Shyam Sankar on Palantir, Reindustrialising America, and AI in Warfare

Shyam Sankar

Well, I think they're really founders. They're founder figures, and they get obsessed with delivering something that makes no sense, because they're fighting — in particular in the military context, you're fighting the bureaucracy, you're going to end your career, you pay extreme prices for it. So there's almost a pathological obsession with winning, which I really relate to personally, but which I also see in great founders.

In the present moment it's really important to encourage the hidden heretics who are in the military today, to recognise: your heresy matters. Frankly, if you look back at history, the only things that ever worked, the things that helped us win all the wars, were the things the heretics actually did. Nothing that went through the machine delivered anything. All change comes from these heretics, and they only later become heroes. Think about Billy Mitchell, who invented the Air Force — he was court-martialled and died penniless and depressed, and only after his death did we build the Air Force based on his original heresy. I'm just fascinated with what motivates these folks to keep doing this. And I relish a little in their rebellion and the victory that the rebellion brings about.

Shyam Sankar

I'd probably pick Rickover's. Hyman Rickover was born in a shtetl in Poland and came to the US when he was six. He had one of these near-miss moments: in those days, when you landed at Ellis Island, you had 10 days for someone to come pick you up. His mother gave someone money to send a telegram to the father, who was already here, and that guy just pocketed the money. So on the 10th day there was another man who arrived at Ellis Island, who knew them and bought them one more day. Then he ran out, got the father, and got them through. This near miss, where the guy almost didn't even end up here — he would have been sent back to Poland.

He was a feisty character. He was about 5'2", very short. He went to the Naval Academy, and he was so unlikable that in the Naval Academy yearbook they'd torn out his picture. In World War I he drove a coal ship. He was not some exceptionally high-performing senior officer commanding a destroyer or a carrier. But after the war he went to Oak Ridge, observed the vestiges of the Manhattan Project, and had this idea: we could build nuclear-powered submarines. He'd had experience on diesel-powered submarines, which are basically surface ships that happen to go underwater for about an hour. This became the thing he was obsessed with.

Talk about chutzpah — Oppenheimer himself thought this idea was going to fail. So you have Oppie telling you you're stupid and this isn't going to work, and you're just going to go forward. He built the first nuclear submarine in seven years, start to finish. The Navy hated him so much that his first office for the project was literally the women's restroom. They asked, what can we do to humiliate this guy into quitting? But he kept going. If you look at his personal memoirs, you can see the humiliation got to him, but in a way that he could channel to be motivating. He wasn't immune to the pain of it all, but he could push through it. Every major project you do is going to push you in these ways, and the question is how you dig deep and grind through the pain.

He created a really unique culture in naval reactors, which is alive today. Our nuclear submarine force is one of our last remaining asymmetric advantages against the Chinese, and we built it in the '50s. What a legacy. He was in this unique role of both being the human who invented the technology and therefore was qualified to think about how to regulate it. The Russian submariners would spend roughly six months at sea and six months in Sochi recovering so their white blood cells could regenerate, because the radiation shielding was so bad. We've had no deaths due to nuclear incidents in our submarine force. He designed it to a standard — he said, I'm designing this thing so my son could be in it — that is a hundred times safer than what we believed the minimum standard to be. That obsession is not a manager's view of the world. It is a founder's view of the world. He was an admiral for 30 years, and even as an admiral he was very difficult. Zumwalt, who was chief of naval operations for a while, said, "The Navy has three enemies: the Soviet Union, the Air Force, and Hyman Rickover."

Shyam Sankar

It surfaces in certain ways. As Karp would say, I'm not everyone's cup of tea. You could see it clearly in the early days of Palantir when you were dealing with the government. It's easy to think about it as a monolith, but you have operators who want to use your software but aren't in charge of buying it, and you have the IT people or the programme who are in charge of building what you'd be building. So you're a structural threat to them. The advice is to go along and get along — hey, those guys are the people paying you, why don't you just do what they want? The problem is, if I did what they want, it wouldn't work. That's where the disagreeableness comes in. It's being ornery about it: no, I'm going to deliver the thing that actually works for the operator. I'm going to piss off literally everyone along the way. They are knives-out going to try to kill you. We had to sue the Army to compete at one point.

In that vein of being truly committed to — is the product excellent? What else could it be doing? — you could think of the whole forward-deployed-engineering idea as continuously solving through back-propagation of what the product should be. That's the motor of innovation.

Shyam Sankar

There's no singular event. It's a combination of family history and the environment I grew up in. Starting with the family history: we fled violence in Nigeria, where my father almost died, to settle in the US, and that happened pretty young in his life — his early-to-mid 30s. It's a jarring experience that gives you a whole new lease on life, a deep amount of gratitude and perspective on the trials and tribulations that may lie ahead. But it also grounds you in a profound understanding of the counterfactual — what makes this country so special. Sometimes it can be unmooring not to have those experiences, because then it's easy to take it for granted, easy to become cynical.

My father didn't have the classic immigrant story where he came here and became successful — which is a great thing if it happens to you, but it's also a facile journey, easy to appreciate. Dad started businesses that went bankrupt. It was very hard. But having that rooting experience created a graciousness and perspective in him. He was always so thankful for the opportunity.

The second thing is growing up in Orlando in the '80s and '90s, a profoundly optimistic period. In the shadow of the Space Coast, you were bombarded with the idea that technology was going to make the world better, that there was a structurally positive-sum view of everything. It was so motivating. People had ambition around these things. That set a frame for how I thought about how I wanted to spend my life and energy. It's a commitment to building. Part of what's interesting about the heretics I'm drawn to is that they're all builders. The conversion from heresy to heroism is the result of an empirical thing they built. John Boyd was alive to see the success of his innovations in Gulf War I, where the US destroyed the fourth-largest army in the world. It's hard to think about now, because we treat it as a foregone conclusion, but leading up to it there was a lot of uncertainty.

