This Week in AI | Ep. 4: Data Centers in Space, AI Excavators & Fixing AI Slop

Host: Jason Calacanis
Guests: Philip Johnston (Star Cloud), Boris Sofman (Bedrock Robotics), Spiros Xanthos (Resolve AI)
Date: March 11, 2026

Episode Overview

This episode brings together three CEO-level experts at the forefront of AI and automation to discuss ambitious trends—data centers in space, the automation of heavy construction equipment, and the reliability of generative AI in high-stakes engineering contexts. Topics span space infrastructure, workforce impacts, trust in AI, regulatory frictions, and the latest advances in autonomous systems. The roundtable is lively, direct, and focused on the evolving relationship between technological breakthroughs and their broader societal effects.

Key Discussion Points & Insights

1. Data Centers in Space: Hype, Timelines, and Economics

Philip Johnston (Star Cloud) shares the case for shifting large-scale compute off-earth.

• Timeline Reality: Space-based data centers won't dominate this decade. But post-2030, the sector is expected to grow rapidly, surging way above terrestrial compute growth rates (02:11).
• Constraints on Earth: “We’re very quickly running up against constraints on where we can build new energy projects, particularly in the US...The marginal cost of adding a new data center in space goes down over time, while on Earth it goes up.” (02:23)
• Unit Economics: Space data centers obviate land and battery costs and access uninterrupted solar (24/7). With falling launch costs (thanks to SpaceX/Starship), the break-even with Earth-based data centers is approaching—"When Starship marginal costs hit $10–$20/kg, we're well within range." (51:02)
• Scale Ambition: Star Cloud filed for 88,000 satellites; Elon/SpaceX filed for a million. "You can probably put at least 10 terawatts of capacity [in low Earth orbit]." (50:14)

Notable Quotes:

• “In 2029, space compute is going to be growing at 500% per year, whereas terrestrial compute will be going at 5% per year.” —Philip Johnston (02:11)
• “The chips are by far the most expensive part...but we're talking about being 10x less in terms of both infrastructure and energy [costs].” (53:59)

2. AI & Autonomy in Construction Equipment

Boris Sofman (Bedrock Robotics) explains the imperative (and challenges) of automating construction equipment.

• Labor Shortages Meet Surging Demand: $700B in data center construction spend this year alone; skilled operators are retiring, driving wages up and creating bottlenecks (04:26).
• Autonomy vs. Teleoperation: Remote/teleop is used today, especially in hard-to-staff locations, but true autonomy unlocks greater efficiency, 24/7 operation, and the ability to take on more projects (06:53).
• Initial Focus: Excavators—a highly utilized, hard-to-learn machine doing the literal heavy lifting in data centers, factories, and warehouses.

Memorable Segment:

• “Oftentimes these projects go on for 10 months, 15 months...there's an astronomical amount of earth to move. You're loading dump trucks 12 hours a day, 18 hours a day. And it's really hard to find this labor.” —Boris Sofman (06:53)

3. Reliability, 'AI Slop,' and Skill Degradation in AI-Assisted Engineering

Spiros Xanthos (Resolve AI) addresses the risks of over-trusting AI, especially in code and infrastructure.

• Generative AI's High 'Blast Radius': Reports of trust issues at Amazon—where AI-generated code caused unanticipated, large-scale incidents (11:56).
• Human Oversight: The bottleneck is no longer coding but maintaining, debugging, and reliably running critical systems.
• Automation Dependency: The analogy with aviation—when too much reliance on autopilot degrades foundational human skills and system understanding (14:01).

