TBPN Podcast Summary

Episode: Inside ChatGPT’s Uses, NVIDIA Pours $100B into OpenAI | Kimbal Musk & Shervin Pishevar, John Shahidi, Laura Deming, Steven Glinert, Austin Petersmith, Ethan Barajas & Jamie Palmer
Hosts: John Coogan & Jordi Hays
Date: September 22, 2025

Episode Overview

This jam-packed episode explores the evolving landscape of artificial intelligence (AI), the business of LLMs (Large Language Models), and the ecosystem around OpenAI’s latest developments. Key topics include AI personalization, OpenAI’s monetization and compute challenges, the growing intersection of AI and consumer products, and a flurry of startup founder interviews covering everything from drone art to on-device LLMs and the business of creator-led brands. With a blend of news, tech analysis, and founder chats, the episode is a whirlwind tour of the current tech zeitgeist.

1. The Future of ChatGPT: Personalization, Monetization, and Competition

[00:00–29:59]

Main Discussion Points

AI Personalization & User Expectations

• Sam Altman teases new compute-intensive offerings for pro subscribers; hints at new features/products with extra fees ([01:25]).
• Ongoing debate: Can LLMs deliver true personalization, or do they all just “remember a few facts” and respond in similar, generic tones?
• Referencing a Matthew McConaughey podcast appearance, the hosts discuss users’ desire for “private LLMs” trained on personal data:

  “I want the answers based on what I’ve uploaded it with, not from the outside world.” – John Shahidi quoting McConaughey ([03:39])

• Current ChatGPT “memory” function deemed “overhyped” and not a differentiator.

  “I haven’t experienced feeling like I am so locked into this LLM because it knows everything about me.” – Jordi ([04:48])

“Voice” and “Personality” in LLMs

• Brief detour into model personalities: Reddit communities deeply attached to specific LLM personalities (e.g., ChatGPT 4.0, Claude 3.5 Sonnet) ([05:33]).
• Prediction: The sticky future is where each user gets a unique, evolving AI “partner” ([09:42]).

AI as Tool vs. Friend

• Divide between using AI for knowledge retrieval (personality doesn’t matter) vs. as a “friend” (personality and memory crucial) ([11:46]).
• OpenAI’s new “Personalization” tab: personality modes (“cheerful and adaptive,” “cynic,” “robot,” etc.), but customized voices are still surface-level ([13:31]).

Product Speculation & Monetization

• Anticipation for Sora 2 (OpenAI’s video model), Agent 2 (autonomous action-taking), and memory upgrades – likely as paid add-ons due to GPU costs ([15:18], [25:34]).

• Candid feedback on the cost and scaling challenges of deploying heavy compute AI models:

  “Maybe the market clearing price is much lower, but it should be above the value of running those GPUs, in my opinion.” – John Shahidi ([15:18])

• Discussion of plug-ins, agents, and deep research – but consensus that personality and memory are not core product lock-ins yet ([10:43], [19:39]).

Notable Quotes & Timestamps

• "What actually is sticky and that feels like it all feeds into lower churn. So I think that's sort of the goal." – John Shahidi ([11:26])
• “People just realize, if I want to post under this identity, I'm still better off going to X and sharing there.” – Jordi on the difficulty of new social platforms ([79:41])
• “I still believe that if you're in the first camp... personality just doesn't matter that much.” – Jordi ([12:00])

2. Founder & Creator Economy: Product, Brand, and Distribution

Notable Guest: John Shahidi, Shots Podcast Network
[58:18–105:40]

Highlights

Building Successful Creator Brands

• John Shahidi describes strategy behind scaling Happy Dad Seltzer, leveraging retail-first approach and targeting under-served demographics (e.g., men in hard seltzers).
• The importance of high product quality for creator brands:

  “A fan of any creator... will try anything once. Now the question is, will they buy it again?... It wasn't good.” – Shahidi on why Travis Scott's seltzer failed ([68:39])

• Retail vs. e-commerce for beverage brands – direct online selling is expensive and not always practical ([63:16]).
• Shahidi explains how relationships with retailers (Walmart, Kroger, 7-Eleven) are key, and offers advice from Dana White to “treat retailers like YouTube, Meta, etc.” ([63:37]).

