Strategy in the Face of Chaos
Audio companions to my writing on strategy, technology, AI, cybersecurity and building technology businesses.
Each edition explores one of my published articles through an AI-generated discussion or debate, offering another way to engage with its central ideas. These are not interviews or original podcast episodes, and the voices are not mine. The written article remains the definitive version.
This channel is currently a pilot, and the format will evolve as I learn what works.
Strategy in the Face of Chaos
Overcoming Innovation Barriers: How Anduril’s Crucible Tackles Scale, Collaboration, and Data Sharing
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
This pilot audio edition explores the argument behind Overcoming Innovation Barriers: How Anduril’s Crucible Tackles Scale, Collaboration, and Data Sharing.
The discussion examines the persistent barriers that prevent organisations from turning promising ideas into scalable products. These include fragmented data, weak coordination between disciplines, growing systems complexity, integration challenges and the difficulty of moving beyond prototypes without sacrificing quality or speed.
It explores how Anduril’s Project Crucible addresses these problems through six-week development cycles, mission-focused objectives, integrated testing and continuous feedback. Design Reference Missions align teams around operational outcomes, while realistic test exercises expose dependencies and performance issues before deployment. A regular release cadence helps the organisation combine rapid iteration with greater discipline and coordination.
The episode considers how Crucible turns complexity into a potential competitive advantage. Its lessons extend beyond defence technology to industries such as healthcare, manufacturing and renewable energy, where organisations must also coordinate specialist teams, test complex systems and scale innovation reliably.
This is an AI-generated discussion based on my published article. It offers another way to engage with the ideas, but it is not an interview or a recording of me. The written article remains the definitive version.
Imagine pouring, I don't know, fifty million dollars into a revolutionary AI-driven drone prototype.
SPEAKER_00Right. Huge investment.
SPEAKER_01Yeah, exactly. Your engineering team has spent two years perfecting the flight algorithms. The aerodynamics are flawless. And you are, you know, just weeks away from a massive product launch.
SPEAKER_00The dream scenario.
SPEAKER_01Well, until the final integration phase, that's when you realize the marketing team promised a payload size that the engineers physically cannot fit into the chassis without uh burning out the battery in five minutes.
SPEAKER_00Oh man, yeah. That is brutal.
SPEAKER_01And this isn't a failure of imagination, right? It is a failure of the innovation supply chain, which is exactly what we're getting into for today's deep dive. So welcome in if you're a technology or business leader tuning in. Uh this one is specifically for you.
SPEAKER_00It really is the nightmare scenario for any executive. But the thing is, it happens constantly. Trevor Burrus, Jr.
SPEAKER_01Constantly. And our mission today is to deliver a really rigorous, fluff-free executive briefing on a fantastic article by Victor Holman titled Overcoming Innovation Barriers.
SPEAKER_00Yeah, Holman really strips away the hype with this piece because you know everyone knows innovation drives growth. But we have this culturally ingrained, almost romanticized view of it.
SPEAKER_01Like the lone genius in a garage.
SPEAKER_00Exactly. We picture the lone genius having his sudden eureka moment, writing an equation on a whiteboard and just disrupting an entire industry overnight. Trevor Burrus, Jr.
SPEAKER_01Right. But the reality is way messier.
SPEAKER_00Totally. You look at the reality of modern enterprise, and that whiteboard is actually a grueling, deeply interconnected logistical network. It's a supply chain of ideas.
SPEAKER_01And Holman's piece is so vital because he treats innovation strictly as an engineering problem that has to be managed, not some spark of magic. So we're gonna unpack these highly documented systemic friction points that actually kill projects. Things like data silos, cross-functional disconnects, systems complexity.
SPEAKER_00And scaling too, plus the terrifying realities of supply chain security.
SPEAKER_01Right. We'll get into all of that objectively and then examine how the defense tech company Anderl uses a specific framework called Project Crucible to systematically dismantle these exact barriers.
SPEAKER_00Which is a brilliant case study. But you know, you can't even get to the point of a chathy payload problem if your teams don't have the raw materials to start building in the first place.
SPEAKER_01Yeah, the absolute foundation.
SPEAKER_00Right. And the most critical raw material we have right now is data. It's human collaboration and data.
SPEAKER_01Which brings up the first major barrier Holman identifies. Massive amounts of data are effectively trapped. It's wild. We are in the data age, but organizations are drowning in information they literally cannot use because of strict privacy laws, proprietary hoarding, or just basic interoperability failures.
SPEAKER_00The interoperability thing is huge.
