TBPN Podcast Summary — Oct 2, 2025

Episode Title: AI’s Power Problem, Apple Goes Meta on AI Glasses | Pat Gelsinger, Josh Isner, Sheel Mohnot, Santiago Nestares, Austin Federa
Hosts: John Coogan & Jordi Hays
Guests: Pat Gelsinger, Josh Isner, Sheel Mohnot, Santiago Nestares, Austin Federa
Length: ~3 hrs
Theme: The real-world costs and crossroads of AI growth—particularly energy and PR, major funding/launches, and the collision of frontier technology and culture.

Episode Overview

This episode delves deep into the mounting tensions between explosive AI/tech progress and its societal and infrastructural costs—especially energy. The hosts unpack how culture and economics are rapidly shifting under AI’s weight, how companies are responding (or not), and where the innovation/mania is heading. Special guests inject expert insight on topics from AI benchmarking for human flourishing, to how tech is transforming public safety, to the future of internet infrastructure.

Key Discussion Points & Insights

1. AI’s Escalating Energy Crisis ("The Slop Trough")

Timestamps: 00:16–16:00, 53:00–60:00

• AI and Data Centers: The hosts lampoon the explosion of "AI slop"—cheap, high-noise content generated at huge energy expense—calling out how AI compute is dragging up electricity demand and consumer bills.
  • “Everyone was worried about water. Turns out electricity may be a bigger deal.” —Jordy, 01:30
  • Average U.S. electric bills up 25% since 2020, tracked closely with AI/digital infra buildup.
• Public Backlash Brewing: Users resentful over footing the bill for “crap” AI videos—one viral post: “Pretty cool that my electric bill has gone up 25% so other people can make this crap.” (02:50)
• Infrastructure Catch-up: Despite migration to LEDs and more efficiency elsewhere, new AI/data center demand outpaces grid improvements. New nuclear and solar builds are in focus, with startups (like Fermi) and hyperscalers attempting to commission both new and old energy sources.
• Regional Impact: In Virginia, data centers jumped from <5% to 40% of statewide electricity consumption in 15 yrs (13:00).
  • “Right now…single family homeowners and businesses are basically subsidizing the cost for these data centers.” —Jordy, 15:15
• Tech vs Public Opinion: If energy costs keep rising without clear consumer benefit, public ire at AI (already on the rise) could spiral. Reference: comparison to NFTs—cool for insiders, deeply irritating for most.

2. Cultural Backlash and PR Trap for AI

Timestamps: 16:00–21:00, 61:00–66:00

• AI PR Dilemma: AI’s usefulness (e.g., summarizing meetings) is evident, but growing sentiment is that it mostly “helps people do unnecessary work faster.” (17:22)
• Negative Narrative: Media seizes on stories about AI-induced “psychosis,” i.e., users getting lost, harmed, or isolated in their interactions.
• Hopeful Paths: Like with crypto’s transition from proof-of-work to proof-of-stake, hosts posit that technical innovation might resolve energy-waste narratives for AI.

3. AI Architecture Debates: The "Bitter Lesson" and RL vs Pretraining

Timestamps: 22:05–38:00

• Karpathy & Sutton: Andrej Karpathy’s summary of Richard Sutton’s “Bitter Lesson”—AI success comes from scaling compute/data, not hand-crafted complexity, but LLMs may not embody this fully as they're dependent on finite human data.
  • “We do not in fact have an actual single clean... algorithm you could unleash upon the world and see it learn automatically from experience alone. Does such an algorithm even exist?” —John (Karpathy paraphrase) at 27:02
• Reinforcement Learning (RL) as the True Path? Sutton and the panel discuss whether pure RL, not imitation learning, is the real bridge to AGI—mirrored in animal learning, e.g., “If we understood a squirrel, we’d be almost done.” (25:52)

4. The Infinite Jest/“Slop Trough” Hypothesis in AI Video

Timestamps: 23:00–33:00

• AI-generated video (e.g. OpenAI Sora): Cost to create is still high, but OpenAI and others are subsidizing it to drive adoption, potentially eating hundreds of millions in cost for user growth.
• Sam Altman "Aura Farming": A term for building fame/reputation (and resilience to criticism or brand weirdness) by leaning into deepfake content, for example, Altman himself starring in hundreds of viral Sora videos.
  • "It's way less strange to watch a feed full of memes of yourself than I thought it would be." —John quoting Sam Altman, 37:00

5. Job Market Realities: Did ChatGPT Kill White Collar Jobs?

Timestamps: 43:42–46:11

• Chart Crimes & Causality: Hosts debunk the “ChatGPT triggered jobpocalypse” narrative. Decline in job postings tracked more closely to macroeconomic cycles (e.g., end of ZIRP, FTX crash) than solely on LLM deployment.
  • “The AI narrative from a job loss standpoint, it’s also been a lot more narrative-driven…” —Jordy, 45:11

6. Apple & Meta AI Glasses Arms Race

Timestamps: 56:00–59:00

• Apple Halts Cheaper Vision Pro for Glasses Push: Shifting resources to AI-centric smart glasses to keep up with Meta. Prediction: Apple will keep design in-house (vs. fashion partnerships), seeking AirPods-like ubiquity on faces.

