Jul 24, 2026 · 41m · american-optimist
America’s $700 Billion Defense Tech Rebuild
Emil Michael · 11m spoken Kevin Mandia · 11m spoken Joe Lonsdale · 9m spoken Shyam Sankar · 6m spoken
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Hosted by Joe Lonsdale, this episode examines the critical modernization of American national security through in-depth discussions with defense leaders Shyam Sankar, Emil Michael, and Kevin Mandia. Together, they explore how artificial intelligence, autonomous weapons, domestic industrial reshoring, and next-generation cybersecurity are reshaping the balance of global power.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Joe holds 23.7% of the talking time here. How this is scored →
Joe as informed peer 5.4 Guest teaching 3.8 Guest disagreement 1.1 Joe pushing back 1.0
speaking balance: gold is Joe, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 8:14 Sankar rejects rogue military conspiracy theories
Sankar emphatically dismisses the host's devil's advocate suggestion of rogue actors in the Pentagon, condemning cynical media tropes that caricature defense personnel as corrupt villains.
Hardest push from Joe ▶ 6:58 Lonsdale pushes devil's advocate rogue actor scenarioLonsdale presses Sankar on public fears of unauthorized rogue AI killer robots developed inside the defense establishment, challenging Sankar's optimistic safety assumptions.
Biggest teaching moment ▶ 5:28 Sankar reframes autonomous weapons debate via AegisSankar re-educates the audience and host by showing that autonomous kill chains have operated safely since the 1970s, dismantling the premise that autonomous weapons represent an unprecedented moral crisis.
Joe holds their own ▶ 17:06 Lonsdale demonstrates early foresight founding EpirusLonsdale establishes deep personal domain authority by detailing how he foresaw asymmetric drone swarm stadium attacks back in 2018 and co-founded high-power microwave firm Epirus to solve it.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Joe as informed peer | Guest teaching | Guest disagreement | Joe pushing back | Why |
|---|---|---|---|---|---|---|
| AI in Doctrinal Targeting and Accelerating the OODA Loop | 5 | 4 | 1 | 1 | Lonsdale opens the defense AI topic and notes Detachment 201 and private sector tensions. Sankar provides an in-depth historical lesson using Manhattan Project scientist Theodore Hall to illustrate the peril of tech hubris without democratic accountability. | |
| Autonomous Weapons Systems and Human-in-the-Loop Realities | 5 | 5 | 2 | 1 | Lonsdale asks how to handle human-out-of-the-loop decisions in naval drone warfare. Sankar reframes the entire issue, noting autonomous weapons have existed since the Aegis system in the 1970s and calling modern hysteria a miscategorization of degree versus kind. | |
| Debunking Rogue AI Conspiracies and Trusting Military Integrity | 4 | 6 | 4 | 3 | Lonsdale plays devil's advocate suggesting a rogue Pentagon bureaucrat might secretly program killer robots. Sankar firmly pushes back against the cartoon caricature of corrupt defense officials, insisting that moral uniformed professionals, not bureaucratic systems, ensure ethical conduct. | |
| Streamlining Defense Priorities: The Six Critical Technologies and Applied AI | 6 | 2 | 0 | 0 | Lonsdale shows granular familiarity with the CTO's critical technologies list, naming specific startups like Gallatin and Palantir's C2 contracts. Michael explains consolidating defense priorities from over twenty down to six focused operational areas. | |
| Biomanufacturing and Rebuilding the Domestic Industrial Supply Base | 7 | 3 | 1 | 2 | Lonsdale highlights his venture investment in biomanufacturing startup Cambium and drills down into terminal velocity physics between ballistic missiles and true hypersonics. Michael clarifies the maneuverability and atmospheric reentry differences. | |
| Scaled Directed Energy and Neutralizing Drone Swarm Threats | 7 | 2 | 0 | 1 | Lonsdale references founding directed energy firm Epirus to combat drone swarms as early as 2018. Michael validates this threat model, discussing microwave shielding and the homeland Golden Dome defense initiative. | |
| Pentagon Procurement Reform and the $200B Strategic Capital Fund | 6 | 3 | 0 | 1 | Lonsdale recalls how nontraditional contractors previously had to litigate against the government for market access. Michael details how the newly stood-up Office of Strategic Capital deploys low-cost debt to crowd private venture capital into domestic minerals and components. | |
