Summary

Sankar’s organizing thesis is that military innovation only ever comes from heretics — “they’re really founders” — with a categorical claim attached: “the only things that ever worked, the things that helped us win all the wars, were the things that the heretics actually did. Nothing that went through the machine delivered anything.” Hyman Rickover built the first nuclear submarine in 7 years from an office the Navy assigned him in the women’s restroom, and that ’50s-era sub force remains “one of our last remaining asymmetric advantages against the Chinese.”
America has lost deterrence and, arguably, only started restoring it in the last year — Crimea 2014, the militarized Spratly Islands, Ukraine, gray-zone operations in the South China Sea — but Midnight Hammer and the Maduro operation were “massively deterring,” and Sankar has seen “more change in the Department of War in the last year than the prior 19 years.” Some new defense entrants are now discussed inside the Pentagon as “plan A” — “unrecognizable from the world I started in 2006,” with hundreds of founders building in the national interest with a US capital stack and over $100B invested in their projects.
The industrial-base decay is quantified: in the Cold War era only 6% of major weapon-systems spending went to defense specialists; today it’s 86%. The 1993 “Last Supper” collapsed 51 primes to 5, and the real damage wasn’t lost competition but “financialization and conformity… you lost the crazy people.” Primes run ~9% operating margins at under 2x revenue — his fix is to make them more valuable and address cost-plus contracting, which he calls structurally “anti-heresy.”
“The biggest lie that we bought from globalization is… we will do the innovation and they will do the production.” Innovation is downstream of production: the “Attention Is All You Need” paper came from a 3% improvement to Google Translate, WuXi went from pipetting-for-hire to 50% of all clinical trials being drugs created in China, and 80% of US generic drugs come from China in one way or another. Apple has spent “the equivalent on an inflation-adjusted basis in the last 5 years of two and a half Marshall Plans” building talent and capacity in China; “how about we try to spend one Marshall Plan here?” AI is “David’s slingshot” — a 50x-more-productive American worker changes the efficient frontier of what can be made here. He is a “re-industrialization maximalist”; friend-shoring “can let us off the hook too easily.”
On AI value capture: “the value is going to accrue at the chips layer and at the ontology layer.” Models are commoditized — model companies are running up the stack while AI application pure-plays run down and reinvent AIP from first principles. Palantir’s delivery metric compounded from “inside of 8 weeks” two years ago to “it feels like a week.”
The forward deployed engineering model he pioneered is “build software through back propagation” — validating outcomes “on the factory floor, in the foxhole,” not at the point of sale. It only makes sense when customers sit on a power law (solving the future-dweller’s problem puts you “5 years ahead of everyone else”) and when you can capture value downstream of customer economics; if you fit a Gartner box, “doing forward deployed engineering would be crazy.”
Palantir’s talent machine: hire people who can turn “an inch into a mile,” irradiate them Hulk-style with problems they’re unqualified for (“your maximal rate of learning will be coincident with your maximum ability to tolerate pain”), and enable superpower/kryptonite discovery — because career ladders are “perfectly designed to make you feel comfortable that you are growing… but actually it’s all fake.” Sankar calls the company the Founder’s Factory.