Scale vs. Niche for Creator-led Startups

• You don’t need to be Tom Cruise-level famous — niche creators can build substantial brands, as long as the product is strong ([66:40]).
• Collaborating with large CPGs often fails due to a lack of “hustle” and entrepreneurial energy from the big partner ([72:22]).

Influencer Businesses, Investing, & Content Platform Strategy

• On influencer investing: surround yourself with sophisticated institutional investors – don’t just take flyers on “your cousin’s business” ([101:58]).
• Platform advice: “YouTube first” for discoverability/long tail, X for live, Spotify and Insta still playing catch-up ([74:53]).
• Advice to creators: focus on owning a great product, not just content ([88:23]).

Memorable Moments

• John retelling advice about not overhyping before nationwide distribution: “If I had to do it again, I wouldn’t have pushed it so hard, because what's the point of Chris promoting this... if it's not available everywhere?” ([92:53])
• Entertaining banter about securing usernames across every platform:

  "You got to log in every day, 179 days... There’s an army of like high school kids trying to snipe the names." – John & Jordi ([85:33])

3. AI, Capex, & The Circular Economy: Investment, Bubbles, & Tech Cycles

[24:07–55:16, 116:56–120:22]

Key Discussion Points

NVIDIA, OpenAI, Oracle’s $100B Flywheel

• Huge circular flow of capital: NVIDIA invests $100B into OpenAI, who uses some of it to pay for Nvidia GPUs, etc. ([50:36])
• Hosts mock the “circular” nature of this, compare it to older ad-tech cycles.
• Ultimately, the only thing that matters is usage and DAUs (Daily Active Users): “if ARR starts to slow, then all bets are off” ([52:50]).

Tech Bubbles and Value Creation

• Quoting Carlotta Perez on financial bubbles: we may be mid-bubble, but the life cycle is like PCs/container ships, not just internet 2000. ([54:05])
• Debate over the value of fundraising announcements vs. product launches in B2B SaaS:

  “When you’re in B2B, you need to take advantage of the business story. And the business story is the fundraising story.” – John Shahidi ([42:00])

• Conversation around changing value of niche and viral in enterprise SaaS vs. consumer apps ([42:00–44:16]).

4. New AI-Driven and Frontier Tech Startup Announcements

a) Laura Deming, Until Last Year

“Making hibernation pods reality by solving reversible organ cryopreservation”
[120:27–147:02]

• Announced $58M round led by Founders Fund & Lux, aiming to demonstrate fully-reversible cryopreservation for human organs.
• Practicality: First products will drastically extend organ shelf-life for transplants (currently only ~12–24 hrs), disrupting expensive, logistics-heavy systems like Blade (medical airlifts).
• Long-term vision: hibernation pods for both medical/clinical (e.g., waiting out diseases) and far-future uses like interstellar travel.
• Insightful technical breakdown:

  “Our enemy in cryopreservation is ice... what we want to do is make glass.” – Laura ([124:32])

• On iteration speed: “Unlike a lot of biology, you can test whole human-sized organs a lot of your protocols weekly.” ([131:59])
• Big ideas: Cryo as a time machine, ethical/sci-fi impacts, U.S. vs. global efforts ([134:49], [139:24]).

b) Kimbal Musk & Shervin Pishevar, Nova Sky Stories

Breathtaking live drone shows as art, at city scale
[151:09–167:47]

• Raised $50M led by Pishevar; just produced a 300,000-person Vatican show simulcast on Disney+.
• Expansion: now running 5 global shows simultaneously, bottlenecked only by drone supply.
• Product is a full-stack of technology, IP, government/venue relationships, and creative:

  “It’s basically building like a Pixar and a Disney all in one inside of a technology company.” – Shervin Pishevar ([165:34])

• Regulatory advantage: Flight permissions, logistics, venue relationships.
• Emotional resonance and public demand for “live, communal” experiences.

c) Stephen Glinert, Sphera

AI-designed analog and mixed-signal chips (for defense/communications)
[169:14–176:53]

• Announced $12M funding (Acme/Future Ventures, $20M total).
• Builds analog front-ends for systems like Anduril, using fully-AI-designed chips old guard still often designs by hand.
• Go-to-market: Defense sector, expanding to data-center networking and comms.

