Shyam Sankar - Celebrating Heretics - [Invest Like the Best, EP.462]
At a Glance
Shyam Sankar, CTO of Palantir Technologies and the company's 13th employee. The main thread of this episode: America must reindustrialize to reverse decades of shifting production overseas, and the key to achieving this lies in reactivating the historical "heretics"—builders who challenge bureaucratic systems. Sankar's core judgment is that the U.S. has lost deterrence and is in an "undeclared state of emergency," with AI serving as "David's sling"—if it can make American workers 50 times more productive than workers anywhere else, it will fundamentally shift the efficiency frontier of "what can be made in America."
Theme 1: Heretics — The True Engine of American Military and Industrial Success
Sankar argues that every military innovation in American history that truly worked came from "heretics" who challenged the bureaucracy, not from formal processes.
Historical Context and Mechanism Breakdown:
- Sankar defines "heretics" as "founder figures" in the military domain — individuals driven by an almost pathological obsession with winning, willing to risk their careers to defy the bureaucracy.
- Key case: Hyman Rickover, who stood only 5 feet 2 inches tall and had his photo torn out of the Naval Academy yearbook. During World War II, he commanded only a coal ship. After the war, while observing the remnants of the Manhattan Project at Oak Ridge, he proposed building a nuclear-powered submarine — a concept even Oppenheimer believed would fail. The Navy gave him an office in a women's restroom, hoping to shame him into quitting. Yet within seven years, he built the first nuclear submarine from scratch.
- Rickover designed nuclear reactor safety standards 100 times higher than the minimum requirements — "I designed this thing so that my own son could serve inside it." The result: to this day, no one in the U.S. nuclear submarine force has died from a nuclear accident, whereas Soviet submariners had to recuperate in Sochi for six months after every six-month deployment to restore their white blood cell counts.
- Other heretics: Billy Mitchell, the inventor of the air force concept, court-martialed, died in poverty and depression; John Boyd, the creator of the OODA loop, three-time top fighter pilot, a "jerk" by nature and highly unlikable, but instrumental in developing the F-16 and the military strategy for the Gulf War.
- Sankar's reasoning: "All change comes from these heretics, who only later become heroes. If you look back at history, the only things that really worked — the things that helped us win all our wars — were actually built by heretics. The things produced through the machine delivered nothing."
Sankar's self-acknowledged heresy: In Palantir's early dealings with the government, there was a conflict of interest between IT procurement officials and frontline operators. "The advice was to 'go with the flow' — those are the people paying you, why not do what they say? The problem is, if I did what they said, it wouldn't work." This "ornerness" — insisting on delivering what actually works for operators — is his own heresy.
Theme 2: American Reindustrialization — A Reversal from "Exquisite Few" to "Mass Production"
Sankar argues that the United States has transformed from the nation that was "the world's best at mass production" during World War II into a country that "manufactures an extremely small number of extremely exquisite things" — while its rival, China, is precisely the world's best at mass production today. This structural mismatch is deeply unsettling.
Historical Context and Data Chain:
- On the eve of World War II, the U.S. appointed Bill Knudsen (a Danish immigrant and master of mass production), the second-in-command at General Motors, directly as a three-star general responsible for war mobilization. He traveled across the country, meeting with other world-class engineers. Through the Lend-Lease Act, the U.S. had already begun mobilizing production before entering the war, and by the time of Pearl Harbor, it was operating at full capacity.
- Comparison with Germany: German engineers were more skilled than their American counterparts, but they only manufactured an extremely small number of extremely exquisite things.
- Key turning point — the "Last Supper" of 1993: After the Cold War victory, defense budgets were cut. The Secretary of Defense summoned 15 of the 51 major contractors and informed them: "Not all of you will survive. We are allowing you to merge." This triggered a consolidation frenzy from 51 companies down to 5.