Deep dive

1. Heretics are founders — the machine delivered nothing

Sankar’s definition: heretics are “really founders… there’s almost like a pathological obsession with winning,” fighting the bureaucracy at career-ending personal cost. The historical claim is stated without hedge: “the only things that ever worked, the things that helped us win all the wars, were the things that the heretics actually did. Nothing that went through the machine delivered anything.”
Heretics “only later become heroes,” and the conversion always runs through something empirically built. Billy Mitchell “was court-martialed and died penniless and depressed, and only after his death did we invent the Air Force based on his original heresy.” John Boyd was alive to see the success of his innovations in Gulf War I — the US destroying the fourth-largest army in the world, which looks foregone now but wasn’t.
His 18 Theses distill to: we’ve lost deterrence, we lack the WWII industrial base, and “the person is the program” — Edward Hall built the Minuteman, Apollo “was actually Gene Kranz’s program.” “There is no process out of this… we prayed at the altar of process,” and that only leads to decline. Churchill built the tank as head of the Royal Navy — calling it a “land ship” because he could only build ships — and the Sidewinder was a rogue project that “totally humiliated the Air Force’s missile in Vietnam.” Bill Perry framed the choice as fixing the system or protecting GPS and stealth.
The closing Boyd-ism that frames Sankar’s own posture, including on being cast as the surveillance villain: “You can either be somebody or you can do something” — doing something “is going to feel really shitty and no one’s going to appreciate it, but you’re going to have the intrinsic reward of knowing that you did something.” The panopticon caricature “used to weigh on me… it’s the lot of the man in the arena.”

2. Rickover: a nuclear submarine in 7 years, with an office in the women’s restroom

The heretic he’d have everyone study. Rickover nearly never made it to America — held at Ellis Island ten days after a courier pocketed the telegram money — was so disliked at the Naval Academy that his yearbook photo was torn out, and drove a coal ship during the war. At Oak Ridge he saw the Manhattan Project’s vestiges and fixated on nuclear submarines — with Oppenheimer himself thinking the idea would fail.
He built the first nuclear submarine in 7 years, start to finish, while the Navy assigned his project office to the women’s restroom to humiliate him into quitting. His memoirs show the humiliation “got to him, but in a way that he could channel it to be like motivating” — “it wasn’t like he was somehow immune to the pain.”
The engineering standard was a founder’s, not a manager’s: “I’m designing this thing so my son could be in it” — a hundred times safer than what they believed the minimum standard to be. Russian submariners spent 6 months at sea, then 6 months in Sochi “so their white blood cells could regenerate”; the US force has had zero nuclear-incident deaths. Zumwalt’s line: “The Navy has three enemies: the Soviet Union, the Air Force, and Hyman Rickover.”

3. A worldview forged by Nigeria and the Space Coast

The family fled violence in Nigeria — “we frankly almost died” — and his father’s story was not the facile immigrant arc: “Dad started businesses that went bankrupt.” What it produced was gratitude and “a profound understanding of the counterfactual, like what makes this country so special” — the absence of which, he notes, “can be unmooring… it’s easy to become cynical.”
Growing up in Orlando in the ’80s–’90s, in the shadow of the Space Coast, he was “bombarded with this idea that technology was going to make the world better” — a structurally positive-sum frame. His father’s signature trait: every spare moment spent helping someone, including the “profound kindness” of telling a stranger with disfigured teeth to get surgery — advice that changed the man’s job prospects.
America’s two load-bearing components: belief in exceptionalism (“no army that lost its morale ever won the war”) and complete plasticity of thought — “in Europe, people are still arguing about things that happened 300 years ago.” The empirics: Europe has created zero companies worth more than €100B from scratch in 50 years; the US created all its trillion-dollar companies from scratch in the same window. Greatness is “an aspiration that’s substantially bigger than oneself” — the antidote to civilization’s “unstable equilibrium” pull toward nihilism.
The kindest thing ever done for him fits the theme: Miss Ethel Danhoff, an Orlando private-school admissions head, admitted him past a blown deadline — “give this kid the admission test, just do it” — for a family needing financial aid. “There’s nothing in it for her… it fundamentally changed my life.”

4. Talent: superpowers are effortless, kryptonite is unfixable

Karp modeled Palantir from the earliest days as “an artist colony” — “we have more in common with a Hollywood talent agency than a typical software company.” The precept: really talented people are highly uneven, and young high-achievers misattribute their superpower to the effortful thing that delivers dopamine hits. Real “superpowers are effortless” — Superman didn’t strain to fly — and “all of your contributions to the world are going to come from your superpower. Everything else is basically a waste of time.”
The mirror image is kryptonite: not “a standard deviation below average” but “six standard deviations below average… The only strategy for Superman around kryptonite was to avoid it.” At the point of acceptance, “you become more valuable.”
Discovery requires exposure — a real mistake with real consequences — so the culture can’t be “you screwed this up, I gotta fire you.” Sankar once confessed a costly screwup to Karp with no grand strategy beyond honesty; Karp “was also in pain as he internalized what this was going to mean,” but valued that he didn’t hide it. Sankar names his own superpower narrowly: owning the seam between forward deployed engineering and product, “almost as a dictator.”

