No PriorsNo Priors Ep. 19 | With Anduril CEO Brian Schimpf
CHAPTERS
- 0:05 – 4:38
Brian Schimpf’s path: coding, DARPA self-driving cars, and Palantir to Anduril
Brian recounts getting into coding young, then experiencing the rapid progress of robotics during the DARPA Grand/Urban Challenges. He describes how his time at Palantir—especially forward-deployed work with government customers—shaped his interest in national security and ultimately led to founding Anduril with longtime collaborators.
- •Early passion for software; coding since age 12
- •Robotics acceleration during DARPA self-driving competitions
- •Palantir experience: mission-driven national security problems
- •Co-founding team origins and why the timing felt right (2017)
- 4:38 – 6:09
Why Anduril’s thesis was “lower cost, higher volume” systems powered by software
The conversation shifts to Anduril’s founding thesis: traditional primes excel at exquisite, expensive platforms, but future conflicts demand many cheaper systems. Brian emphasizes that intelligent software is the key enabler and that defense ultimately buys integrated capabilities, not just components.
- •Traditional defense strengths: carriers/jets; weakness: speed and software iteration
- •Future need: higher quantities of lower-cost systems
- •Software as the critical bottleneck for scaling capability
- •Defense procurement favors end-to-end capability (hardware + software + networking)
- 6:09 – 8:24
Lattice and early deployments: sensing, classification, and broad situational awareness
Brian explains Anduril’s core software platform, Lattice, as a way to ingest and interpret massive sensor data to track objects and activity. He connects that foundation to initial product areas like base protection and border security, then to a growing suite of autonomous platforms.
- •Lattice’s goal: understand “everything going on” from many sensors
- •Early focus: force protection/base protection and border monitoring
- •Classification over wide areas (people, vehicles, animals)
- •Expansion path from small drones to larger autonomous systems
- 8:24 – 13:34
Why higher-volume systems matter: manpower limits, industrial capacity, and lessons from Ukraine
Sarah presses on the rationale for higher volume, and Brian ties it to real constraints: pilot shortages, recruiting shortfalls, and limited industrial capacity to build large platforms. He argues modern air defenses and attrition rates make exquisite systems harder to use, while battlefield transparency pushes forces into distributed, smaller units.
- •Manned exquisite systems are costly and constrained by personnel shortages
- •Industrial base limits: shrinking capacity to build subs and other high-end platforms
- •Ukraine: high air-defense risk changes how air campaigns and attrition work
- •Commercial and military ISR reduce ability to hide; drives disaggregation and scale
- 13:34 – 16:04
Choosing where to apply a general-purpose sensing stack: sensors, simulation, and urgent gaps
Brian frames defense sensing as a general problem with a small set of sensor modalities, making software reuse powerful across missions. He then describes Anduril’s approach to picking opportunities: model the battlefield, identify capability gaps, and prioritize areas where urgency and iteration speed are decisive.
- •A small set of core sensors (EO/IR, RF, radar) underpins many missions
- •Reusable algorithms/tooling can be adapted across environments and targets
- •Defense as “analyzable”: simulate outcomes and identify capability gaps
- •Prioritization: urgent problems where speed of adaptation matters most
- 16:04 – 18:48
Counter-drone as a “software-speed” battlefield: rapid iteration and integrated defeat chains
Using counter-drone as a case study, Brian describes how cheap drones created a pervasive air-defense challenge, even for non-state or lower-tier actors. He argues success requires quickly updating detection/classification models and integrating multiple defeat options (jamming, guns, missiles) into a single operational package.
- •Drones as low-cost cruise-missile analogs, proliferating widely
- •Adversaries iterate in months, not decade-long programs
- •Deploy hardware once; iterate software/models rapidly in the field
- •Integrated counter-UAS kill chain: detect, classify, and defeat via multiple methods
- 18:48 – 20:41
Autonomy in defense: human accountability, trust, and bounding non-determinism
Elad asks about autonomy, and Brian outlines why U.S. doctrine and accountability constrain fully autonomous lethal decision-making. He emphasizes operator trust, predictability, and a crawl-walk-run adoption path, noting that non-deterministic techniques (RL, LLMs) require careful bounding and controls.
- •U.S. preference: human accountability for lethal force decisions
- •Adoption bottleneck is trust and predictability for operators
- •Crawl-walk-run approach to increasing autonomy
- •Non-determinism (RL/LLMs) introduces risk; must be bounded and validated
- 20:41 – 23:28
Long-duration autonomous missions: translating human intent into machine-executable plans
Brian dives into the hardest autonomy problems: specifying mission intent, encoding contingencies, and achieving high reliability for weeks without communication—especially underwater. He connects the challenge to “nines of reliability” and the shift from research prototypes to fieldable systems.
- •Mission intent as a UI/translation problem from human goals to machine plans
- •Underwater comms limits force higher onboard autonomy
- •Reliability requirements: weeks-long operation with minimal failure tolerance
- •Engineering effort to move from demos to operationally trustworthy capability
- 23:28 – 27:27
LLMs in defense: synthesis, decision support, and safer pathways into autonomy
Brian separates LLM value into synthesis (querying large knowledge bases) and generative/creative tasks. He highlights intelligence-analysis use cases while stressing hallucination risk, and he points to cyber and mission-planning assistance as promising areas with the right human-in-the-loop controls.
