This Week in AI E3 – Detailed Episode Summary

Overview: Main Theme and Purpose

Episode Title: AI in Warfare, OpenClaw & The Stargate Mega-Campus
Date: March 4, 2026
Host: Jason Calacanis
Guests:

• Chase Lockemiller (Co-founder and CEO, Crusoe)
• Naveen Rao (CEO, Unconventional AI)
• Anastasios Angelopoulos (CEO, Arena)

A high-caliber, experts-only roundtable on pivotal developments in AI, this episode dives into the explosive impact of agentic AI platforms like OpenClaw, the rise of modular and mega-scale AI infrastructure (Stargate campus), and urgent conversations around AI’s role in warfare and the future of labor. The group debates open v. closed models, how AI labor may reshape global economics, challenges around AGI, and ethical questions as AI becomes integral to society and national security.

Key Discussion Points and Insights

Introductions and Company Backgrounds

• [02:10] Chase Lockemiller on Crusoe
  Crusoe builds vertically integrated AI infrastructure, handling both data center hardware and cloud software, serving major clients like Oracle and OpenAI. Their mega project, the 1.2 GW 'Stargate' campus in Abilene, TX, is the world's largest GPU cluster in operation or development.

• [03:46] Naveen Rao on Unconventional AI
  His company is rethinking computer architecture entirely, aiming for 1000x power efficiency by mimicking biological computation. The long-term aim: achieve “joules per token” at unprecedented levels, potentially unlocking new substrates for AI computation.

• [06:32] Anastasios Angelopoulos on Arena
  Arena started as a Berkeley research project for neutral LLM benchmarking using human preference voting (Arena platform), now central to industry model evaluation with tens of millions of monthly users ([10:51]).

The Rise and Impact of OpenClaw & Open Source Agentic AI

• OpenClaw’s Explosive Adoption and Token Costs

  • Jason describes how rapid application of OpenClaw internally increased efficiency but at significant cost, sparking the need for local or open source agentic models ([12:09], [13:48]).
  • “We were trending, I think with 15 people getting on it towards $500 bucks a day each, $7,000 a day, $2 million in tokens… this is like in month one of the project…” – Jason [13:12]

• Open vs Closed Source Model Debates

  • Chase: “I think there's… room for... multiple solutions in the future. ...open source and a lot of closed source.” [13:48]
  • The ability to distill large models into performant small models that can run locally (e.g., on an iPhone in airplane mode) is a key disruptive trend [13:48], [24:43].
  • Agentic platforms are a step change: “...actually producing digital labor, not just digital intelligence.” – Chase [43:38]

• The Benchmark/Goalpost Problem in AGI

  • Definitions slide constantly as models get smarter; solving grand unsolved problems like the Millennium Problems is suggested as a real test of AGI ([15:51], [17:15]).
  • “When we create benchmarks for AGI... We need to make sure that the data is constantly fresh to avoid overfitting.” – Anastasios [17:15]

Economic & Workforce Disruption

• “A large fraction of human labor is going to be commoditized by, by AI, and that's going to have very wide-reaching economic impacts and probably require a new economic system to deal with it.” – Anastasios [18:11]

• Jevons Paradox and Accelerated Productivity

  • Token usage and productivity are increasing, driving demand for more compute and changing employment paradigms ([37:20], [39:51]).
  • “I want to see the same amount of people doing more work. …my customer base is going to expect more.” – Naveen [39:51]

• On the Future of Work

  • Founders are mostly optimistic but acknowledge the destabilization of “repeatable tasks” as foundational to current labor [41:19].
  • “We're going to enter a crisis of labor where… labor does not equal value anymore…” – Anastasios [41:19]
  • The future will favor creativity, curiosity, self-agency, and radical self-reliance rather than rote or repetitive work (consensus among all hosts/guests at [52:46–54:44]).

Next-Generation Infrastructure: Mega and Modular Data Centers

• Chase details how Crusoe is building two paradigms:
  • Mega-campuses for training breakthrough foundation models (Abilene 'Stargate', Wyoming, others).
  • Modular, mobile data centers: “Crusoe Spark is a fully self-contained modular AI factory … can be rapidly deployed anywhere … powered by solar and batteries … swarms of these things” [26:02–27:32].
  • This “bimodal” future means both enormous and ubiquitous micro-data centers will coexist, supporting everything from massive training to distributed inference/robotics [29:38–31:17].
• There are significant utility/grid and energy transmission constraints, which ever-smaller, power-efficient models and centers can help address [30:48].

The Shrinking Model and On-Device AI

• Models are getting both bigger and smaller:
  • Large, complex models drive advances; meanwhile, compression/distillation means performant small models run on consumer devices (Quinn 3.5, etc.) [24:43], [34:21].
  • “Within the year we could have… models that are a billion parameters or hundreds of millions… same performance as… hundreds of billions…” – Anastasios [34:27]
  • On biology vs AI approaches: humans/animals are “orders of magnitude” more power efficient at intelligence tasks than today’s silicon. “We’ve brute-forced our way… now we’re understanding, oh actually I didn’t need all that…” – Naveen [35:15]

Should Everyone Learn to Code?