5. Gamma rays, not progressive overload

The development model: “How did Bruce Banner become the Incredible Hulk? It wasn’t like progressive overload… It was a near-fatal dose of gamma rays.” Palantir is “taking people and just irradiating them” — with enough transparency and information exhaust that leadership can push in if things go off the rails without being sought out.
Judged against the counterfactual, ladders lose: the structured career progression “is perfectly designed to make you feel comfortable that you are growing… but actually it’s all fake.” Instead: “your maximal rate of learning will be coincident with your maximum ability to tolerate pain” — jump off the deep end serially at problems “you’re frankly not qualified to go after” and “you turn into a superhero. That’s actually why Palantir is the Founder’s Factory.”
His own gamma-ray moment was the Co-Op deployment — all on him, unclear until “basically the last second” whether it would work. The lesson: “success looks like a lot of pain,” and the instinct to make the next one less painful “is exactly the wrong instinct.” The bet-on-talent heuristic: “If I give this person an inch, can they turn it into a mile? The worst sort of bet: I gave this person a mile and somehow they turned it into an inch.”
The structural hope for the next decade is a “quantum org structure” — not flat, but one that “crystallizes into the right structure to solve the problem as it exists today and is able to reform as the problem manipulates tomorrow.” Big-company reorgs are “probably wrong even at inception, and every day the entropy is against them.”

6. Forward deployed engineering: build software through back propagation

His critique of the “conventional software industrial complex”: sitting in Palo Alto, your feedback loop is “can I sell it?” The stronger loop is “did it deliver the outcome?” — assessable only “in the field with the end user, on the factory floor, in the foxhole.” This is where his disagreeableness lives: the IT buyer pays you, the operator validates you, and “if I did what they want, it wouldn’t work” — Palantir sued the Army to compete.
Two types of engineers: MacGyver hackers, for whom “solving the right problem is 80% of the work,” and artists — “don’t you see how beautiful the thing I built is?… never mind it doesn’t solve a problem anyone cares about.” FDEs skew hacker, product engineers skew artist; it’s vector math in unstable equilibrium, and “if either side dominates, you don’t really have a business” — hence the dictator at the seam.
The model pays when customers sit on a power law: if the value of the next marginal request is Gaussian, survey everyone and average — traditional product management. If some customers “are living in the future,” solving their problem now means “effectively a monopoly… we’ll be 5 years ahead of everyone else.” And it surfaces Thielian secrets: pre-COVID, “the one thing you could not tell anyone is that their manufacturing software doesn’t work” — $5–10B spent on ERP while everyone on the final assembly line uses Excel. That eject button is the tell; Warp Speed was born from that secret.
The advice to entrepreneurs: use FDE only if the problems are big enough that “our economics will be downstream of our customers’ economics,” and if your ambition “extends beyond a linear box.” If you fit neatly into a Gartner enterprise-architecture box, “doing forward deployed engineering would be crazy.” Palantir’s answer to “what box do you fit in?” — “we don’t fit in any box. We fit in all the boxes.”