So that worldview is centred on an understanding that the US is the greatest force for good that exists in the world, and that as a country we understand there's something special about founders. There's a reason we call them the founding fathers. Empirically, Europe has created zero companies worth more than a hundred billion euros from scratch in the last 50 years — an astonishingly bad track record. We've created all of our trillion-dollar companies from scratch in the last 50 years. The primacy of people, being in a culture and environment that allows that to happen, that doesn't subjugate the human or snuff out human flourishing — that's what the heretics are all a version of. And it's not Pollyannaish, as though there's no resistance to these creatives. It's that despite the resistance, you can succeed here, which really isn't true basically anywhere else.

Shyam Sankar

Every spare moment this man had, even though he was so busy and just grinding, he'd be on the phone trying to help someone. I have this memory: my dad had a sort of engineering autism that a lot of us have, but he met someone who had totally disfigured teeth, and he had the profound kindness to tell him, you should get cosmetic surgery. And that guy was so thankful, because no one else had the balls to say it. It actually changed his job prospects. It changed his life. Dad was one of these people in whom you could clearly see the positive-sumness — I'm just going to keep paying it forward.

Shyam Sankar

A belief in exceptionalism and greatness. No army that lost its morale ever won the war. Believing in yourself, believing greatness is possible, is a precondition to being able to express and realise it, and a lot of cultures don't have that. The other thing, which sounds simple, is a complete plasticity of thought. Culturally, America is a place where people do change their minds — hey, this isn't working, let's do something different; or, I observe that thing working, even though it goes against the orthodoxy of everything I've learned, so we're pivoting. In Europe, people are still arguing about things that happened 300 years ago. Some cultures are so rooted in the past that they can't update their priors. A culture capable of learning — you could call it plasticity, the first derivative of learning — how can you make progress without that?

Shyam Sankar

An aspiration substantially bigger than oneself. The focus of American greatness is on American prosperity and the American worker. How can we do things that enrich our civilisation and our people, raising their aspirations, inspiring them to go on and do great things? There's an unstable equilibrium in civilisation that pulls you towards nihilism, and greatness is the antidote — the countervailing force that says, no, it's actually worth investing in our institutions, worth going through this painful journey, because what we get on the other end is something better, something you'll be proud to pass down to your children.

Shyam Sankar

The obvious one to me is Alex Karp. His ability to manage and unlock talent is something I've never seen before, and those are some of the most profound lessons I've learned about Palantir. From the very early days, he modelled our entire company as an artist colony. We have more in common with a Hollywood talent agency than with a typical software company in how we look at humans. It's a first-principled approach to the individual — how do you grow? — and a deep resistance to a cargo-cult process in every manner and fashion that corrodes the things that actually create value. That matters, because if you drive process to the extreme you get something more like the Soviet Union. At the other end, how do you maximise the potential of the people in front of you? You advocate for their growth. That involves a structural personality where you're not threatened by other people's successes — you relish them. That's the parallel between Alex and my father.

Shyam Sankar

Start with how we think about talent to begin with. Really talented people are highly uneven. They're very good at some things, okay at another set of things, and really bad at some others. The trick is where they are in their own journey of understanding what those things are. Particularly for high-achieving people when they're young, they misattribute their superpower. They think it's a thing they do that requires a fair amount of effort, so they get a dopamine hit when they succeed — but that's probably something they're not even bad at, just okay at. Superpowers are effortless, so in a sense it's almost not even rewarding to exercise your superpower. My analogy: Superman could fly, could see through walls, but that wasn't arduous for him. It's just something he could do. You usually learn it comparatively — oh, this thing that's lizard-brain effortless, almost thoughtless for me, other really smart people can't seem to do, or I do way better. Embracing that is the first step. Then you help them understand that all of your contributions to the world are going to come from your superpower; everything else is basically a waste of time. So how do you get into that configuration? There's a releasing of ego around that.

The other part is kryptonite. There's some set of weaknesses that aren't just, hey, I'm average or a standard deviation below. It's that you're six deviations below average here. It's not something you can work on. The only strategy for Superman around kryptonite was to avoid it. So can you be okay with that? As the artist, what's your process of accepting it, and how do we help you accept it? At the point of acceptance, you become more valuable, and we can unleash you further.

Shyam Sankar

Exactly. You don't want a culture that says, you messed this up, I've got to fire you. It's, oh, I'm so glad you learned that this is not what you do for the company. I remember doing something I really screwed up once, that had consequences for the company. I sheepishly went to Alex and was completely honest. At the time there was no grand strategy behind it other than my commitment to always being honest. He was very appreciative. He was also in pain as he internalised what it was going to mean. But he valued that I wouldn't try to hide it. I took away from that how important that environment was going to be, and how to scale it through the organisation.

Shyam Sankar

There's a narrow sliver. In the Palantir context, the intersection of forward-deployed engineering and product. There's an interface there, and owning that almost as a dictator — being one-in on both sides of that interface — that's my superpower.

Shyam Sankar

Let me start with an analogy. How did Bruce Banner become the Incredible Hulk? It wasn't progressive overload; he didn't lift a little more weight every week. It was a near-fatal dose of gamma rays — 50% chance he dies, 50% chance he turns into a big green monster. There's absolutely an element where we take people and just irradiate them. Prospectively, you're not sure they'll come out the other end, but you have some belief they have the raw talent and potential. Then you create an environment with enough transparency and flow of information that, if you see the thing going off the rails, without them having to seek you out, you can push in and help them. A lot of environments don't have that by default — you'd have to manage them, do reports and check-ins. You've got to create an environment where it's okay to ask for help.

