Uncapped with Jack AltmanAnduril & Founders Fund’s Trae Stephens on Choosing Good Quests in the Age of AI | Ep. 35
CHAPTERS
- 0:00 – 0:35
Why “If you build it, they will come” fails in defense contracting
Trae opens with a vivid example to explain how defense procurement works: novel tech doesn’t automatically translate into government adoption. He argues that without a deliberate go-to-market strategy inside DoD acquisition, incumbents can absorb ideas and win the contract anyway.
- •Defense procurement isn’t a meritocracy of best product; process and contracting dominate
- •Incumbents can outbid or “rebuild from white paper” what startups invent
- •Selling to government requires navigating acquisition pathways, not just building tech
- •The core challenge is getting the government to buy from non-traditional players
- 0:35 – 3:05
Choosing “good quests” when AI makes shallow startups too easy
Jack tees up Trae’s “Choose Good Quests” idea as especially relevant in the AI gold rush. Trae argues AI’s distortion is less about enabling hard problems and more about making low-effort, crowded “LLM wrapper” companies too easy to start.
- •AI increases the supply of easy, low-differentiation company ideas
- •“Whiteboard founding” leads to consensus categories with dozens of competitors
- •Building primarily to get rich is framed as the opposite of a good quest
- •Opportunity cost: top talent pulled away from high-impact work
- 3:05 – 4:09
Opportunity cost and the ‘level 100 wizard’ problem
Trae extends the “good quest” argument into a resource allocation critique: society’s most capable builders are finite. He wants top-tier talent aimed at foundational, civilization-advancing work rather than what he calls “AI slop.”
- •Framing founders as RPG characters highlights scarcity of elite talent
- •Misallocation to trivial products reduces progress on hard infrastructure problems
- •Examples of higher-impact quests: semiconductors, deep tech manufacturing
- •AI boom resembles prior crypto booms, but may not self-correct as cleanly
- 4:09 – 8:03
AI at moral boundaries: grief tech, companions, and social side effects
The conversation shifts to AI products that trigger visceral discomfort, like recreating deceased loved ones or AI companions. Trae distinguishes personal appeal from societal risk, focusing on how these tools may reshape relationships, dating, and social stability.
- •Personal motivation: desire to speak to lost loved ones can be genuine
- •Risk: AI companions may reduce incentives to sustain real relationships
- •Online dating already pushed sorting toward shallow metrics; AI could amplify effects
- •Concern about large populations of ‘unmatched’ young men and social instability
- •Long-run implications include birth rates and broader social cohesion
- 8:03 – 10:32
A practical ethics framework: the ‘feels vs is’ two-by-two
Trae offers a decision framework: ‘feels good/bad’ versus ‘is good/bad.’ He places defense and social order in the ‘feels bad, is good’ quadrant, while highlighting how addictive vices can land in ‘feels good, is bad.’
- •Two-by-two: feels vs reality clarifies why people disagree
- •‘Feels bad, is good’: duty-driven domains like defense, discipline, law & order
- •‘Feels good, is bad’: gambling, pornography, addictive hedonism as businesses
- •Policy is often the only lever to constrain harmful-but-profitable markets
- •Edge cases (e.g., prediction markets) can have real societal value
- 10:32 – 11:28
How regulation actually happens (and why tech must engage Capitol Hill)
Trae argues policy formation is reactive: technology advances first, society discovers harms, then guardrails follow. He notes Congress has limited technical expertise, so industry engagement and real-world evidence drive workable regulation.
- •Policy typically follows innovation; guardrails are built after boundary-crossing
- •Capitol Hill lacks deep AI expertise; few have CS backgrounds
- •Democracy works best when societal debate shapes constraints over time
- •Defense has longstanding autonomy precedent, so some AI governance is pre-baked
- 11:28 – 16:53
Working with regulators in defense: bipartisan reality vs Valley perception
Jack asks how Anduril experiences regulation and Washington compared to Silicon Valley’s cultural skepticism. Trae explains defense policy is unusually bipartisan, and Anduril’s biggest pushback historically was recruiting and cultural legitimacy in tech, not DC.
- •Defense is one of the most bipartisan areas (e.g., NDAA passing reliably)
- •Autonomy has existed for decades (e.g., CIWS), with established legal norms
- •Silicon Valley backlash peaked around Project Maven and anti-DoD sentiment
- •Anduril focused on articulating alignment with just war theory to recruits
- •Debate about military power is healthy and improves policy outcomes
- 16:53 – 18:32
What Anduril still has to prove: manufacturing at tens-of-thousands scale
Trae says Anduril’s next leg is production, not prototypes: moving from dozens to tens of thousands of units. He describes the need for design-for-manufacturing and introduces Arsenal One, a major Ohio manufacturing campus planned to come online in phases.
- •The core bottleneck is scaling manufacturing capability, not concept validation
- •Design decisions must enable cost-down and high-volume production
- •Arsenal One: planned multi-million-square-foot manufacturing campus in Ohio
- •Venture-backed companies often lack the muscle for industrial-scale execution
- 18:32 – 22:23
Lessons from Tesla and SpaceX: the hidden difficulty of scale and verticalization
Jack presses on why scale is so hard; Trae contrasts building a first product with building a million. SpaceX’s approach—verticalizing critical components to control risk and quality—illustrates why supply chains and reliability dominate at scale.