d) Ethan Barajas & Jamie Palmer, Icarus Robotics

Space labor: dexterous robots to replace astronaut time
[177:35–182:59]

• Raised $6.1M seed.
• Building remotely-operated, dexterous, free-flying robots for logistics and maintenance inside space stations, freeing astronauts for higher-leverage science.
• Details on testing: air-bearings, NASA's neutral buoyancy pool, and upcoming ISS trials.

e) Austin Petersmith, Secretary (Howie)

AI + human-in-the-loop executive assistant for scheduling
[186:30–198:44]

• Raised $6M seed.
• Practical hybrid model: AI models for scheduling logic, with human validation for reliability (“like Waymo’s driver in an icy road”).
• Price: $35/mo for base, $145/mo for “pro” (white label, advanced customization).
• Anticipates gradual shift to more AI automation, but sees permanent need for “humans holding the steering wheel” for hard cases.

5. LLMs, AI Voice, Consumer & Tech Culture Miscellany

[55:16–120:27], [147:10+]

• Discussed Apple’s on-device LLM (A19 Pro chip, Apple "Foundation model"), and why local models might not matter for most consumers ([45:22]).
• Grok (Elon's X AI) voice, 11Labs as “the Midjourney of voice,” and the legal/cultural complexities of branded AI voices ([55:13]).
• Discussion of OpenAI, Anthropic, Google Gemini, and Claude – and users’ strange attachment to “favorite models” as characters ([34:03]).
• “Sharing” as the future social feature of LLMs – curating research, context, and collaborative prompting ([77:23]).
• Meta’s new “Gaussian splats” for VR/AR world scans; real emotional impact of digital spatial environments ([48:02]).

6. Fun, Observations, and Culture

• Oura Ring hits 3M sales, $1.5B revenue, $10B valuation — from Kickstarter to mainstream ([111:37]).
• Trends in live entertainment, the rise of immersive, drone-powered shows over cities and mass experiences ([151:09–167:47]).
• Humorous, rapid-fire takes:
  • The endless struggles with home printers
  • Fake/AI-generated viral images spreading online
  • The “Green Line” test of Trump/Elon handshake photos
  • The trading of platform usernames as digital property
• Banter about anchorless consumer startups, recurring launches, and soundboard gongs as product milestones ([38:02]; [148:46]).

Notable Quotes By Timestamp

• “Crush your backlog with your personal AI engineering team.” – John Shahidi ([22:24])
• “If you want to develop some software, if you want to design something, head over to figma.com. Think bigger, build faster. Figma helps design and development teams build great products.” ([35:16])
• “You should actually definitely run. Ajni Mihdha over at Andreessen had a different take. He said, true for the median startup, but it’s dangerously wrong for frontier technology companies. Capital is a weapon.” ([42:00])
• “I feel like there’ll be some sort of power law of, like, the best environments and there’ll be a few.” – John Shahidi on virtual world scans ([49:46])
• “If the product is great, it'll sell itself.” – John Shahidi ([66:49])
• “Every entrepreneur… has been through this before. It’s a comeback story, baby.” – [149:24]
• “This episode should be called: everything is circular, everything is compute-bound, and everyone is launching, always.” – Jordi (paraphrased theme in closing).

Timestamps for Key Segments

• OpenAI Personalization & Monetization: [00:00–29:59]
• Anthropic vs. OpenAI Marketing: [28:51–34:56]
• Creator Brands Deep Dive (John Shahidi): [58:18–105:40]
• Nova Sky Stories (Kimbal Musk/Shervin Pishevar): [151:09–167:47]
• Laura Deming Cryopreservation/Aging: [120:27–147:02]
• Analog Chips/Defense (Stephen Glinert/Sphera): [169:14–176:53]
• Space Robotics (Icarus Robotics): [177:35–182:59]
• Secretary/Howie (Austin Petersmith): [186:30–198:44]

Conclusion

This episode offers a vibrant snapshot of the current moment in tech: the limits and promise of big LLMs, the economics of AI compute, the perpetual hype cycle of launches, and the messy intersection of tech with real-world industries. Through sharp banter and an impressive guest lineup, the show captures the optimism, anxiety, and relentless pace of technology in late 2025.