This model is valuable because the alternative doesn't actually work — you judge by the counterfactual. The structured career ladder with progressions is perfectly designed to make you feel comfortable that you're growing. It gives you the belief there's a linear path to follow. But it's all fake. On the other hand, if you're willing to suspend disbelief and throw yourself off the deep end — your maximal rate of learning will be coincident with your maximum ability to tolerate pain. Now you're in over your head, working on a problem you don't even understand. This is the environment where you'll learn at the fastest possible rate, limited only by your motivation and ability to endure the discomfort. Retrospectively, you look back and think, holy hell, I grew in all these ways, I learned all these things — none of which you could have predicted specifically in advance. If you're willing to do that serially, just keep jumping off the deep end, going after problems you're frankly not qualified for but have the raw potential to do, you turn into a superhero. That's why Palantir is the founders' factory.

Shyam Sankar

The deployment at the Co-op. That was a seminal one. I was leading the deployment; it was all on me, and it was never obvious until basically the last second whether it was going to work. It let me learn a lot of lessons along the way. That's where I internalised that success looks like a lot of pain, and that the instinct to reduce the pain — how do I do this next time with less pain? — is exactly the wrong instinct.

Shyam Sankar

It's the conspiracy of the people around you who give you the chance. I don't know why Kevin Hart hired me — some Stanford grad student coming to talk to him — but he took the bet and threw me out there. That was the first time I got to jump off the deep end. So you're assessing the people who are hiring you as much as they're assessing you. Is this an environment that's going to bet on me, or one that wants me to be a veal cow? Are they going to tolerate my eccentricities, or try to squash them?

Shyam Sankar

At some point it becomes lizard brain; you get enough reps. But early on, the indicators are the first derivative — proof of some sort of sui generis, did they create anything from zero to one — and then expression of agency and initiative. The simplest way I'd say it: if I give this person an inch, can they turn it into a mile? The worst bet is when you give someone a mile and they somehow turn it into an inch.

Shyam Sankar

Culturally, because all value creation is downstream of the culture. Continuing to be an environment where a new person can come in and have huge responsibility, be in charge of people who've been here 10 years or not. I call it a quantum org structure. I don't think a flat org structure is optimal — that doesn't actually work. What you want is an org structure that can crystallise into the right structure to solve the problem as it exists today, and reform as the problem changes tomorrow. You can see the pathology of big companies as they go through a big reorg every five years. Even if you steelman it and say this is the right structure for this company today, every day the entropy is against them, and every day it's slightly more wrong than before. If you're so ossified and committed to your structure that you have to wait until it's totally broken before reorganising, that's a fundamental weakness. So that plasticity, that commitment to working backwards from the problems — I hope that's how we keep generating value, and it has to be consistent. If we can protect that and promulgate it to future generations, we're going to keep winning.

Shyam Sankar

When you have a structure like this, you don't get to spend nearly as much time with people as you'd want, so the rate of irradiation is very heterogeneous. There's a bimodal distribution for how long people stay. The probability you'll stay a decade-plus if you've stayed three years is incredibly high. But around the three-year mark, everyone faces a crucible where they look around and say, this place is so messed up and crazy — and they have to understand that that's the feature of it. Then they stay. Or they say, this is the bug, I've got to get out of here. The critique I take most seriously is that with a little more help, a lot more people could make that hump. And that's the future of the business — it's to those people.

Shyam Sankar

Start with my critique of the conventional software-industrial complex. You're sitting in Palo Alto building software, and your essential feedback loop on whether it's valuable is, can I sell it? If someone's willing to pay, it must be valuable. That's an okay feedback loop. There's a stronger one: is this software valuable — did it deliver the outcome? The only way to assess that is in the field with the end user, on the factory floor, in the foxhole. In our case, given the weird division between the IT buyer and the operator, the validation was going to come from the operator. So the idea was to build software through back-propagation. You go to the person who has the problem. You show up with your best thesis of what you've built in the metaphorical Palo Alto, but you're not sending a sales engineer whose job is to ring up the sale — you're sending someone obsessed with whether it'll generate an impact in the world, willing to continuously solve backward, a tech lead who understands the full stack. How do we take the specific and learn to generalise it into the product over time?

There really are two types of engineers: people who know how to build the right thing, and people who know how to build it the right way. Their dopamine hits come from different things. The people who build the right thing are more like MacGyver — their high is, I solved the problem; in their minds, solving the right problem is 80% of the work. At the other end are engineers who are like artists in the traditional sense — don't you see how beautiful the thing I built is? The architectural elegance, the scalability, that's their high, never mind that it doesn't solve a problem anyone cares about. I think about it like vector maths. Forward-deployed engineers skew towards the hacker mentality, product engineers towards the artist mentality, and you have an unstable equilibrium where you're trying to align these vectors to go upwind. If either side dominates, you don't have a business. That's why that seam needs a dictator-like figure to manage and decide.

This model is super valuable when your customers exist in a power law of how useful the next marginal request is to the overall product. If it's a Gaussian, you should just survey all your customers, average it, and figure out what to build — the traditional product-development process. If instead some percentage of your customers are living in the future — only a handful have this problem today, but you can assess it's really valuable to solve, and no one else has — you should solve it now and effectively have a monopoly on that capability. You'll be five years ahead of everyone. You can only surface that in this forward-deployed way.

The other thing, in the Thielian sense, is the importance of secrets. If you're going to do something really impactful, there has to be something true about the world that you believe and others disagree with. Warp Speed was born out of one of those secrets. For the better part of 15 years, our software has been used to make things in the world. The one thing you could not tell anyone before COVID is that their manufacturing software doesn't work. People would point at their ERP system and say, look how great it is, I spent $5 billion, $10 billion on it. But as a forward-deployed engineer, when I'd go to the final assembly line on the factory floor, everyone's using Excel. There's a contradiction no one is forced to reconcile: what they're really saying is that the software is so ill-fit for what they're trying to do that they hit the eject button, pull the data out to Excel, and hand-roll their own solution. That should be an obvious sign something's not working.