7. What Palantir is: decisions, not data

“Actually the hardest question in the world for me.” The answer: an enterprise operating system whose thesis is “what makes this stuff really valuable is decisions, not data.” The ontology models both the data and the actions — “the kinetics” — and “almost becomes like an API layer for the business. It makes the business programmable,” changing the rate at which you compound. Every decision in the value chain should have its own OODA loop, with the first derivative — learning to decide better tomorrow — mattering most.
The Airbus example: in 2015 the whole company was “effectively riding on” the A350 ramp (the A380 was “a wonderful plane, structurally unprofitable program”). Palantir started by sitting with “wrench turners” in Toulouse, triaging non-conformities — training-fixable teething pain versus design defects needing suppliers — out of SAP reports hand-worked in Excel. Quality led to production planning led to in-service uptime. The philosophy: “alpha not beta” — “software that makes you more different, not more similar,” from a “wildly inductive company” that would never arrive with a preconceived aerospace ontology.
The deeper problem he says outsiders miss (Palantir is neither collecting data nor doing surveillance): institutional legitimacy. “Why do doors fall off planes?” The C-suite diligently turns a steering wheel that is “a Jungle Cruise prop from Disneyland — it’s not connected to anything,” while the people below ask how their leaders can be so dumb. The 20th-century managerial revolution concentrated power in “an amorphous middle blob”; unfixed, it “breeds nihilism… we should just burn it all to the ground.”

8. Culture must be maintained; AI is the self-disruption test

“The entropy of the universe is against you… if you don’t invest in your culture every single day, you will lose it. There’s nothing about culture that is self-sustaining — that’s as true for America as it is for your company.” The best cultures are “almost cult-like”; recruit believers, not “structural mercenaries.” Every new employee’s first AMA ends with Sankar telling them to tell him to fuck off to his face — “how can you say you have a flat culture if you can’t tell the boss to fuck off?” — plus twice-yearly “weeks of revolt” to prove “we’ve been doing it wrong.” The critique he takes most seriously: the ~3-year crucible where employees decide the craziness is feature or bug — “with a little bit of help, a lot more people can make that hump.”
On the whiplash of constant re-prioritization, the Greg LeMond quote: “You know it doesn’t get easier, right? You just go faster.” Championship performance feels shitty — the discrimination that matters is whether the pain is generating results or is just managed decline (“the worst world is where you have no signal”). Every change is “a mini insurgency inside the company”; campaign for it, add weight behind existing believers, and “momentum is the thing that aligns the whole company. People will catch up.”
AI adoption ran on beginner’s mind: he didn’t bother converting cynics, but let the youngest hires — “the best people doing this” had been there three months — generate “overwhelming productivity that almost in some sense humiliated or inspired other people to get on board.” Against would-be Palantir competitors enabled by the same technology: “the only way you can do that is by disrupting yourself… the most valuable thing is not to defend a moat but to go charge after the new opportunity.”
The value-capture call: “at the chips layer and at the ontology layer.” The model is commoditized, so model companies run up the stack; AI pure-play applications run down the stack as they scale, “reinventing the AI infrastructure, essentially AIP, from first principles.” Palantir’s 20-year head start (Foundry-to-AIP since 2015) shows up in the metric he watches: delivering enterprise autonomy went from “inside of 8 weeks” two years ago to “it feels like a week.”

9. Deterrence was lost — and the last year started restoring it

“We are the best military in the world. Our uniform service members are better than we deserve” — the eye-watering talent of an E-4 building apps — but encumbered by process and deprived of agency. The scoreboard is failed deterrence: Crimea 2014, the Spratly Islands militarized after China told President Obama it wouldn’t, a pogrom in Israel, the invasion of Ukraine, “unprecedented gray zone phase zero operations” in the South China Sea.
“Arguably in the last year we’ve started to restore deterrence.” Midnight Hammer was “massively deterring” — but a single strike isn’t peer deterrence: “can you generate that strike every single day until the enemy submits to your will.” The Maduro raid “happened the day the Chinese were in town,” and Venezuela’s Chinese and Russian weapons “don’t seem to work at all — and that sends a very strong message to everyone who thinks their security is coming from buying Chinese and Russian weapons.”
On the CCP: “it’s not enough for China to be prosperous, America must fail.” Buying Brazilian over American soybeans is fair business; “smuggling in agricultural funguses so we can’t grow soybeans” is not. Their conception of war is deception and “system destruction warfare,” waged below the threshold of conflict — while the American Calvinist sensibility turns the other cheek until it snaps into what his favorite admiral calls the American OODA loop: “observe, overreact, destroy, apologize.”
The asymmetry map: China’s advantage is long-term planning — after watching Gulf War I, “they said, okay, we have to systematically map all of the critical dependencies of America’s military technology and invest against it. And they have been doing that since ‘91,” broadcasting intent the whole way. America’s advantage: “we’re crazy.” AI was in nobody’s long-range plan, “but when it happened, our whole economy pivoted on a dime” — “even the Chinese models are a result of distillation.” The unpredictability comes from the heretics, not the bureaucracy.