- •Synthesis: interrogate large volumes of reports and open-source intel
- •Key constraint: hallucinations are unacceptable in operational intelligence
- •Cyber as a natural fit for pattern generation/defense and automation
- •LLMs as mission-planning assistants that draft plans for human review
- 27:27 – 30:32
Anduril’s scaling reality: headcount, deployments, and the episodic nature of defense demand
Sarah asks about the business, and Brian shares Anduril’s scale and footprint, including international growth (notably Australia). He explains why defense revenue scales in jumps—10 units to 5,000 units—creating manufacturing and support challenges that require capital and careful readiness planning.
- •Company snapshot: ~1,600 people; Orange County base with broader footprint
- •Working across services: Army, Navy, Air Force, Marines, SOCOM
- •Mature deployments: border/base protection and counter-drone; newer: underwater/autonomy
- •Defense procurement is lumpy; scaling requires manufacturing + support readiness
- 30:32 – 34:43
Programs of record and winning big: early major contracts and what it takes to be a “capability” provider
Elad probes the “program of record” milestone and Brian cites large awards, including border security and a major SOCOM counter-drone effort. Brian then explains why succeeding requires far more than a great product: lobbying/engagement, navigating budgeting timelines, and building training, logistics, integrations, and security into the offering.
- •Program-of-record style scaling as the real measure of success
- •Examples discussed: ~$250M CBP border security; ~$1B SOCOM counter-UAS
- •Budgeting is slow and byzantine; success needs early Hill engagement
- •DoD buys full capability: support, training, logistics, integration, security
- 34:43 – 39:02
Next defense-tech sea changes: cheap launch, satellite constellations, and hypersonics at startup speed
Brian identifies two major shifts beyond drones/AI: dramatically cheaper access to space enabling proliferated constellations, and the emergence of credible hypersonics startups. He argues the combination of sensing, battle-space understanding, and long-range engagement reshapes deterrence and operational concepts—while also creating new counter-strategy challenges.
- •Low-cost launch (e.g., Starship-era economics) enables mega-constellations
- •DoD hasn’t fully internalized how fundamental the space cost curve shift is
- •Hypersonics becoming cheaper/faster via new entrants and iteration speed
- •Fusion of sensing + C2 + long-range fires changes force posture and deterrence
- 39:02 – 43:50
Why tech’s relationship with defense is resurging: Ukraine, China, and DoD’s gradual adaptation
Elad asks what changed culturally, and Brian points to Ukraine as a pivotal moment resetting assumptions about state conflict. He adds that China’s modernization created urgency inside the DoD to engage new vendors, and he credits the department’s ability—albeit slower than Silicon Valley—to reorganize in response to strategic pressure.
- •Ukraine invasion as a societal and strategic wake-up call
- •China’s pace and A2/AD capabilities increase U.S. urgency
- •DoD is a massive bureaucracy but has demonstrated ability to pivot over decades
- •Greater openness to startups now than 10–15 years ago due to competitive pressure
- 43:50 – 46:26
Where urgency is still lacking: sensing the Pacific, allied capabilities, and deterrence options
Sarah asks where the U.S. is underinvested, and Brian highlights the scale of the Pacific as a sensing and response problem. He emphasizes empowering allies with defensive capabilities tailored to deterrence—without necessarily enabling escalatory deep-strike options—while giving U.S. leaders credible choices that don’t rely on risking large troop deployments.
- •Pacific theater scale makes sensing, awareness, and response exceptionally hard
- •Deterrence improves when leaders have credible options short of risking troops
- •Allies need tailored defensive tools (coastal defense, cables, shipping lanes)
- •Balance: allied resilience vs. escalation and differing U.S. vs ally mission needs
- 46:26 – 48:23
How the tech industry can help: infrastructure, logistics modernization, and Washington engagement
Brian argues that beyond weapons systems, commercial tech can modernize DoD infrastructure (e.g., cloud) and logistics. He also urges Silicon Valley to engage Washington proactively—educating lawmakers and shaping workable regulation—rather than treating policy as an afterthought.
- •Commercial cloud and infrastructure adapted for classified environments
- •Operational efficiency opportunities: logistics and supply chain modernization
- •Washington ‘has a vote’: regulation will shape AI/crypto/social outcomes
- •Proactive engagement helps policymakers and reduces backlash risk
- 48:23 – 50:06
Conviction over popularity: leading through controversy in defense and border security
Elad closes by asking how Anduril’s leaders handled early criticism and unpopularity. Brian responds that controversial missions require clear conviction and ethical clarity, and that serious risks should motivate better people—not deter them—from working on the problems.
- •Defense and border security work draws controversy; clarity of mission is essential
- •Leadership requires conviction, not popularity
- •Ethics and accountability as core justifications for building these systems
- •Controversy can spotlight the seriousness of the problem being addressed