Shyam Sankar

You have to believe you'll be able to capture meaningful value for the problems you're solving. The problems are really big, implying a lot of value behind them, and if you solve them deeply you'll capture it. To get at a lot of the problems we solve, you're not going to get paid a lot up front — you're actually going to invest a huge amount into the customer, because you believe that in the end you'll capture a portion of the value you're creating. Our economics will be downstream of our customers' economics.

The other part is that your ambition extends beyond a linear box. We always struggled coming into the commercial world, where the Gartners of the world would say, this is my enterprise architecture, I've broken it into all these boxes, which box do you fit in? The answer for us is: we don't fit in any box, we fit in all the boxes. You're solving for the problems in the seams. You're basically solving for the fact that the architecture diagram was made by some mid-level product-marketing manager, never used or validated, never actually solved a problem in the real world — but the cargo cult of the IT buyer keeps believing they should buy it. If you fit into that box, doing forward-deployed engineering would be crazy. But if you're like us and don't fit into it, you need some way of scraping and clawing to deliver enough value that the institution has to confront that, even though it's heterodox, even though it's heresy to their enterprise architecture, it works.

Shyam Sankar

This is actually the hardest question in the world for me. We've built an enterprise operating system, and our core thesis is that what makes this stuff really valuable is decisions, not data. It goes back to John Boyd's OODA loop. We provide an operating system that lets you bring all your enterprise data together. It solves for the impedance mismatch: all this data was collected and stored in a way that made sense to the underlying transactional system, or to some part of the institution, but doesn't represent how the humans of the institution think about the problems they face. We represent that in what we call the ontology, which is not just a model of the data but also a model of the actions — we call that the kinetics. If you're going to make a decision, you might be allocating inventory; how do you do that? This ontology layer becomes an API layer for the business. It makes the business programmable, and that changes the rate at which you can compound on top of it. Now you have an abstraction layer to build applications that manage your value chain from the hand of your supplier to the hand of your customer. You can call it a value chain, or a decision chain. Your whole business, if you squint at it, is just a series of decisions. For every one of them you want an OODA loop — I want to make the best possible decision today, but more importantly I want the first derivative to be good: I want to learn tomorrow how to make a better decision than the one I had today.

Shyam Sankar

Two things first, to make this make sense. One: we want to make software that makes you more different, not more similar. We think of it as alpha, not beta. Two: we are a wildly inductive company. We'd never show up to a customer with a preconceived notion of, this is the aerospace ontology. It's rather all the values and understanding of what makes your aerospace company different from all the others. That's the differentiation you're competing on.

When we first came to Airbus, they were ramping the A350. The A380 was a wonderful plane but a structurally unprofitable programme, so the A350 was a bet the whole company was riding on in 2015. The first time you make an airplane, you always have issues; there's a learning curve to production. One of the most important things is to figure out, as quickly as possible, what is just a non-conformity — hey, I'll give this guy some training and he'll figure it out, it's teething pain — versus a recurring issue that's actually a design defect you need to work upstream with your suppliers. How quickly you solve that becomes fundamental to the shape and cost of the ramp.

So the first thing we did was sit with users on the final assembly line in Toulouse. These are wrench-turners. What makes your life suck? What's hard about working these non-conformities? They were going through Excel, looking at reports in SAP, trying to bucket all the issues coming in. So we helped them automate all of that. The ontology at that point was related to parts, sequencing of work, defect rates — everything related to quality. Once you had that immense quality asset, you could pivot it: now I'm up the learning curve, and it naturally leads to the next problem, production planning. From quality to production planning to in-service. Now you have the greatest possible data asset for that tail — as the plane enters service, how do I maximise uptime for my customers and compete against other manufacturers? That's the natural evolution with most customers: you start with one problem, but of course it exists in a connected decision chain.

Shyam Sankar

There are a lot more non-market forces in the military. The military is the only institution in the world that profoundly divides supply and demand. The beating heart of most companies is that integration — the sales-and-operations planning process. In the military you have the combatant commands, which deal with real-world events; if you're fighting Russia, that's European Command, EUCOM. And you have the services — Army, Navy, Air Force, Space Force — whose job is to man, train, and equip. That's the supply side. So a group with no responsibility for demand is essentially deciding what to build and how much, then presenting those forces to the combatant commands and saying, this is what you have to solve the problem. If the demand side says, I really want this capability, they can't buy it. That adds complexity — all these non-market forces. But one thing that's really advantageous about being both a commercial and government company is that the intrinsic rewards on the government side are so high, so motivating, that it enables you to get through and solve really hard problems in a way that's a rising tide for the whole business.

Shyam Sankar

We are the best military in the world. Our uniformed service members are better than we deserve. Someone once asked me, when are you going to stop being shocked at how good this E-4 is at building this app? The amount of talent is eye-watering. I think we encumber them in a lot of process and deprive them of agency, and I have critiques of the bureaucracy. But the structural challenge we need to observe from a distance is this: the whole point of the military is to deter our adversaries from conflict. Over the last 10 years we've not been able to deter a lot of conflict. The annexation of Crimea in 2014; the militarisation of the Spratly Islands after they told President Obama they would not militarise them; a pogrom in Israel; the invasion of Ukraine; unprecedented grey-zone, phase-zero operations by the Chinese in the South China Sea.

Arguably, in the last year we've started to restore deterrence. Whatever you think about it, Midnight Hammer was a massively deterring operation. Maduro is massively deterring. With Midnight Hammer, it's the overwhelming strength, the precision, the incredible complexity — no other military in the world could have done that. But we need to view that as a single strike; the deterrence capability against a peer adversary is, can you generate that strike every single day until the enemy submits to your will? With Maduro, the raid happened the day the Chinese were in town. Maduro has all these Chinese and Russian weapons, and they don't seem to work at all. That sends a strong message to everyone who thinks their security comes from buying Chinese and Russian weapons.