Notable Guest Segments

Pat Gelsinger (Ex-Intel CEO, Deep Tech/“FaithTech” Investor at Playground/Glue)

Timestamps: 90:14–135:39

• AI Benchmarks for Human Flourishing: Gelsinger leads an effort to benchmark AI models along dimensions tied to flourishing—happiness, relationships, character, finances, faith, health—rather than just speed or cost.
  • “The absence of bad does not demonstrate the presence of good.” —Pat, 92:19
• Reality Check on Energy & CapEx Forecasts: AI growth is truly energy-limited; every “gigawatt” data center announced should conjure a nuclear reactor in your mind! The U.S. is drastically underbuilt for this (118:45+).
• Faith & AI, "AI Gods": Gelsinger reflects on rising faith/religion discussion in Silicon Valley, warns of both cultish uses of AI and overblown “apocalypse” narratives, advocates for tech being steered toward positive, relational human goals—not isolation or idol worship.
  • “Technology is neutral. Are we going to bend that arc toward good?” —Pat, 96:24

Josh Isner (President, Axon)

Timestamps: 61:37–79:26

• Transformation from Hardware to SaaS: Axon (formerly Taser) now manages 40x Netflix’s media volume in digital evidence for police. AI is shaving police report-writing time from 50% to 10%.
• Prepared Acquisition: Acquired for $630M—real-time 911 automation aims to dispatch drones and officers instantly via AI.
• Openness and M&A: Axon wants to be “Switzerland” for public safety APIs, integrating across platforms.

Sheel Mohnot (Better Tomorrow Ventures, Fintech VC)

Timestamps: 149:43–159:39

• $140M New Fund: Fintech is back in favor after a rough patch; surviving was key for portfolio companies.
• Stablecoins and Stripe: Startups face tough competition as giants (like Stripe) and public cos (Circle, Tether) move in. “Believe in the category, but not many new startup opportunities.”

Santi Nestares (Dual Entry, AI ERP)

Timestamps: 160:10–167:24

• $90M Series A Announced: AI ERP company with agentic “next day” migration, enabling companies (even public ones) to switch fast—distinguishing by drastically reducing pain/risk of ERP migration.

Austin Federa (00, Blockchain Infra)

Timestamps: 167:29–178:52

• $6B Token Launch: Double Zero protocol builds a blockchain-based, decentralized fiber internet backbone for high-performance networking, with major crypto and trading firms as contributors.
• Crypto Business Model: Network aligns operators with tokens; contributors drawn by the chance to compete with cloud giants’ lock-in. Growing interest as regulatory signals shift positive.

Selected Memorable Quotes

• "JP Morgan estimated that 70% of last year's increased electricity cost was the result of data center demand." —John, 04:47
• "If demand is certainly there today... get ready for expensive electricity. But if it's a sigmoid curve... America probably overbuilds and then prices fall." —John, 11:00
• "I don't think people understand the logistical hurdles in place for data centers now... Power companies are overwhelmed and tired." —Jordy, 145:09
• "AI is really good at helping people do unnecessary work faster." —Unnamed entrepreneur, 17:22
• “We do not in fact have an actual single clean... algorithm you could unleash upon the world and see it learn automatically from experience alone. Does such an algorithm even exist?” —John quoting Karpathy, 27:02
• “If we understood a squirrel, we’d be almost done.” —John (Karpathy/Sutton paraphrase), 25:52
• “The first job of the leader is to represent reality where it is.” —Pat Gelsinger, 127:11

Key Guest Segment Timestamps

| Guest | Topic Highlights | Start-End | |--------------------|----------------------------------|---------------| | Pat Gelsinger | AI, Human Flourishing, Nuclear | 90:14–135:39 | | Josh Isner | Axon AI, SaaS, Drones | 61:37–79:26 | | Sheel Mohnot | Fintech trends, Stablecoins | 149:43–159:39 | | Santi Nestares | Dual Entry AI ERP, $90M raise | 160:10–167:24 | | Austin Federa | 00 Blockchain, Token Launch | 167:29–178:52 |

Closing Thoughts

The episode paints a vivid, sometimes satirical but insightful picture of the next phase of AI—where practical challenges (energy, infra, costs), big-picture cultural impact (public opinion, job market), and philosophical questions (alignment, meaning, flourishing) all collide. The stakes have moved from code and compute to questions about society itself, while capital and ambition remain as abundant as ever.

For full context, listen to the marked segments for a deep dive into the energy crisis, AI’s PR struggles, and the nuanced behind-the-scenes of tech’s biggest new launches.