- Data comparison: Before the fall of the Berlin Wall, only 6% of major weapons system spending went to "defense-only enterprises" (companies that did business exclusively with the Department of Defense), while 94% went to "dual-use enterprises" — Chrysler simultaneously built missiles and minivans, General Mills manufactured torpedoes and inertial navigation systems, and Kodak was a key supplier for the first spy satellite, Corona. Today, 86% of major weapons systems come from defense-only enterprises.
- Sankar's reasoning: "We lost the crazy people. They went to the tech industry, to other parts of the economy with enough positive-sum energy and growth. What remained were dividends, buybacks, and financial engineering, not real engineering."
The Urgency of Reindustrialization — Core of Sankar's "18 Theses":
- The biggest lie of globalization: "We do innovation; they do production." But innovation itself is a product of productivity — Google's motivation for the "Attention is All You Need" paper was to achieve a 3% incremental improvement in Google Translate. "If you don't have the input stimulus from improving production, you cannot actually capture innovation."
- China has evolved from "cheap labor" to an innovator: Wuxi AppTec went from being a cheap pipette contractor in the pharmaceutical industry to a company where 50% of clinical trial drugs are now created in China.
- 80% of U.S. generic drugs come from China in some form — "Today, your child has an ear infection, and you think, 'I'll just take some generic antibiotics.' But if a great-power competition erupts between the U.S. and China, the American people's will to fight will face the reality of 'my five-year-old might die from an ear infection.'"
Asymmetric Advantages and the Leverage of AI:
- Sankar believes the U.S. still has manufacturing capability — "Look at Elon and his successors. We do know how to make things here."
- AI is "David's slingshot": If it can make American workers 50 times more efficient than workers anywhere else, it will shift the efficiency frontier of "what can be manufactured in the U.S."
- Specific policy recommendation: Manipulate patent terms to incentivize pharmaceutical companies to move global supply of branded drugs back to the U.S. — "If you manufacture the global supply of your branded drug in the U.S., I will extend your patent term so you can recoup the additional capital expenditure. Every pharmaceutical company would agree immediately."
Theme 3: Palantir’s Cultural Code — Forward-Deployed Engineering and the "Quantum Organization"
Sankar argues that the core of Palantir’s success is not technology, but a culture that enables talent to achieve maximum growth under high pressure — marked by the "forward-deployed engineering" model and the "quantum org structure."
Origin and Mechanism of Forward-Deployed Engineering:
- Traditional software feedback loop: "I write code in Palo Alto, someone is willing to pay for it → it must have value." Sankar believes there is a stronger feedback loop: "Does this software have value? Did it deliver results?" — This can only be assessed on-site, alongside end users, on the factory floor, or in the trenches.
- Vector conflict between two types of engineers:
- "MacGyver" type: Their life highlight is "I solved the problem," believing that solving the right problem accounts for 80% of the work.
- "Artist" type: Focused on architectural elegance and scalability — "Don’t you see how beautiful what I built is?" — even if it solves a problem no one cares about.
- Forward-deployed engineers lean toward the MacGyver mindset, while product engineers lean toward the artist mindset. The "interface" between the two requires a "dictator-like" figure to manage — which Sankar calls his own "superpower": "The interface between forward-deployed engineering and product, one inch on each side, own it like a dictator."
"Quantum Organization Structure":
- Sankar opposes flat organizations and also opposes major reorganizations every five years. "The correct structure today is being eroded by entropy every day, and each day it becomes slightly more wrong than the day before."
- Ideal state: The organizational structure can "crystallize" into the optimal form for solving the current problem and can "reorganize" when the problem changes — "It’s like constantly tearing up the route map and reordering it based on new data."
The "Gamma Ray" Model for Unlocking Talent:
- Sankar uses the Hulk’s Bruce Banner as an analogy: "He doesn’t do progressive overload — adding a little more weight each week. It’s a near-lethal dose of gamma rays. 50% chance of death, 50% chance of turning into a big green monster."