- •Prototype success doesn’t translate to mass production; it’s a different mindset
- •Manufacturing optimization becomes central (throughput, welding, sourcing, QA)
- •SpaceX verticalizes because supplier failure can be catastrophic
- •In weapons systems, cost reductions often come from manufacturing choices
- •Scaling requires material supply strategy as much as engineering design
- 22:23 – 24:47
The future of warfare: low-cost autonomy across every domain
Trae predicts warfare will shift toward distributed, low-cost autonomous systems in air, land, sea, and space. He argues legacy force projection (e.g., aircraft carriers in great-power conflict) faces a brutal cost-exchange disadvantage against modern missiles.
- •Space matters, but every domain will see low-cost autonomous platforms
- •Aircraft carriers lose viability in great-power conflict due to cost asymmetry
- •US operations likely remain Special Forces-heavy with autonomy doing the rest
- •Autonomy targets ‘dull, dirty, dangerous’ tasks to reduce casualties
- •Ukraine–Russia highlights that high-casualty wars still exist despite trends
- 24:47 – 29:06
Juggling Anduril and Founders Fund: the accidental ‘investor-operator’
Trae explains he didn’t set out to be an investor-operator; Anduril emerged from Founders Fund’s search for a hardware-enabled defense prime. He describes how he manages both roles through priority-setting and intense context switching.
- •Anduril began as a gap identified from investing in Palantir and SpaceX
- •Founders Fund encouraged Trae to build what didn’t yet exist
- •Day-to-day management relies on strong operational support and prioritization
- •Operating provides depth that complements VC’s breadth
- 29:06 – 31:15
Founders Fund’s system: founder-first philosophy and incentives for depth
Trae unpacks Founders Fund’s name: not just founders investing, but a fund designed to back founders rather than replace them. He claims the firm’s pattern of partners building companies is an emergent outcome of a culture that rewards going deep, not an incubation strategy.
- •“Fund for founders”: invest because the founder is the best operator for the job
- •Low desire for heavy control; fewer board seats than typical VC firms
- •Reactive support: introductions, pricing, strategy—without over-governance
- •Partner-built companies (e.g., General Matter, Varda, Anduril) as ‘exhaust’ of depth
- •Time/permission to explore enables deep thesis development
- 31:15 – 37:29
Why Founders Fund performs: access, debate-driven evaluation, and contrarian taste
Trae attributes performance to edges in three buckets: access, evaluation, and follow-on. He emphasizes open debate without hierarchy, avoidance of consensus categories, and a willingness to take technical risk while betting heavily on exceptional founder-CEOs.
- •Access edge rooted in early fund history, Peter Thiel brand, PayPal network
- •Evaluation culture: rigorous debate, no sacred cows, hierarchy minimized
- •Preference against consensus/competitive markets; ‘competition is for losers’
- •Founder-CEO focus; strong willingness to underwrite big tech risk early
- •In competitive AI, exceptions require ‘absolute killer’ founders (e.g., Cognition)
- 37:29 – 43:55
King-making, ‘kamikaze rounds,’ and the discipline of concentration
Jack asks about brand-driven king-making; Trae warns oversized, overpriced rounds can become anchors that destroy companies. He then explains Founders Fund’s concentration strategy: proactively build big winners and avoid scattering capital into small, non-moving checks or automatic pro-rata follow-ons.
- •King-making is less relevant if you avoid crowded categories, but still debated
- •‘Kamikaze rounds’: huge checks at inflated valuations that can harm companies
- •Founders must treat fundraising as long-term strategy, not just price maximization
- •Concentration goal: top positions can become 40–50% of a venture fund
- •Tactics: proactively add capital to best companies; don’t auto-follow mediocrity
- •Founder tip: if a company asks you not to do pro rata, that can be a bullish signal
- 43:55 – 47:21
Where the alpha is now: anti-FOMO, avoiding categories, and insisting on real origin stories
Trae says venture is hard and ‘alpha’ comes from resisting hype and moving before categories form. He describes Founders Fund as low-regret and anti-mimetic, using a founder’s origin story as a quick test for authenticity versus trend-chasing.
- •Strategy: stay slow, ignore hype; good deals exist in every vintage
- •Founders Fund is structurally ‘no-FOMO’ and rarely dwells on missed rounds
- •Categories are a lagging indicator—once a category exists, it’s often too late
- •Mimesis awareness matters: ask for the origin story first
- •Good companies have compelling narratives; ‘whiteboard founding’ is a red flag
- 47:21 – 52:24
Traditional faith revival in tech and just war theory’s Christian roots
In the closing segment, Jack asks about Trae’s openness regarding Christian faith and why it resonates now. Trae points to a post-COVID revival of interest in traditional religion and connects Christianity’s intellectual tradition to just war theory and Western laws of armed conflict.
- •Faith comes up organically; Trae isn’t afraid to discuss it publicly
- •Post-COVID shift: renewed interest in traditional faith after ‘spiritual but empty’ trends
- •Tech culture stereotypes Christians; curiosity rises when respected peers are believers
- •Distinction between cultural Judaism and belief-requiring Christianity in coastal tech
- •Just war tradition (St. Augustine) underpins Western law of war and informs Anduril’s moral framing