10. Reindustrialize or lose everything

“The biggest lie that we bought from globalization is… we will do the innovation and they will do the production.” Innovation is itself a consequence of productivity: the attention-is-all-you-need paper came from “an incremental 3% improvement to Google Translate” — “you cannot think of something more banal leading to something more revolutionary.” WuXi went from “a cheap pair of hands for pipetting” to 50% of all clinical trials being drugs that are created in China; they went from making a battery to making the whole car. “It’s not that it was taken away from us. We just seeded it.”
The history he wants remembered: at WWII’s dawn America direct-commissioned GM’s Bill Knudsen a three-star general, and Lend-Lease meant we could mobilize production before Pearl Harbor, so that after Pearl Harbor we were basically at full-rate production. Until the fall of the Berlin Wall, only 6% of major weapon-systems spending went to defense specialists — 94% to dual-purpose companies (Chrysler built missiles and minivans; General Mills built torpedoes; Kodak supplied the Corona spy satellite). Today that 6% is 86%, and after the 1993 “Last Supper” collapsed 51 primes to 5, the real consequence was “financialization and conformity… you lost the crazy people.” Kelly Johnson built 41 airplanes in a lifetime, the U-2 in 13 months — “you can’t look at that and think anything other than this is an artist.” Now “we sell like the Germans” — few, exquisite — while the adversary is the world’s best at mass production.
The incentive fixes: primes run ~9% operating margins at under 2x revenue — “we actually need to make our primes more valuable” — and cost-plus contracting is “anti-heresy”: capped upside, no risk, no skin in the game. The Higgins boat had no Navy spec (that’s why the Navy rejected it) and became 92% of all WWII boats; Anduril is pioneering the modern version — “these are my specs, I’m absorbing the risk, buy it if you like it.” His president-for-a-day order: extend patent length for pharma companies that make their worldwide branded supply in the US — “a simple business case every pharma company would underwrite in a heartbeat” — because 80% of our generic drugs come from China in one way or another and in a great-power fight “your 5-year-old might die of an ear infection.”
Why he went loud after 19 years behind the scenes: “I can’t look my kids in the eye and not say that I’m ringing the bell that we’re in an undeclared state of emergency.” But the note is profound optimism: for the first solid 10 years “we were alone in the darkness — it was SpaceX on the launch side and us”; now hundreds of founders with a US capital stack and over $100B invested in their projects are treated in the Pentagon as “plan A,” and he’s seen “more change in the Department of War in the last year than the prior 19.” AI is “David’s slingshot” — a 50x-more-productive American worker changes the efficient frontier of what can be made here. Be a “re-industrialization maximalist” (friend-shoring “can let us off the hook too easily”); even a 10% improvement “dramatically reduces the adversary’s leverage” — rare earths, stared at for 15 years, are more or less solved for roughly $2B, “a rounding error.” Apple spent “the equivalent on an inflation-adjusted basis in the last 5 years of two and a half Marshall Plans” building talent and capacity in China: “how about we try to spend one Marshall Plan here?”

A tour through the media landscape with TSOH's Alex Morris Agent Inference at the "Speed of Light" — How NVIDIA moves like a $4.3 Trillion Startup