- At Palantir, newcomers are "irradiated" — given high-risk projects far beyond their experience. "The maximum learning rate coincides with your maximum capacity to endure pain. You are only limited by your drive and your ability to tolerate discomfort."
- Key signal: Judging whether someone is worth betting on — "If I give this person an inch, can they turn it into a mile? The worst bet is: I gave them a mile, and they turned it into an inch."
- Superpowers and Kryptonite: Superpowers are "effortless" — like Superman’s ability to fly or see through walls; it’s not a hard struggle. Most people misjudge their superpowers when young, mistaking things that require considerable effort to succeed at as superpowers. "All your contributions to the world will come from your superpowers. Everything else is a waste of time." Kryptonite is "a weakness six standard deviations below the mean" — the only strategy is to avoid it, not to improve it.
Self-Disruption in the AI Era:
- Sankar believes the only way to avoid being disrupted is to disrupt yourself. "The most valuable thing is not to defend the moat, but to charge after new opportunities."
- AI adoption strategy: Let the youngest employees use it first — "I didn’t bother with those who were cynical or dismissive. I let the young people generate overwhelming productivity, almost in a way that shames or motivates others to catch up."
- The company holds "weeks of revolt" twice a year, allowing employees to do whatever they want — "The whole purpose is to prove that we’ve been messing up, and to inspire us with what you’ve built."
Theme 4: Capturing AI Value — The Chip Layer and the Ontology Layer
Sankar argues that AI value will primarily accumulate at the two ends of the stack: the chip layer and the ontology layer, rather than the model layer or application layer in between.
Mechanism Breakdown:
- Stack structure (bottom to top): Chips → Model Providers → AI Infrastructure (where the ontology layer resides) → AI Applications
- Model companies are forced to move upstream — pure models have become commoditized, and true value lies in the solutions built around them
- Pure AI application companies, in turn, move downstream — as they scale and process more customer data, they reinvent AI infrastructure from first principles (i.e., AIP)
- Palantir’s positioning: The platform it has built over the past 20 years (especially from Foundry in 2015 to AIP) happens to occupy this "ontology layer"
- Metric: The speed of delivering value to clients. "Two years ago, we could do it in eight weeks. Now it feels like one week. That rate is compounding dramatically."
Sankar’s Reasoning: "Even if these models help you write code faster, there is still a secret behind the code — the secret of why we write code that way. Where will we be in two years? We will be in the next phase two years from now."
Theme 5: US vs. China — Asymmetric Advantages and the "Undeclared Emergency"
Sankar argues that China's primary asymmetric advantage is its long-term planning capability, while the US core advantage is "unpredictability"—a quality that stems from heterodoxy rather than bureaucracy.
China's asymmetric advantage:
- Since the 1991 Gulf War, China has systematically mapped all key dependencies of US military technology and continuously invested in them—"the frog has essentially been boiled"
- China's concept of warfare is "deception"—"winning without firing a single shot." A large number of actions fall below the threshold the US considers conflict, but "this is absolutely conflict"
- Key evidence: China not only buys Brazilian soybeans (a commercial decision) but also attempts to smuggle agricultural fungi to prevent the US from growing soybeans
US asymmetric advantage:
- The AI phenomenon was not in China's long-term plan—"nor was it in our long-term plan. But when it happened, our entire economy pivoted in an instant."
- "Our own unpredictability—because we ourselves don't know what we're going to do—is our key advantage."
- Sankar's warning: "The Calvinist conscience of America is structurally positive-sum. We will turn the other cheek again and again. But there will be a line—at some point we just react. And when we react, the best quote is: 'Observe, overreact, destroy, apologize.'"
The "Undeclared Emergency":
- Sankar believes the US has lost its deterrence capability—Crimea in 2014, militarization of South China Sea reefs, the Israeli massacre, the invasion of Ukraine—"over the past 10 years, we have failed to deter many conflicts"
- Deterrence has begun to recover in the past year—Operation "Midnight Hammer" demonstrated overwhelming precision strike capability; the Maduro operation was conducted on the same day Chinese personnel were present, and Chinese and Russian weapons in Venezuela "seemed completely ineffective"
- But Sankar emphasizes: "The deterrence capability against a peer competitor is: can you generate that kind of strike every day until the enemy yields to your will?"