Shyam Sankar

People want to talk about narrow things — acquisition reform, the mix of weapons. I think it comes back to people. It comes back to the heretics. At the dawn of World War II, we were the best at mass production; our country effectively invented it. To create an industrial base, we took the number two at GM, Bill Knudsen — a Danish émigré, previously number two at Ford, the perfecter of mass production — direct-commissioned him a three-star general, and put him in charge of war mobilisation. He was a world-class engineer, and we gave him the authority to go meet other world-class engineers all around America. One key advantage in timing was Lend-Lease: there was a period when we mobilised production while we were not fighting, building stuff we gave to our allies, the Brits and the Soviets. That let us build and retool factories and hit a production ramp, so that after Pearl Harbor we were basically at full-rate production. Put that in context of our adversary: the Germans were actually better engineers than we were, they just made a very small number of exquisite things. Now, in the present moment, we sell like the Germans — a very small number of eye-wateringly exquisite things — and our adversary is the world's best at mass production. That should give us all discomfort.

The second part is the nature of the industrial base. Today we call it the defence industrial base. It used to be the American industrial base until the fall of the Berlin Wall. Only 6% of spending on major weapon systems went to defence specialists — companies that only did business with the Defense Department. 94% went to what I call dual-purpose companies. A missile is single-use, there's no dual-use missile, but the idea that Chrysler would build missiles and minivans meant all their R&D on production, materials, machine tooling could be leveraged and was subsidising our national security. We recognised that freedom and prosperity were two sides of the same coin. General Mills, the cereal company, used to build torpedoes and inertial guidance systems. Kodak, the film company, was a key supplier to Corona, our first spy satellite. That's the industrial base that won World War II, won the early Cold War, and deterred our adversaries. It's only a consequence of becoming the sole superpower that everything changed. That 6% number is 86% today.

And the nature of it: we think Northrop Grumman, but more accurately it was Jack Northrop, Leroy Grumman, William Lear, Henry Ford, Henry Kaiser — founder figures, their own heretics, what we'd recognise in the Valley today. We were betting on those people, not just their institutions. In 1993 there was a famous dinner at the Pentagon called the Last Supper. We'd won the Cold War, and in a democracy it's reasonable for people to ask, what exactly is our threat, so we slashed the defence budget. At that dinner, the Secretary of Defense told a subset of the primes — we had 51 primes then, today we have five — you're not all going to survive, the budget's getting cut, we give you permission to consolidate, some of you should exit the defence business, and some of you may go out of business and we're not going to save you. That set off a merger frenzy, from 51 down to five.

Most people take away the wrong conclusion — that this is when we lost competition. The more profound consequence is that this was the moment of profound financialisation and conformity in the industrial base. You lost the crazy people. They went to tech and other parts of the economy where there was positive-sum energy and growth, where it wasn't about dividends and buyback ratios and financial engineering, but about real engineering. Why should we celebrate a Hyman Rickover or a Kelly Johnson? Kelly Johnson founded Skunk Works at Lockheed. In his lifetime he built 41 airplanes, many of which we still fly. He built the U-2 in 13 months. You can't look at that and think anything other than: this is an artist, this is divine. So much of what we've gotten wrong is that those things were messy, chaotic, hard, and full of friction. Over time we installed lots of process to make them less messy, and as a consequence they end up not working. We need a little more crazy back.

Shyam Sankar

It is daunting but critical, so we have no choice but to lean in and start. The best time to start would have been yesterday; the next best time is today. Let me give the case for why it's critical, because I think we're screwed if we don't do this. The biggest lie we bought from globalisation is the idea that we will do the innovation and they will do the production. What you have to recognise is that innovation is itself a consequence of productivity. In terms your audience would understand: what motivated Google to do the research behind the "Attention Is All You Need" paper in 2017? If you ask them, the team was working on an incremental 3% improvement to Google Translate. You can't think of anything more banal leading to something more revolutionary. If you don't have the input stimulus of improving production, you're not going to be able to capture the innovation. That's literally what we see playing out: they went from making a battery to making the whole car. WuXi went from being a cheap pair of hands for pipetting contract research in pharma to now, 50% of all clinical trials being drugs created in China. We seeded the innovation. It wasn't taken away from us; we seeded it because we had a completely incorrect precept.

We should re-attack that with some asymmetric advantage, and I think we have a couple. It's not true we're not great at building things — look at Elon and the progeny of Elon. And it doesn't all have to be new: Ford, General Motors, these are world-class manufacturers too. But we also have new ways of thinking — much deeper vertically integrated production, treating the production process itself as something iterative and software-like that you keep upgrading, co-locating R&D with production. And the last piece: AI is David's slingshot here. If we can give the American worker superpowers, make them 50 times more productive than a worker anywhere else, you change the efficient frontier of what can be made here. As a consequence, we should be re-industrialisation maximalists. The free traders say maybe we can friend-shore — I think that lets us off the hook too easily and fails to solve the problem.

From Patrick McGee's book, Apple in China: people love to say we don't have the machine-tool engineers in America, you could hardly get two people in a room to talk about it, whereas you could fill a stadium in China. But Apple has spent, on an inflation-adjusted basis over the last five years, the equivalent of two and a half Marshall Plans building talent and capacity in China. How about we try to spend one Marshall Plan here? The problem is we're not even trying, partly because we're daunted, partly because parts of our economy don't recognise this isn't going to keep working — you're going to lose everything, not just something. It's why I feel an urgent call to action: we are in an undeclared state of emergency. It's not just manufacturing physical things. 80% of our generic drugs come from China in one way or another, whether the APIs or the finished product. Today, your child's ear infection is trivial — some generic antibiotics and it goes away. But in a great-power competition with China, the will of the American people to fight has to contemplate that a five-year-old might die of an ear infection. I'm glad we're starting with rare earths, though there we had to get punched in the face to mobilise. One galling thing: we've been staring at the rare-earths problem for the better part of 15 years, and we more or less solved it for about $2 billion — a rounding error. So part of it is just starting. A marginal improvement of even 10% dramatically reduces the adversary's leverage.

