This Week in AI

Fighting Foundation Models Is a Losing Strategy

Stop Fighting Foundation Models. Do This Instead.

The bluntest strategic warning of the week: if your AI company is positioned against foundation models rather than with them, you've already lost. The framing isn't about coexistence — it's about actively finding ways to leverage the tailwinds these models create.

• The speaker draws a hard line: companies that haven't figured out how to collaborate with foundation models are "screwed"
• Vertical AI companies with the most upside are those that reach deeper into the stack to control their own P&L and differentiate at the infrastructure layer
• Remaining dependent on foundation model providers without stack ownership leaves you exposed to margin compression and commoditization

"If you are fighting against them, you've already lost. If you haven't figured out how you're going to win with them... then you're screwed."

---

The $50M R&D Trap Is Killing Defense Startups

a16z's David Ulevitch: Why the $50M R&D trap keeps killing defense startups

Government R&D funding sounds like a lifeline for small defense companies — David Ulevitch argues it's closer to a death sentence. The overhead of auditing, accounting, and compliance that comes with federal dollars is manageable for large primes but crushing for startups.

• A $50M government R&D grant brings compliance burdens that are "almost impossible" for small companies to absorb
• Large defense primes love the system — they simply charge more to cover overhead, passing costs to taxpayers while locking out nimble competitors
• Ulevitch's prescription: private capital funds R&D, government commits to buying finished outcomes — eliminating bureaucratic drag while keeping risk with investors
• His case for VCs as better pickers: "We've been doing it a lot longer. We know the founders. We're able to evaluate them."

"All we need from the government is we don't need them to fund R&D. We need them to commit to buying outcomes and buying solutions."

---

Anduril Is Running a Marvel Cinematic Universe for Weapons

How Anduril uses anime trailers to tease its next weapon

Anduril has turned defense product launches into a serialized cinematic universe, hiding post-credits teases of future weapons inside anime-style promotional videos. The Fury video teased Copperhead; the Copperhead video picked up exactly where Fury left off — and teased an autonomous surface vehicle.

• The Copperhead video literally continues the narrative from the Fury video's final frames, an unprecedented storytelling move for a defense contractor
• The surface vehicle tease was confirmed this week via an official Anduril-Hyundai partnership to build autonomous surface vessels
• New entrant puts Anduril in direct competition with Saronic in the autonomous maritime domain

"the video for the Copperhead products literally picked up where that video ended."

---

Mosaic Warfare Is the Strategic Logic Behind Anduril's Expansion

How Anduril uses anime trailers to tease its next weapon

The surface vehicle partnership isn't a random product bet — it's the next node in a deliberate multi-domain network strategy. Every new Anduril platform is designed to plug into a single command-and-control fabric spanning air, sea, and subsurface.

• The doctrine: "mosaic warfare" — a military internet of things where every domain has an autonomous asset woven into a unified response network
• Air (Fury), munitions (Copperhead), surface (Hyundai ASV) — each announcement fills a domain gap in the mesh
• The strategic moat isn't any single platform; it's the integrated network that ties them together

"You want every domain, every surface, every force level to have some relevant asset that you can weave into this network of potential responses."

---

Brilliant Teams Still Fail Without Organizational Architecture

Max Levchin on why brilliant people alone won't save your startup

PayPal co-founder Max Levchin delivers a counterintuitive warning: talent density is necessary but not sufficient. Without mission alignment and leadership that correctly combines individual skills, even exceptional teams collapse.

• ~85% of US startups fail — Levchin uses this baseline to frame how easy failure is, even with great people
• The failure mode isn't lack of talent; it's failure to organize, align, and harness what you have
• Leadership's core job: give people a mission that "feels true to them" while aligning the group toward a single direction

"Even if you do have the people and you don't harness their unique skills and kind of don't correctly combine them, you might fail anyway."

---

AI Shifts Humanity From Consumers to Creators

The era of the artist (2026-05-11)

The most optimistic framing of AI's cultural role this week: not productivity tool, but universal creative equalizer. The speaker argues AI removes the technical barriers that historically separated those who consumed culture from those who made it.

• The through-line: da Vinci, Van Gogh, Walt Disney, Steve Jobs — all driven by intrinsic motivation, not validation or reward
• Van Gogh sold one painting in his life; da Vinci carried the Mona Lisa until he died, never showing it publicly
• Disney spent his last day in a hospital bed using ceiling tiles as a grid to plan Disney World — the archetype of vision untethered from circumstance
• AI's promise: "Everyone can make now" — the creative drive that defined these figures is no longer gated by technical skill

"We don't have to be consumers anymore. We can be creators. Everyone can make now. It's amazing."

---

Teotihuacan Dismantles the Myth of Old World Technological Superiority

They Built This Without Wheels or Metal - David Reich

Geneticist David Reich uses Teotihuacan as a case study in convergent human achievement — a civilization as monumental as ancient Egypt, built without metal tools, draft animals, or wheels. The implication cuts against any narrative of inherent Old World advantage.

• The peoples of the Americas separated from East Asian ancestors ~20,000 years ago and from West Eurasians ~40,000 years ago — yet independently produced monumental architecture from the same shared biological and cultural starting point
• Reich's framing: "a long fuse delay" — the same toolkit, different timelines
• His prescription for cultural bias: take anyone who holds Old World superiority "to these places and they will not have it anymore"

"It's without metal. Not only without metal, but it's without animals and without wheels, which is crazy."

---

Key Takeaways

• Collaborate with foundation models or die — vertical AI companies that fight the tide instead of owning their stack deeper are structurally disadvantaged and running out of time.
• Defense procurement is broken by design — the overhead burden of government R&D funding protects large primes at the expense of startups and taxpayers; outcomes-based contracting is the structural fix.
• Anduril is building a domain-spanning autonomous network, not just products — each new platform (air, munitions, surface) is a node in a mosaic warfare mesh, and the cinematic launch strategy is a deliberate signal of that roadmap.
• Talent without organizational architecture fails — Levchin's 85% failure rate stat is a reminder that assembling brilliant people is the floor, not the ceiling; leadership that aligns and combines skills is the differentiator.
• AI's deepest promise is creative democratization — the shift from passive consumption to active creation is the cultural argument for AI that goes beyond productivity metrics.
• Convergent human achievement is underrated — Reich's genetics framing of the Americas as an independent civilization experiment with the same biological toolkit challenges technological determinism and Old World exceptionalism.

---

Sources

• The era of the artist
• How Anduril uses anime trailers to tease its next weapon
• Stop Fighting Foundation Models. Do This Instead.
• Max Levchin on why brilliant people alone won't save your startup
• They Built This Without Wheels or Metal - David Reich
• a16z's David Ulevitch: Why the $50M R&D trap keeps killing defense startups