Mentioned Positions
| Position |
Guest Stance |
Key Data |
| Palantir |
Bullish (core holding, but emphasizes "not a defense company, but a software company") |
4,500 employees; AIP delivery cycle compressed from 8 weeks to 1 week; two "rebellion weeks" per year |
| Anduril |
Bullish (as a benchmark for the new model) |
Builds products with its own specifications and own risk; "You should buy it, if you like it" |
| SpaceX |
Positive mention (as an early companion "walking alone in the dark") |
Spent the first 10 years "alone in the dark" alongside Palantir |
| Apple |
Neutral (as a case study) |
Invested the equivalent of 2.5 Marshall Plans in China over the past 5 years to build talent and capacity |
| Google |
Neutral (as a case study) |
The "Attention is All You Need" paper originated from an incremental improvement to boost Google Translate by 3% |
| Lockheed (Skunk Works) |
Positive (as a paragon of heresy) |
Kelly Johnson built 41 airframes in his lifetime; the U-2 was completed in 13 months |
| General Mills |
Positive (historical case) |
Once manufactured torpedoes and inertial navigation systems |
| Kodak |
Positive (historical case) |
Was a key supplier for the first spy satellite, Corona |
| Chrysler |
Positive (historical case) |
Simultaneously built missiles and minivans |
Judgments Worth Remembering
1. Sankar on Heretics and Bureaucracy: "All change comes from heretics, who later become heroes. Looking back at history, the only things that truly worked—the things that helped us win all our wars—were actually done by heretics. What was produced through the machine delivered nothing."
2. Sankar on the "Gamma Ray" Model of Talent Unlocking: "The Hulk, Bruce Banner, is not about progressive overload—it's about a near-lethal dose of gamma rays. 50% chance of death, 50% chance of turning into a big green monster. At Palantir, we irradiate people. The maximum learning rate coincides with the maximum pain you can endure."
3. Sankar on Superpowers and Kryptonite: "A superpower is effortless—like Superman can fly and see through walls. All your contributions to the world will come from your superpowers. Everything else is a waste of time. Kryptonite is a weakness six standard deviations below the mean—the only strategy is to avoid it, not to improve it."
4. Sankar on the Structural Mismatch Between the US and China: "During World War II, we were the country best at mass production. Today, we are like Germany back then—making a very small number of extremely refined things. And our adversary is the country best at mass production in the world today. That should make everyone uneasy."
5. Sankar on the Biggest Lie of Globalization: "'We do innovation, they do production.' But innovation itself is a product of productivity. Google's motivation for 'Attention is All You Need' was to improve Google Translate by 3%. If you don't have the input stimulus from improving production, you cannot actually capture innovation."
6. Sankar on AI as "David's Sling": "If we can make American workers 50 times more efficient than workers anywhere else, we will change the efficiency frontier of 'what can be made in America.' We should become reindustrialization maximizers."
7. Sankar on "Doing" vs. "Being" (John Boyd's Binary Choice): "You can either 'be somebody' or 'do something,' but you cannot have both. If you want to 'be somebody,' you have to play the game—it's all performance, you'll get promoted, you'll get followers, you'll feel good. Or you actually do something—it will feel terrible, no one will appreciate you, but you have the intrinsic reward of knowing 'you did something.'"
8. Sankar on the Asymmetric Advantage of the US vs. China: "China's asymmetric advantage is long-term planning—they actually cannot pivot on a dime. Our asymmetric advantage is that 'we ourselves don't know what we're going to do'—this unpredictability comes from heretics, not from bureaucracy. The AI phenomenon was not in their long-term plan, nor in ours. But when it happened, our entire economy pivoted in an instant."