SPEAKER_01Yeah, he gives some really stark examples, like AI-driven drug discovery. You have researchers trying to identify life-saving chemical compounds using molecular data sets.
SPEAKER_00But those data sets are scattered across competing pharmaceutical systems.
SPEAKER_01Exactly. Or they're locked down by patient privacy regulations. So a researcher might spend six months just begging a legal department for access. The whole process slows to a crawl. Medical breakthroughs are delayed simply because digital systems aren't allowed to talk to each other.
SPEAKER_00And you see the exact same wall in the renewable energy sector. Researchers are trying to optimize national smart grids for fluctuating power sources, like wind and solar.
SPEAKER_01Which requires granular real-time consumption data, right?
SPEAKER_00Right, but that data is held by dozens of different regional utility companies, and none of them use standardized protocols. So the predictive models just fail.
SPEAKER_01So how do you fix that?
SPEAKER_00Well, Hallman points out that it requires a multifaceted solution, universal standards, secure yet open infrastructure, and public-private partnerships. It really takes a heavy executive mandate.
SPEAKER_01Which actually brings up a solution he discusses that honestly fundamentally changes the game: federated learning models.
SPEAKER_00Yes, federated learning is fascinating.
SPEAKER_01When I first read about this, it sounded like a paradox. Training AI collaboratively without exchanging any raw data. But it makes total sense if you think of it like uh sharing a recipe without giving away your secret sauce, or actually think of it like sending a student to a highly restricted library.
SPEAKER_00Aaron Powell Oh, I like that analogy. Break that down for me.
SPEAKER_01Well, the student, which is the AI algorithm, needs to learn from a rare, highly classified book. But the library rules dictate that the book can never leave the building.
SPEAKER_00Aaron Powell Okay, so the data stays put.
SPEAKER_01Exactly. So the student goes in, reads the book, learns the underlying lessons, and then leaves. They don't take the book with them, they just bring back the knowledge they gained, the updated neural network weights, basically, and share that new intelligence with the rest of the class.
SPEAKER_00Aaron Powell That is a great way to put it. The raw data stays completely localized and secure, but the system as a whole gets smarter.
SPEAKER_01Right. But my question when reading this was is this actually working in the real world yet?
SPEAKER_00Oh, absolutely. The brilliance of that localized training is it completely bypasses those legal bottlenecks. And Holman provides concrete proof points. Like in material science, we're seeing the impact of the materials project.
SPEAKER_01Oh, right. I remember that part.
SPEAKER_00Yeah, they use open databases and AI models to rapidly accelerate the discovery of sustainable materials for cleantech. And they do it without forcing chemical companies to hand over their proprietary IP.
SPEAKER_01He also mentions a platform called GYME in the agricultural sector, doesn't he?
SPEAKER_00He does. GYME is a perfect example of overcoming interoperability issues. It aggregates remote sensing data to map field-scale crop yields. So farmers are pooling data on soil health, hyperlocal weather patterns, all of that.
SPEAKER_01To boost productivity and reduce fertilizer waste.
SPEAKER_00Exactly. Federated learning and secure sharing protocols are actively unlocking collaborative innovation that was honestly legally or technically impossible just five years ago.
SPEAKER_01Wow. Okay, so let's transition smoothly here. Say you implement federated learning. You've completely solved the data bottleneck. Perfect data is still totally useless if the human beings using it are misaligned, right?
SPEAKER_00Oh, 100%.
SPEAKER_01This is the second barrier Hallman talks about cross-functional RD coordination, which brings us right back to my drone example from the start of the deep dive. Getting research and development, marketing and operations to actually speak the same language.
SPEAKER_00It's notoriously difficult because historically those departments inhabit completely different realities. And engineer is optimizing for technical performance, right? Thermal efficiency.
SPEAKER_01While marketing is optimizing for user desires, aesthetic appeal, competitive feature lists.
SPEAKER_00Exactly. And meanwhile, you have operations agonizing over margins, supply chain sourcing, and manufacturability at scale. When those three realities don't intersect until the final stages of a project, you get the hurting cats phenomenon. The friction inevitably derails the product.
SPEAKER_01So to prevent that hurting cats issue, Holman details the necessity of cross-functional teams, CFTs, and combining those with agile methodologies.
SPEAKER_00Right. Now agile is a term that gets thrown around a lot, usually just as a corporate synonym for working faster.
SPEAKER_01Yeah, a buzzword.
SPEAKER_00But Holman focuses on the actual mechanical function of Agile, breaking massive projects into highly structured iterative phases to force continuous feedback loops. Canalize 3M is a prime example of this.