• Nat. Language as the New Programming Language:
  • “Karpathy said, the next programming language is natural language.” – Anastasios [62:21]
  • Lower-level coding (React, Ruby, etc.) is being abstracted away; deep systems understanding (machine architectures, transfers) will be valuable for a minority—but everyone will be a developer, in a sense, via AI agents [59:12–63:09].

AI in Warfare and Ethical Dilemmas

• Should AI Builders Enable Military Use?
  • Chase: “Any global conflict is going to be… fought with AI… it's an inevitability.” [65:55]
  • Naveen: Moral decisions are only vital at the application layer (e.g., targeting software vs. general chip design); ultimate responsibility lies higher up ([68:11]).
  • Anastasios: Corporations are part of nations and, ultimately, the nation (via law, e.g., Defense Production Act) gets to decide—individuals and companies can and should object, but the collective will decides ([71:27]).
  • The group acknowledges the new complexity of private, global corporations developing foundational technology with outsized state-level power and responsibility.

Notable Quotes & Memorable Moments

• On agentic AI’s leverage:
  “In the future, my kids are going to have access to the workforce equivalent of millions of people worth of labor, just like at their fingertips, like on their phone.” – Chase Lockemiller [00:00]

• On redefining coding:
  “The next programming language is natural language.” – Anastasios Angelopoulos [62:21]

• On model progress:
  “A model 1/10th the size getting similar performance only a few months later. Compute is getting exponentially better... The cost is only going down and the performance is only going up.” – Chase Lockemiller [36:19]

• On shrinking models:
  “Within the year we could have… models that are a billion parameters or hundreds of millions… same performance as… hundreds of billions…” – Anastasios Angelopoulos [34:27]

• On foundational change in value:
  “Labor does not equal value anymore… moving forward… a lot of what we now call labor will be done in data centers, or on Naveen’s chips…” – Anastasios Angelopoulos [41:19]

• On AI and warfare’s inevitability:
  “Any global conflict is going to be at the foundation of it… fought with information, fought with AI and AI is going to be embedded in weapon systems… it's an inevitability.” – Chase Lockemiller [65:55]

• On advice for the next generation:
  “The things I’m optimizing for [in my kids] are… a high sense of curiosity… and agency. … If you have incredibly high agency and curiosity to solve a problem that … speaks to some other human, you’re going to be able to just will that into existence.” – Chase Lockemiller [50:28, 00:00]

Timestamps for Important Segments

• [02:10] – Crusoe’s Stargate campus, mega AI infrastructure
• [03:46] – Rethinking computer architecture for AI (Unconventional AI)
• [06:32] – Arena and the “battleground” for LLM benchmarking
• [13:48] – The open vs. closed source debate, rise of small, on-device models
• [15:51], [17:15] – Defining AGI and benchmarks for intelligence
• [24:43] – System 1 vs. System 2, shrinking models, on-device inference
• [26:02]–[27:32] – Modular “data centers in a box”; rapid deployment
• [34:27]–[35:15] – Model compression, performance leaps, biological vs. engineered intelligence
• [41:19]–[43:38] – The future of labor and value, mass automation tension
• [50:28]–[54:44] – Guidance for workers and students: embrace agency and leverage new tech
• [59:12]–[63:09] – Coding, abstraction, and the future of software development
• [65:55]–[75:46] – AI and warfare, corporate/nation responsibilities, moral debate

Advice for Listeners & Takeaways

• Embrace Curiosity, Agency, and Continuous Learning:
  The next era belongs to those who rapidly adopt and adapt. Learn the tools, question everything, and stay at the “bleeding edge” ([00:00],[54:10]).

• Understand Systems, Not Just Tools:
  Coding is moving up abstraction layers. The few who master the underlying systems and architecture will lead innovation, while many will “code” in higher-level, more natural languages ([59:12]).

• Be Proactive About AI’s Impact on Your Field:
  Whether you’re a driver, paralegal, or engineer – learn how AI will transform your role, and reposition yourself to ride the coming wave ([54:10]–[55:38]).

• Agentic Platforms are Here—Participate or Fall Behind:
  “...the three or four who took to Openclaw first, in 30 days ...became literally five times more valuable than the people who didn’t.” – Jason Calacanis [52:46]

• Society Must Grapple with New Types of Power (and Risks):
  As AI becomes central to warfare, government, and economics, new responsibilities and debates emerge for technologists, companies, and citizens ([65:55], [71:27]).

Closing Thoughts

An episode rich in hard truths and optimism: AI is simultaneously exploding efficiency, destabilizing labor, and moving shockingly fast in capability. The panel sees disruption as inevitable, with massive opportunity for those who wield agency, curiosity, and agility. Ethical debates and societal restructuring loom large as AI embeds itself in both everyday productivity and global power struggles.