Shyam Sankar

When we won World War II, we spent our own money to rebuild Germany and Japan. Microelectronics moved to Southeast Asia, as Apple in China and Chip War document, because we helped move it there. Yes, we got cheaper goods, but the economic development and flourishing of those countries, to create global stability, was part of the calculus — the same with admitting China to the World Trade Organization. The problem is they cheated the whole way through. If you go back to a pre-Deng Xiaoping era, they call us the big enemy. Part of it is our own naivety, our belief that capitalism and human flourishing are a natural consequence of prosperity — okay, they're saying these things, but as they get richer they'll change and we'll all find a positive-sum view. That's essentially what didn't happen.

The challenge with the CCP is that it's not enough for China to be prosperous; America must fail. You see it in simple acts. In agriculture, it's totally their prerogative — a business decision — whether to buy American soybeans or Brazilian, and I don't begrudge them that one bit. It's a totally different thing when you're trying to smuggle in agricultural funguses so we can't grow soybeans. Go down the laundry list, and that's the shape of what's happening. They're a formidable adversary. Our conception of war is kinetic — the lone heroic cowboy winning against the odds through heroism. Their conception is deception — the general who tricked the adversary into the valley and then flooded it. You want to win without firing a shot. A lot of what's happening is below what we'd consider the threshold of conflict, but it absolutely is conflict. In their system, it's system-destruction warfare, waging war through all means, where they have asymmetric strength. The American Calvinist sensibility is structurally positive-sum — we'll turn the other cheek time and again. But there will be some line, some point at which we react. And when we react, as an admiral I like says, it's no longer John Boyd's OODA loop — observe, orient, decide, act. It's the American OODA loop: observe, overreact, destroy, apologise. What we want is to prevent that by having enough deterrence that we're never provoked into a position where we even need to react.

Shyam Sankar

Their principal asymmetric advantage is long-term planning. They actually can't pivot on a dime. Go back to Gulf War I: they saw the devastating violence we did in basically four days to the fourth-largest army in the world, and they said, we have to systematically map all the critical dependencies of America's military technology and invest against it. They've been doing that since '91. The frog is kind of boiled. It's an easier game because they're playing a home game, but that's the map they've been on — a steady journey, and they broadcast their intent the whole time. No surprises.

Our asymmetric advantage is that we're crazy. The AI phenomenon wasn't in their long-range plan; it wasn't in ours either, but when it happened our whole economy pivoted on a dime. Even the Chinese models are a result of distillation. Our ability to be unpredictable — because we ourselves don't know what we're going to do — is our key strength. And that comes back to the heretics. Where does that unpredictability come from? Not our bureaucracy, which is predictable; not steady-Eddie managed decline. It's the crazies.

Shyam Sankar

Two incentives. One is just recognition that we need the heretics. Rickover enjoyed his celebrity for a big period; being known as the guy who did this is itself a reward. The second is financial. The cynical view of the defence-industrial complex is that these are greedy warmongers — well, they picked the wrong industry, because if you're greedy you should be in tech. These are crappy businesses: 9% operating margins, valued at less than two times revenue. We actually need to make our primes more valuable, and that would help attract the heretics.

How do you make them more valuable? The facile critique is that the primes are the problem. But the primes are downstream of the incentives the government gives, and the worst incentive is cost-plus contracting. Your upside is capped, you're not supposed to take any risk, the government pays for your R&D so you have no skin in the game. You can still make a small number of exquisite things, but the system is anti-heresy — you can only do the conventional thing the government agrees you should do. Think about the Higgins boat: Andrew Higgins built it on his own, there was no spec, which is why the Navy kept rejecting it — it wasn't the Navy's idea. But 92% of all boats in World War II were the Higgins boat. The heresy is the boat that won the war. How do you create an economic incentive where that's the case? Companies like Anduril are pioneering it on the hardware side: I'm building a product, these are my specs, these are my opinions, I'm absorbing the risk, buy it if you like it.

Part of my profound optimism now: I wouldn't have started this advocacy if I thought I was screaming into the wind. For the first solid 10 years we were alone in the darkness — it was SpaceX on the launch side and us. Today there are hundreds of founders building in the national interest with a US capital stack — another advantage — over a hundred billion dollars invested in projects that are changing how the department thinks about the future. Increasingly, in conversations at the Pentagon, some of these new entrants are talked about as plan A, not plan B — not, we'll see if this works in the future, but, this is the plan and we'll have backups if it doesn't. That's unrecognisable from the world I started in in 2006.

Shyam Sankar

I have narrow ideas. You want to bring pharmaceutical production back to the US? For generics the business case will never work, it's too much of a subsidy. But what you really want is to redevelop a skilled workforce that understands how to make drugs again, and production capacity that in a crisis could go from making branded drugs to generics. So why don't we manipulate patent length? Tell the drug companies: if you make your worldwide supply of branded drugs here in the US, I'll extend your patent length so you recover two extra capex investments to do it. That's a simple business case every pharma company would underwrite in a heartbeat, and it provides the catalyst to generate the capacity.