| Tech Startup Onramps, DARPA Innovation, and Defense Talent Acquisition | 5 | 2 | 0 | 0 | Lonsdale explores talent acquisition hurdles inside the Pentagon and quips about imposing term limits like DARPA's across civil agencies. Michael describes commercial-first software adoption to leverage external frontier engineering without rebuilding foundational models. | |
| The AI Cyber Revolution: Autonomous Attack Swarms and Asymmetric Threats | 5 | 4 | 1 | 1 | Mandia warns of AI-automated cyber offensive swarms operating at inhuman scales. Lonsdale verifies the speed of recent breakthroughs, citing social engineering attacks and voice cloning vectors. | |
| Nation-State Restraint and Shielding Critical National Infrastructure | 5 | 4 | 1 | 1 | Lonsdale categorizes cyber bad actors into financial thieves, vandals, and nation-states, asking which poses the greatest threat. Mandia highlights conflict thresholds, warning that while nation-states hold back in peacetime, critical utility infrastructure remains vulnerable during kinetic escalation. | |
| Moving Beyond Compliance Dashboards to Realistic Red Team Simulation | 6 | 4 | 1 | 1 | Lonsdale asks whether safety guardrails in commercial LLMs like Claude hinder testing compared to unconstrained Chinese models. Mandia reveals that autonomous offensive agents routinely surprise internal red teams and drive developer output up by more than a hundredfold. | |
| The Future of Autonomous Red Teaming for Corporate Cybersecurity | 4 | 6 | 2 | 0 | Mandia delivers a blistering critique of corporate cybersecurity compliance dashboards, ridiculing four-dimensional pie charts that mask true vulnerabilities. He explains that empirical, continuous red teaming is the only authentic measure of corporate resilience. |
Statements from this episode (31)
Sankar: Epic Fury is a leapfrog, with 10x-100x improvement still ahead
“And I would say Epic Fury is a leapfrog in capability, but I think it also shows that there is another 10 X, hundred X to go that's illuminated by what's already been done here.”
Theodore Hall shares blame for deaths caused by communism
“Every death since 1949 from communism, which is the year the Soviets got the bomb, you could say that Blood is at least partially on Theodore Hall's hands.”
Technologists unilaterally deciding global policy is 'tyranny by tech bro'
“You know, cause the alternative, you could say Theodore Hall was essentially tyranny by tech bro. You know, he decided global policy and it had huge consequences. On the other end, there's a true deep commitment to democracy, even when it feels uncomfortable.”
The US military has deployed autonomous weapons since the 1970s
“We've had autonomous weapons since the seventies and eighties. You know, you think about, like, the Aegis system that defends a carrier when it senses an incoming threat coming at the speed of a missile, it's firing off to protect the ship, and you don't have …”
Sankar: Modern AI warfare is a difference of degree, not kind
“So I would posit that this is more of a difference of degree than kind. And because we're miscategorizing it as a total difference of kind, we kind of have some deranged conversations around it.”
A garage PhD student poses more AI danger than military personnel
“I think like a random PhD student in a garage is probably more danger with AI than anyone in these that we were serving within the department of war.”
Sankar: Robotic weapons will not eliminate human soldiers in warfare
“The part, I don't buy the idea that these sorts of robotic weapons are going to somehow mean that humans don't have to do the fighting. You certainly want the robots to draw first blood. You want them to serve as a deterrent, but the will of the nation to actu…”
A Pacific conflict will force the military into extreme remote operations
“And if there's ever a conflict in, in, in the Pacific, we'll, by definition, be remote. And frankly, you know, the most remote place you could imagine.”
Biological organisms can produce TNT, gunpowder, and other energetics
“Biological organisms can create TNT, gunpowder, you know, energetics, if you will.”
Michael: Adversaries Dump Rare Earths to Create US Tech Dependencies
“And we've been behind again because some of our adversaries have been dumping Dumping, like with rare earth metals to get us dependent on battery technologies, on critical minerals, pharmaceuticals, you name it.”
Missiles fired by Iran at Israel are not hypersonic or maneuverable
“Not maneuverable, but they're not going at hypersonic speeds also.”