SPEAKER_01Right. A massive legacy corporate structure that leverages CFTs to basically mimic the agility of a startup.
SPEAKER_00Because that mechanism of continuous feedback is what actually dismantles the silos. If you have a cross-functional team running on a structured two-week sprint, the engineer doesn't find out about the marketing payload requirement in week 40.
SPEAKER_01They find out in week two.
SPEAKER_00Exactly. When the cost of pivoting is negligible, you're structurally forcing the engineer, the marketer, and the supply chain manager to negotiate their constraints continuously rather than sequentially.
SPEAKER_01But the stakes of those negotiations are so much higher now, aren't they? Because of what these teams are actually building. Once your teams and data are aligned, the products themselves are more complex than ever.
SPEAKER_00Oh, absolutely.
SPEAKER_01We aren't just stamping out a plastic widget or writing a standalone mobile app anymore. Modern innovation is overwhelmingly dominated by cyber physical systems, which is the third barrier, Holman outlines.
SPEAKER_00Yeah, cyber physical complexity completely changes the engineering paradigm. You are no longer managing hardware and software as two separate tracks. Think about an autonomous vehicle.
SPEAKER_01Right, it's not just a car.
SPEAKER_00It's a network supercomputer on wheels. It relies on real-time, flawless coordination between physical LIDAR sensors, complex AI decision-making algorithms, mechanical steering actuators.
SPEAKER_01All operating with zero latency while hurtling down a highway at 70 miles per hour.
SPEAKER_00Exactly, which requires an entirely new way of designing products.
SPEAKER_01And this is where Holman introduces model-based systems engineering, MBSE. The way he describes it, it allows engineers to create a highly sophisticated digital twin of the product. They can fully visualize and simulate all those complex system interactions in a virtual environment before a single piece of physical material is cut.
SPEAKER_00Yes, the aerospace industry uses this extensively to drastically reduce late-stage design flaws.
SPEAKER_01Aaron Powell But let me play devil's advocate for a second here. Doesn't simulating everything before you build just lead to paralysis by analysis? Why not just build a quick prototype, fly it, see if it crashes, and iterate?
SPEAKER_00Well, that cultural resistance you're describing is actually one of the primary hurdles Holman explicitly addresses. Traditional engineering teams, particularly those steeped in the Silicon Valley, move fast and break things. They often view MBSE as way too slow.
SPEAKER_01Because it is resource heavy, right?
SPEAKER_00Extremely. Yeah. It requires massive upfront investment in new software tools and demands rigorous training.
SPEAKER_01So why do it?
SPEAKER_00Because in complex cyber physical realms like aerospace, autonomous transportation, or defense moving fast and breaking things means catastrophic failure. Breaking a piece of social media code means what, a user can't log in for five minutes.
SPEAKER_01Right, minor inconvenience.
SPEAKER_00But breaking a cyber physical system means a jet falls out of the sky. The strategic necessity of MBSE is that it catches the catastrophic integration failures when they are just pixels on a screen.
SPEAKER_01Not twisted metal on a runway.
SPEAKER_00Precisely. And Holman provides the operational antidote to the paralysis by analysis concern. You don't just stay in the simulation forever. You pair MBSE with what he calls integrated test exercises, ITE.
SPEAKER_01Which is regular end-to-end physical testing.
SPEAKER_00Yes. You physically test whatever components are ready throughout the development lifecycle, and you feed that real-world data back into the digital twin to refine the model. It's a strategic necessity for fields where failure literally isn't an option.
SPEAKER_01Okay, so you've got your federated data, your cross-functional teams are communicating flawlessly, you've used MBSC to design a perfect cyber physical prototype in a simulation.
SPEAKER_00Things are looking good.
SPEAKER_01Right. But now you face what Holman describes as the scaling chasm: moving from a functioning prototype to sustainable large-scale production. Because producing it sustainably at scale without exposing your company to cyber threats is the ultimate executive test.
SPEAKER_00It really is where a lot of brilliant ideas simply evaporate. A prototype merely proves a concept obeys the lies of physics. Scaling proves it obeys the laws of economics.
SPEAKER_01And it requires a massive shift in operational discipline, balancing rapid growth with adaptability.
SPEAKER_00Exactly. You have to establish reliable supply chains, allocate immense capital for manufacturing infrastructure. So to bridge that chasm, Holman unpacks the strategic use of additive manufacturing.
SPEAKER_01Which is basically 3D printing on an industrial scale, right?