A lot of the American economy could look like that. Chrysler built minivans and missiles, Ford built satellites until 1990, General Mills, Kodak, down the list. It's a consequence of being the sole superpower — and the old Pentagon being the last bastion of communism except for Cuba. Every country gave up on communism, including Russia and China, except Cuba and the Department of Defense. That made it both a bad business and seemingly no threat, so you could just be a prosperity company, a single-purpose company that only cared about being a company. I'm not sure that's a stable or healthy civilisation. We have to recognise we have obligations across both freedom and prosperity.

Shyam Sankar

Some problems require sustained effort. If you can pivot really quickly, you can pivot towards solving the problem, but you can also pivot away from it, get distracted, and lose sight of what you're trying to accomplish. So it's a feature-bug, not an unequivocal feature. But in a competition you want to play to your strengths, which usually involves things the adversary can't do. It helps that the nature of populating the US was that all the heretics from the old world came here — an unfair advantage.

In maintaining a culture, the critical lesson I've learned over and over is that the entropy of the universe is against you. If you don't invest in your culture every single day, you'll lose it. Nothing about culture is self-sustaining — that's as true for America as for your company. The best cultures are almost cult-like. Your culture should be highly differentiated, hard for someone to inorganically insert into, otherwise it's not differentiated at all. That requires flag-bearers, true believers, so it starts with recruiting. Not every human has the disposition to want to be part of a cult. Some are structurally mercenaries, evaluating everything through the lens of, what's my next job? You want to build your company with people who think of the present job as the calling. That engenders the primacy of winning — they're always thinking about how to win today, because that's the whole game. If you win today, you have the right to win tomorrow. The other side is getting lost in what I wish the world could be — an aesthetic of how you wish the world worked, rather than how it does work and what's required to win today.

Shyam Sankar

The most important decisions from my seat are which projects are worth outsized investment because we think they'll define the future of what our customers need, and therefore the future of the business over the next five years. Then, how do you get the right talent on those projects, and how do you discover what those projects are? My backlog is something I re-prioritise dynamically relative to reality. Here's my roadmap, all the things I want — it's an approximation, at a point in time, of what I think are the most valuable things to build. The second some new data comes in, I'm going to rip it up and re-prioritise. Having a culture that lets you do that, because it's really painful — people don't like it, their world is shaken up, it generates chaos — but it's the right thing to do. Managing that, so your culture enables it, and holding the conviction that the cost is worth it because the return is so outsized: that's the whole business.

Shyam Sankar

There are always degrees. But to the first-order issue: is it hard because you want to exist in a business that isn't painful? There's a great Greg LeMond line — a world-champion cyclist. Someone asked him about a championship performance, and he looked almost incredulous and said, you know it doesn't get easier, right? You just go faster. What do you think a championship performance feels like? It feels shitty. It's really painful. Your ability to win is basically, can you survive more pain than your competitors? There's a weird phenomenon: if you're doing something world-class, it's really painful, and if you're totally failing, it's also really painful. So you need discrimination to know which of the two worlds you're in. The worst world is where you have no signal — you're plodding along, but it's managed decline. Is the pain we're experiencing generating results or not? Orient people to that.

The other part: you can't move the whole company at once. Even a company as small as ours, 4,500 people, every change is a mini-insurgency inside the company. Think about it like campaigning. Who already believes what you're saying? How do you add incremental weight behind them, help them succeed more, generate momentum? My experience is that momentum is the thing that aligns the whole company. So you should wake up every day, in the midst of these pivots, working on momentum. People will catch up.

Shyam Sankar

Celebrating the heretics, the successes, and creating venues where everyone has some ability to revolt. Every employee's first AMA with me, I end it by telling them to tell me to f*** off to my face. It's very uncomfortable for them — they're almost not sure if I'm entrapping them into something. But how can you say you have a flat culture if you can't tell the boss to f*** off? It's a small gesture, but it sets the condition: we are a company committed to taking the very best idea, regardless of where it comes from, and then putting all our wood behind those arrows. Most of those ideas are some variant of what we're currently doing — otherwise it'd already be the current best idea.

You have to be careful how much you institutionalise, but at minimum, twice a year we have these weeks of revolt where people can go build whatever they want. I want people to have a chip on their shoulder — the whole point is to prove that we've been doing it wrong, and to inspire us with the thing you've built. If you have a culture that is profoundly truth-seeking, the truth is so charismatic that now that I've seen it, I can't unsee it — this is obviously the right answer, let's go. It's not as hard as it sounds to move and pivot on these things.

Shyam Sankar

The only way you can do it is by disrupting yourself. To the point of institutionalising rebellion and holding that up, it's the perpetual motor of self-disruption. The most valuable thing is not to defend a moat, but to charge after the new opportunity. One thing unique about AI is how much it plays to the beginner's mind. The first adopters of AI for us were the youngest hires. To generate momentum, I didn't bother with the people who were cynical or dismissive of it. I let the young people generate overwhelming productivity that in some sense humiliated or inspired others to get on board. There was an almost out-of-body experience for someone who'd been there 20 years, or even three years, to say, you know what, the best people at this are really the young people — by which they meant people who'd been there three months. That recognition — here's someone who's not a gatekeeper, but is trying to find the insurgents, empower them, and place them where they can do better work so we can all learn together. It's been really fun, the rate of transformation and what AI has enabled.

Shyam Sankar

We think the value accrues at the chips layer and at the ontology layer. If you squint at the stack — chips, model providers, AI infrastructure (which is where we think the ontology sits), and AI applications — you observe that the model companies have to run up the stack, away from providing a pure model, because the model is commoditised; it's everything they build around it that creates value. The AI pure-play applications are running down: as they scale, deal with more customers and more of their customers' data, they're reinventing the AI infrastructure — essentially AIP — from first principles, recognising why they need it. Through some divine intervention, we built this platform over the last 20 years, the meat of it since 2015 with Foundry leading to AIP, which gives us a huge head start. Even if these models help you code faster, there's the code, and there are the secrets behind why we built the code. We feel we're in a very good position, and it's expressed by how quickly we can deliver value: two years ago we could fundamentally change a customer's business inside eight weeks; now it feels like a week. That rate is compounding dramatically.