Quotes:

• “We advanced code generation quite a bit without necessarily letting the rest of the stack catch up... Now let's have AI that is on call. When something goes wrong in production, reacts a lot faster, has the full context of the whole system and helps us do that.” —Spiros Xanthos (14:01)

4. Trust, Job Displacement, and Messaging AI's Impact

• Surveys & Perceptions: Only 9% of $500M+ revenue companies plan to cut jobs due to AI; 55% expect to increase hiring (21:45). Yet, public trust is low—fears include misinformation, privacy, job loss, and lack of control.
• Industry's Role: The panel argues that the sector focuses too inward. Messaging needs to emphasize broad societal benefits—cheaper homes, better health and education, more accessible technology (23:40, 32:58).
• Historical Parallel: “Every revolution—industrial, computing, mobile—had a shock. When the dust settles, productivity rises, jobs increase, and society wins.” (24:39, Boris Sofman)

5. Regulation, Government, and Future-of-Work

• Government’s Role: Consensus: regulation should exist for safety but must avoid crippling innovation. Let disruption and benefits roll, then address dislocation with support (like UBI or targeted safety nets) (37:27).
• Philip on Policy: “If people started seeing their energy prices going through the roof, that would have been a disaster for the PR...[But] we should also consider Andrew Yang-style universal basic income financed by taxing tech companies.” (34:56)

6. Physical World Models, Simulation, and Robotic Agents

• The Next Leap: Massive investment (e.g., Yan LeCun’s $1B round) aims to model the real world for AI agents, enabling them to interact robustly in complex, chaotic environments—key for driving, robotics, household automation (39:05).
• Simulation’s Importance: In physical domains, data is scarcer and hardware heterogeneity greater than in LLM training; breakthroughs will depend on improved simulation and imitation learning (42:00).
• RL and Scientific Discovery: The group discusses autonomous agents learning through large-scale experimentation—even automating the scientific method (59:54).

7. Anthropic, Ethics, and Government Use of AI

• Military & AI Providers: Tension between Anthropic (Claude) and the US government: Should AI companies restrict use by military clients? Is branding models as a “supply chain risk” (with legal ramifications) appropriate? (61:51)
• Ethical Perspective: The guests agree private companies have the right to choose clients, but governments also have interests. Overly adversarial regulation or retaliation risks harming innovation and trust.
• Media Misinformation: Viral but unsubstantiated stories (e.g., about LLMs selecting military targets) highlight challenges in public understanding and responsible discourse (66:14).

Notable/Memorable Quotes

• On Skill Degradation
  “That's going to be the skill in some ways, Philip, isn’t it? Knowing when to stop trusting this, knowing [when not] to blindly trust the system...” —Jason Calacanis (28:55)
• On the Scale of Space Data Centers
  “We’ve just filed for a constellation of 88,000 [satellites] with the FCC. Elon’s just filed for a constellation of a million.” —Philip Johnston (03:21, 48:14)
• On the Promise and Responsibility of AI
  “I am a huge believer and a huge optimist in AI...but I think it comes with a lot of downsides...We don’t do enough to probably explain or think through what the world is going to look like.” —Spiros Xanthos (22:10)

Timestamps — Highlighted Segments

• [02:11] Data Centers in Space: Timeline and Economic Rationale
• [06:53] AI Excavators: Labor Shortages and Automation’s First Beachhead
• [14:01] “Automation Dependency” & Risks of Over-Reliance on AI/Autopilot
• [21:45] Survey: Corporate AI Adoption, Hiring vs. Job Cuts
• [32:58] 'Big Three' for PR: Communicating AI’s Impact on Health, Housing, Education
• [37:27] On Regulation & Social Support (UBI) vs. Heavy-Handed Slowing Innovation
• [39:05] Real-World AI Models and Robot Humanoids: State and Hype
• [59:54] Automating the Scientific Method: AI Agents Running Experiments
• [61:51] The Anthropic–Government Tension: Ethics, Policy, and Supplying the Military

Closing & Hiring Pitches

Each guest briefly pitches open roles at their companies:

• Star Cloud: Engineering and mission ops for “Dyson spheres and Matryoshka brains.” (67:54)
• Bedrock Robotics: ML, hardware, simulation, ops, and legal—the “giant autonomy problem” (68:17)
• Resolve AI: Infra, agentic workflows, go-to-market; “changing the way software works” (68:51)

Final Thoughts

The episode paints a forward-looking, pragmatic but optimistic picture: AI’s largest impacts are yet to come, and the societal shocks—from job shifts to government pushback—are real but manageable, if the industry tells its story better and continues to invest in building robust, safe, and valuable technology.

[Listen and subscribe for more expert roundtables at thisweekinai.ai]