SPEAKER_00Yeah. It enables companies to move from low volume prototyping to scalable production without needing to build a billion-dollar factory first. You just need the advanced printers and the digital CAD files.
SPEAKER_01He notes this is particularly vital for aerospace and healthcare, where you might need highly specialized, complex parts, but you don't necessarily need them in batches of 10 million.
SPEAKER_00Right. It drastically reduces time to market because you eliminate that massive traditional tooling phase.
SPEAKER_01He also emphasizes circular economy principles for scaling, doesn't he?
SPEAKER_00He does. As environmental and regulatory constraints tighten globally, scaling has to incorporate circular economy principles. Like in the electronics sector, designing products from day one with modularity in mind.
SPEAKER_01Meaning you design, say, a smartphone or a smart grid component so that the battery or the processor can be easily swapped out and upgraded rather than throwing the whole unit in a landfill.
SPEAKER_00Exactly. It maintains scalability and product lifecycle value while reducing the environmental footprint. He also mentions using digital twins for supply chain predictive analytics to reduce costs and emissions.
SPEAKER_01But Holman adds a major organizational caveat here, doesn't he? About the culture clash.
SPEAKER_00Yes, very important one. The culture required to rapidly prototype is fundamentally at odds with the culture required to scale. Teams that are brilliant at agile, unstructured iteration often just reject the rigid operational discipline required for manufacturing.
SPEAKER_01So how do they mitigate that?
SPEAKER_00He advises phased rollout strategies and ecosystem collaborations like grid scale battery networks. You scale gradually to stress test the operational systems before you flood the market.
SPEAKER_01But let's pivot to the dark side of all this rapid-scaled innovation because building this massive interconnected cyber physical machine introduces a terrifying vulnerability, the sock door supply chain.
SPEAKER_00Oh, this is a huge issue right now.
SPEAKER_01You've just connected your entire RD supply chain, your suppliers, your digital twins, to the outside world. Rapid iteration and third-party dependencies create vulnerabilities that are actively targeted by nation states and cyber criminals.
SPEAKER_00Right. Because sophisticated adversaries understand that hacking a massive corporation directly is incredibly difficult. But modern software relies heavily on a sprawling web of open source libraries and third-party dependencies.
SPEAKER_01So if they find a weak link.
SPEAKER_00If they can compromise a tiny piece of code maintained by a small vendor in that supply chain, they gain backdoor access to every major enterprise that uses it. They can steal IP, insert malicious logic, or disrupt physical operations before the product even ships.
SPEAKER_01It's the ultimate security tax on agility. So what is the executive mandate here? How do you lock this down without killing the speed of innovation?
SPEAKER_00Well, security can no longer be a final audit that happens a week before launch. It has to be embedded. The mandate is implementing secure by design principles from the very beginning.
SPEAKER_01Which involves zero trust architectures, right?
SPEAKER_00Yes. Zero trust means no user or system is trusted by default, regardless of whether they are inside or outside the corporate network. You also have to enforce cryptographic integrity checks on all code.
SPEAKER_01And he also mentions end-to-end artifact provenance, which, you know, is a bit of a dense term. But essentially we're talking about a digital birth certificate for every single line of code and every physical component, right?
SPEAKER_00That's a perfect way to describe it. It's a verifiable record that travels with the component, proving exactly who wrote it, where it came from, and guaranteeing that it hasn't been altered at any point in the development pipeline.
SPEAKER_01So by automating these security protocols directly into the life cycle, you maintain speed without exposing the company to catastrophic risk.
SPEAKER_00Exactly.
SPEAKER_01Okay, so we establish all the theoretical barriers and the theoretical solutions. But how does a highly complex organization actually operationalize all of this at once? This brings us to the core case study of Holman's piece Androil RS Project Crucible.
SPEAKER_00Right. And Dural operates in defense technology, which is arguably one of the most complex, high-stakes, and highly regulated industries on the planet.
SPEAKER_01And Holman frames Project Crucible strictly as an AI-driven, structured approach designed to break down silos and support rapid iteration. They operate on these strict six-week sprint cycles.
SPEAKER_00Which is such an interesting time frame.
SPEAKER_01It is, because six weeks seems to be the exact sweet spot. It's long enough for an engineering team to actually build a meaningful feature, but short enough that if a team starts drifting, management can course correct before millions of dollars are wasted.
SPEAKER_00And Crucible's core mechanisms map directly back to the barriers we just discussed. Take their design reference missions or DRMs.