Shyam Sankar

I'm a bit reluctant about the shift, but I can't look my kids in the eye and not ring the bell that we're in an undeclared state of emergency. The first major public thing I put out was the 18 theses: we're in emergency, we've lost deterrence. I'd been doing this for roughly 19 years at that point. Here's my diagnostic of what's wrong; the shot clock is ticking; we've got to coalesce around this problem. On a note of profound optimism, I've seen more change in the Department of War in the last year than in the prior 19. To the point that Americans do the right thing having exhausted all the other options, we are definitely starting to do the right things, and the first and even the second derivative are positive.

Shyam Sankar

To synthesise: we've lost deterrence. The reason this won't go the way World War II did is that we don't have the industrial base we had then. We have to recognise our unique American strengths. The person is the programme. Edward Hall built the Minuteman; Schriever was the Air Force general who oversaw our ICBMs; we call it the Apollo programme, but it was actually Gene Kranz's programme. The primacy of those people, and getting the right people into those roles, is determinative. It would be like thinking you could found a company by recipe — the company's future is determined by the founders. That's almost the congenital limiter on what's possible, and of all cultures we should understand that.

Downstream of that, we should recognise there is no process out of this. The reason we're screwed is that we prayed at the altar of process — we thought if we follow this requirements process, or just let the bureaucracy work, it would lead to a good result. The only thing that leads to is decline. Look at the great military innovations: Churchill built the tank as head of the Royal Navy, because the British Army wasn't smart enough to realise horses wouldn't win the next war — he called it a land ship, because he could only build ships. That's not a one-off. Every major innovation, including the Sidewinder missile — a rogue project some guy built with cigarette lighters and a new seeker, that humiliated the Air Force's missile in Vietnam and became the standard. When Bill Perry became Assistant Secretary of Defense, he was staring at this and said, I can either try to fix the system and lose, or take the two things I really care about and make sure they work: GPS and stealth. At every crossroads, the person who just wants to engage with the system, we lose; the person willing to put it all on the line to make it work, we win. It's a call to the latent heretics in the department and in defence tech: this is the moment to express your heresy.

Shyam Sankar

First of all, we don't really think of ourselves as a defence company. We're a software company. The fact that half our business is in government, and two-thirds of that half is in defence, is actually more normal to the historical pattern we've discussed than the present is. But the part outsiders misunderstand is that they think we're either in the business of collecting data or doing surveillance. That's the least well-understood part of what we do. The core problem we're set up to solve is the legitimacy of our institutions. Why do doors fall off planes? Why does it seem like our basic government services don't work? If you don't address those problems, if you don't build state capacity — private and public — it breeds nihilism, the sense that we should just burn it all to the ground, that there's nothing worth investing in. That's a strictly worse world.

Our diagnosis: give the C-suite, government or private, the benefit of the doubt. They have the steering wheel, and they're diligently trying to steer. What they don't understand is that it's a Jungle Cruise prop from Disneyland — it's not connected to anything. Then the people down below look up and say, how could my leaders be so dumb? That's where that neologism starts. The point of this operating system, the reason we're trying to make decisions better, is to create something more responsive for both the people leading these institutions and the people who do the work. So much of the 20th century was a wild managerial revolution that concentrated power away from both ends into an amorphous middle blob — that's where the disconnection in the steering wheel happens. Reasserting that connection, so our institutions work for the people they're designed to serve and our companies are the best in the world — that's how you underwrite freedom and prosperity.

Shyam Sankar

It used to weigh on me. After you've done it for 20 years — how long can this keep weighing on you? But it's fair to say it weighs on every new employee when they come in; there's a journey of dealing with it. That's the lot of the man in the arena. To end with John Boyd: to do, or to be. Are you trying to be someone, or to do something? We're trying to do something.

Shyam Sankar

John Boyd's advice to other officers, other uniformed service members, was always: you can either be somebody or do something, but you can't really have both. If you're going to be somebody, you have to play the game — it's all theatre, but you'll get promoted, get the accolades and rewards, and feel good about yourself. Or you can actually do something, and it'll feel really shitty and no one will appreciate it, but you'll have the intrinsic reward of knowing that you did something.

Shyam Sankar

It would be hard to overstate the impact Dad and his example had on me. Such a selfless man, who gave everything to his family, who could weather any humiliation in service of his children and his wife. As a kid, you learn this stuff subconsciously, and like any normal kid you think the thing Dad did is annoying, you don't fully appreciate it — but the culmination of all these events over time profoundly shapes who you are. I hope I can set an example that inspires my kids to do the same. A lot of fatherhood is being a man who inspires his family to be better.

Shyam Sankar

Miss Ethel Danhoff was the head of admissions for a private school in Orlando. My dad took me for a tour, having learned it was the best school — that was his other fixation, that education was how you achieve prosperity: I'm going to give my son every opportunity I possibly can. It turned out we'd missed the admission window by a mile and everything was done. I was actually reluctant to go on the trip; I was happy at the school I was at. But when I saw the science facilities, I thought, I need to go here, I need to learn this stuff, this is inspiring. So now I was spun up and excited, and then I learned, hey, you totally screwed this up, you'll have to wait a year. And Ms. Danhoff — she had no data on me, no standardised test scores — said, give this kid the admission test, just do it. Maybe she had a modicum of data after the test, but we were a family that needed financial aid. So here it is: let's admit a kid we're going to have to subsidise to come here. And that opened the entire future. Later in life, maybe people have cynical reasons to help you — you're further along, they think there might be something in it. There was nothing in it for her. It was just a profound act of kindness that fundamentally changed my life.

Shyam Sankar

Thank you.