incident. Grady: "we're kind of sitting here in the wake of the Hugging Face hack, where these civilizations of agents, so to speak, broke out of their sandbox and wreaked havoc." 00:32:22

4. People Identified

Uwe Horstmann

Co-founder of Stark; former venture capitalist in Europe with a military background. Why mentioned: he invested in defense before it was obvious. Luciana Lixandru: "you saw the opportunity in defense before other people. You were investing in defense way before it was obvious, I would say, and you also have a background in the military yourself." 00:07:27. He frames his path as converging "interest in military and security... and then the technology part." 00:09:28

Co-founder of Stark, leading engineering. Why mentioned: he built the program-and-platform organization and led the 2-week-plus turnaround sprint. His account: "That was a forging moment for Stark. Stark, in many regards, was born out of fire." 00:16:32. On the core of the company: "At its very core, Stark is an engineering company. It's everything we do." 00:13:28

Luciana Lixandru

Sequoia partner and investor in Stark; early peer of Horstmann from European venture. She asks the questions that surface the company's manufacturing and primes positioning: "Can we talk about a topic that may sound boring on the surface, but it's really make it or break it for a company like you, which is manufacturing?" 00:29:40

Pat Grady

Sequoia partner and host of Training Data. He surfaces the AI-safety and vertical-integration questions: "How do you guys think about where you need to be vertically integrated, and where you want to leverage the ecosystem or the supply chain that's already in place?" 00:26:50

The Enlisted Soldiers and Staff Sergeants (unnamed)

Why mentioned: Horstmann treats them as the most important customers for product definition: "the staff sergeants, the enlisted personnel, who then actually use those systems are maybe even more important to us." 00:05:22

5. Operating Insights

Win Trust by Promising a Faster Iteration Loop, Not a Better Pitch

When soldiers called the product bad, Stark converted that into loyalty by promising rapid return visits. Horstmann: "sometimes we show up and they say like, 'You know what? This is absolute shit.' But you can really win trust by saying, 'Tell us how exactly does it shit and we'll show up again tomorrow or the day after tomorrow. We create a new version, and will be a little bit less shit tomorrow.' And that goes a long way." 00:06:21. Treat the user's time and feedback as the most valuable thing they give you.

Define the "Make-or-Break" Moment Explicitly and Commit Without Budget Constraints

After the failed public test, Schaback called an all-hands-on-deck effort with a clear exit criterion: "we said that we will only leave this place if we had three consecutive spot-on strikes... no questions asked, you know, money doesn't matter. We have to get to the precision." 00:19:00. They ran three shifts, improvised infrastructure (containers, generators, Starlink), and co-located flight test, perception software, hardware actuation, and simulation teams. The binary completion criterion plus removing the cost debate collapsed the timeline to under 4 weeks.

Run Hardware Like a Venture Portfolio With Shared Compounding Assets

Autonomous programs with their own P&L-style missions, plus capability teams whose outputs flow back to every program, let engineers choose between frontline customer work and platform work. 00:14:18. Cross-domain reuse (air to maritime for mission planning, denied-environment navigation, detection, homing) means each new program starts with inherited capability. 00:28:59

Build Safety Into the Architecture Early, Borrowing From Adjacent Industries

Schaback: "hard-coded constraints that, depending on where the system is, for example, if it's flying over a particular inhabited area, it may not arm... IEC 61508, and several things we borrowed from autonomous driving." 00:33:06. Treating certification as a product feature ("a very boring topic in a way, but it's very very important," Horstmann 00:10:22) is part of what cut procurement from years to months.

Make Cost-Effectiveness a First-Class AI Planning Output

Their ground control system Minerva decides how many systems to use against a threat: "how many systems do I shoot at this target? What's the most cost-effective way of addressing this threat?" 00:34:51. Embedding economics in the autonomy and planning layer is a design pattern worth copying in any attritable-asset business.

6. Overlooked Insights

Software Has a Weekly Update Cadence While Hardware Has a Three-Year Life, Which Breaks Traditional Defense Contracts

In passing, Horstmann notes: "if you look at a drone at the moment, probably, you know, the motors they're good for 3 years, but a software probably should be updated every week with new training data." 00:40:53. This implies defense contracts must shift from one-time hardware sales to recurring software and data subscriptions with surge clauses, and that the long-run value capture sits in the update layer, the training data from the field, and the sustainment contract. This is the SaaS-ification of munitions, and the company that owns the feedback loop, with the data flowing back from users and live conflicts, gets a compounding advantage.

Shelf Life and Logistics Are the Hidden Product Constraints

Schaback briefly mentions that a system may "have a shelf life of, say, 10 years, until it actually gets deployed to the front line," and must survive "shocks and vibrations, the pressure," transport by air or sea, and be disassembled and reassembled. 00:24:06. Combined with the wooden-box anecdote, this suggests a durable moat in logistics-aware, storable, long-shelf-life design, an area where software-first entrants often underinvest and where stockpile-based deterrence models will force competition. Whoever masters stockpilable, long-shelf-life autonomous munitions plus remote software refresh has a distinctive position that pure-play AI or drone startups lack.