Michael: US must produce cheap hypersonics at scale to deter China
“So we know how to build our own. But we have to do it more cheaply. We have to do it at scale. We have to do it, and it's a deterrent effect, right? It's just like we had nuclear missiles, and the Russians did, or Soviets did. If China, the China has them, we …”
The Pentagon issued an RFI for space-based missile interceptors
“We just did a RFI for space-based interceptors, which is how do you put something in space that could take something down in space.”
Palantir, Anduril, and SpaceX sued the government for initial major contracts
“And one of the examples I use Palantir and Roland, SpaceX, all had to sue the government for their first big contract, because the primes were sophisticated at selling, sophisticated at protecting their interests.”
Office of Strategic Capital has $200B for defense and critical minerals
“Half of it, half of two hundred billion dollars. So a hundred billion dollars is to make sure that we have solid rocket motors, brushless motors, batteries, all of the components that go into our weapon systems. The other half of it's for critical minerals, an…”
Emil Michael: Office of Strategic Capital Loans Cost Treasuries Plus 100 Bps
“So it's a basically treasuries plus a hundred bips.”
Emil Michael: The United States Currently Has 250 Medicines in Shortage
“We have 250 medicines on shortage in the United States.”
The Pentagon will not spend $100B building in-house frontier AI models
“For these frontier AI companies, for example, I'm not gonna try to build our own model here and spend a hundred billion dollars. Why would I? Why wouldn't I work with the best minds in industry, have them make something that, for our use case, that builds on w…”
Pentagon Is Deploying Frontier AI Integrated With Intelligence to Warfighters
“I'm not only taking the frontier models and just making them available. I'm doing it across classification levels. I'm combining some of the most, you know, important data that the United States has from an intelligence standpoint, we're deploying it to warfig…”
Every DARPA employee has their exit date on their badge
“So everyone at DARPA walks around with their exit date on their badge.”
Six-second voice clones will automate social engineering attacks against help desks
“You're gonna start seeing social engineering automated, where six seconds of Joe Lonsdale's voice is gonna be used to get a help desk to do something.”
Fully autonomous cyber defense systems will arrive in two to three years
“I think AI will advantage the offense in the near term, but everybody knows that and everybody's waiting to be able to, we're gonna have a total shift change in cybersecurity where you'll have AI agents on offense Being the cheapest way to create that autonomo…”
Major nations will restrain their advanced AI cyber offensive capabilities
“And in reality, no modern nation, even the folks that can create this capability, China, Russia, others will leverage it. Even they don't want to really, no one's ready for it.”
Defensive teams will run offensive AI agents to train defense systems
“There is no question that In the future, you'll have AI agents on offense run by the good guys to train the AI on defense run by the good guys.”
Mandia: Criminals will inevitably deploy ungated frontier AI models
“The models may have gates that prevent you from doing things, but even those folks that make them probably need to remove some of the gating factors because it's so important to make sure there will be models that can think, learn, and do incredible things tha…”
Mandia: Frontier AI models are outpacing human network security testers
“The frontier models are doing an exceptional job already in helping test the security of networks. And they're changing so fast, Joe, that I'm very confident regardless of which frontier model you use you're gonna find it's outpacing humans doing the work, you…”
Mandia's startup is building an offensive nation-state hyper-attack platform
“We want to build the whole hyper attack platform. Everything a nation state would build on offense. We want to build.”
Coordinated AI agent cyber swarms are coming at inhuman speeds
“That drone swarm in cyber will be coming, and the difference with that total recall it has that humans don't have, the instantaneous communication between different agents that have different objectives so they can coordinate at a speed humans cannot, this has…”
AI-native developers are over 100x more productive than traditional engineers
“It's not 10 X. It's more than a hundred X productivity with an AI native developer working on software. We've had over a hundred commits in a day building our software. We have a different demo every two to three days. I never had that before.”
AI attack agents will operate 1,000 times faster than humans
“The day where a human would break in, and if you were 10 minutes behind the human on defense, you could stop the impact of a breach, is going to end. There's going to be an agent that breaks in, then another agent uploaded because it had remote code access, Re…”
Mandia: Resisting a Five-Day Offensive Red Team Is Peak Security
“If the answer is no, feel good at three to four former operators on offense. Couldn't break into your network in five to six days. It's pretty, I'd feel pretty damn good. That's the best you can do, by the way. I don't know what else you're supposed to do. And…”