SPEAKER_01Right. How do those solve the hurting cats problem?
SPEAKER_00The DRM service shared authoritative blueprints that align all cross-disciplinary teams around specific mission-driven objectives.
SPEAKER_01So instead of an engineer just being handed a spec sheet that says, uh, build a thermal sensor with X resolution, they get the holistic mission blueprint. The DRM says, this sensor needs to identify unauthorized vehicles in a desert environment at night while integrated into this specific drone chassis, communicating over this specific network.
SPEAKER_00Exactly. It ensures that the software engineers, the hardware designers, and the AI specialists are all pushing toward the exact same real-world outcome from day one. It entirely eliminates the ambiguity that breeds those departmental silos.
SPEAKER_01But a blueprint is only as good as its execution, right? Okay. Which is why Crucible mandates realistic testing, mirroring the ITE concept we talked about.
SPEAKER_00Right. They do not rely solely on sterile laboratory conditions or digital simulations. They simulate actual operational conditions as early as possible.
SPEAKER_01Aaron Ross Powell Like testing drone navigation systems in real adverse weather.
SPEAKER_00Yes, to catch those subtle cyber physical integration issues early on.
SPEAKER_01And all of this culminates in their closing bell event. I absolutely love this concept. It's a synchronized, all hands live test at the end of the sprint. To me, this sounds exactly like a theatrical dress rehearsal.
SPEAKER_00Well, the parallels there are very strong. How so?
SPEAKER_01Well, in a theater production, the lighting crew, the actors, the set designers, they all practice their specific crafts in isolation for weeks. But during a dress rehearsal, everyone is forced into the same room to run the show live.
SPEAKER_00Right, and they finally see how it all interacts.
SPEAKER_01Exactly. If an actor misses their mark by two feet, the lighting cue completely fails, and the sound effect happens in the dark, it forces an immediate brutal debrief where the teams have to fix the integration points.
SPEAKER_00That is exactly what the closing bell event does for engineering. It forces holistic product readiness and shatters any remaining silos because there is literally nowhere for a failing component to hide.
SPEAKER_01And once that dress rehearsal is successful and the prototype works flawlessly, how do they deliver that complex system at scale?
SPEAKER_00That is where Crucible Lis release train system comes into play. It applies lean project management principles and AI-driven analytics to maintain operational cohesion.
SPEAKER_01Holman uses the example of iteratively updating defense surveillance towers in the field, right?
SPEAKER_00Yes. The release train ensures that software and hardware updates happen frequently and reliably across a distributed network of physical assets. It's how they maintain the agility of a startup while scaling seamlessly.
SPEAKER_01It's incredible. They've essentially transformed systemic complexity from a barrier into a distinct competitive advantage.
SPEAKER_00They've engineered a supply chain of innovation that inherently neutralizes the friction points that stall their competitors.
SPEAKER_01Which brings us to the ultimate executive takeaways for you, the listener. If you are steering the ship of a modern enterprise, innovation requires treating your organizational structure, your data standard protocols, and your security as primary engineering challenges.
SPEAKER_00Not administrative afterthoughts.
SPEAKER_01Right. If you don't engineer the structure correctly, the product will inevitably fail.
SPEAKER_00And Holman's ultimate conclusion is that the lessons from Project Crucible extend far beyond the defense sector. The principles of cross-disciplinary integration, realistic iterative testing, secure by design scaling, they provide a highly viable roadmap for any industry.
SPEAKER_01Healthcare, heavy manufacturing, logistics, renewable energy.
SPEAKER_00Exactly. They can all leverage these structured frameworks to turn the complexities of scaling and collaboration into a massive competitive advantage.
SPEAKER_01It's all about balancing relentless speed with immense operational discipline. Well, thank you so much for joining us for this deep dive into Victor Holman's work.
SPEAKER_00It was a great discussion.
SPEAKER_01But before you go, we want to leave you with a final unscripted thought to mull over, something that builds on the material we've explored today.
SPEAKER_00Let's hear it.
SPEAKER_01We started this discussion by talking about that romanticized image of the lone, chaotic genius in a garage. But if highly structured, AI-enhanced, relentlessly monitored frameworks like Project Crucible become the required standard for successful innovation. Does this mean the era of the chaotic lone genius startup is officially over? If innovation is now a meticulously managed, deeply secure supply chain, how do you ensure you don't accidentally engineer the actual unpredictable creativity out of the process entirely?
SPEAKER_00That is a really tough question.
SPEAKER_01Something for you to explore on your own. Thanks for listening, and we'll see you next time.