Podcast Summary: Intelligent Machines 847 – "Caked Up Football Man"

Podcast: All TWiT.tv Shows (Audio)
Episode: 847
Date: November 27, 2025
Host: Jeff Jarvis (with guest Paris Martineau, Anthony Nielsen)
Special Guest: Emad Mostaque (Founder of Stable Diffusion, Intelligent Internet II.inc)

Main Theme & Purpose (00:00 – 03:20)

This episode centers on the rapid evolution of artificial intelligence, the implications of generative models like Stable Diffusion, and how small, targeted AI systems — especially in healthcare — could shape economic, societal, and cultural futures. Featured guest Emad Mostaque (founder of Stable Diffusion and II.inc) discusses his new book, "The Last Economy", and his vision for a coming "intelligence inversion": a paradigm shift that positions AI as the most economically valuable "actor" on earth, fundamentally altering the role of human cognition in work.

Key Discussion Points & Insights

1. The “Intelligence Inversion” & Economic Shifts

[04:58 – 07:17]

Past to present: Human economies moved from agrarian (land/farming), to industrial (factories), to services, and now toward an “intelligence inversion” — where AI, not humans, will dominate cognitive labor.
Most jobs are at risk: As AI models evolve from single-task to agentic, multi-step processes, human cognitive work could become the least valuable, potentially having "negative value" in team contexts.
Mostaque:

“As we move from these models that can one-shot things to longer-range agents... our cognitive labor is probably going to have negative value, because we'll be the dumbest people on any team.” (05:09)

2. Diffusion Models, Call-Center Extinction, & the Digital Double

[07:31 – 09:54]

First jobs to automate: Remote, routine, and call center roles due to multimodal, cost-effective, “digital double” AIs.
AIs learn from vast user data (voice, keyboard, video) and build “digital twins” — virtual worker replicas indistinguishable from real people.
Mostaque:

“The first thing to go is remote jobs. Anything that can be done fully remotely. Those are most at danger.” (08:32)

3. The Plummeting Cost of Intelligence

[10:14 – 12:47]

GPT-3: ~$600 per million tokens; Grok models now ~$0.50 per million — orders of magnitude cheaper.
Ten million spoken words per year? ~$5. Fifty bucks for everything you think, annually.
Models learn context better than humans realize; context storage, pre-training strategies, and “error rates” have drastically improved.
Mostaque:

“The thing that’s really scary though is people don’t understand how cheap a unit of intelligence has become.” (10:15)

4. Shift from Prompt-Response to Agentic, Multi-Modal AI

[12:20 – 14:36]

New models (Gemini 3, GPT-5, Grok) handle million-token context windows — entire codebases, books, or manuals at once.
Agentic, proactive AIs can verify, re-prompt, and handle complex, multi-hour workflows.
Economic impact: AI is poised to hit creative, knowledge-based, and remote jobs soon.

5. Cultural Identity, Job Meaning, and Societal Fallout

[15:31 – 17:13]

Technology’s “promise” was to free humans for higher pursuits, but identity based solely on work/material achievement is increasingly precarious.
Mostaque:

“If your identity is tied up in your work ... then you're in trouble. If you have a strong community, strong faith, strong family, you're less likely to be affected by all of this.” (15:31)

6. The Future of Money & Economics Amid AI-Dominated Firms

[15:57 – 19:17]

Human-run firms can’t compete with 100% AI ones (like “MacroHard”, Elon Musk’s rumored venture).
“Virtual workers” (AI agents) could soon replace humans at the scale of trillions of dollars in labor spending.
UBI isn’t economically sustainable as currently envisioned.

7. Scale vs. Research — The Frontier Debate in AI

[17:13 – 18:51]

Adam Brown (DeepMind) argues for scale, Yann LeCun (ex-Meta) for new paradigms beyond LLMs.
Mostaque: “Satisficing” models (good enough, fast enough) for ‘machine jobs’ are already here; breakthroughs needed only for “AI god.”
Mostaque:

“We built a 7-billion parameter medical model … performs at ChatGPT levels on healthcare — on a Raspberry Pi. That’s tiny.” (17:56)

8. "Little AIs Everywhere" and the Democratization of Intelligence

[19:47 – 21:55]

Cites Kevin Kelly’s "little AIs" future: small, edge-based, domain-specific intelligence models (vs. monolithic Big Tech LLMs).
“Jarvis future”: Everyone will soon have a personal AI "buddy" — small, running locally, highly persuasive, and sovereign.
Local, sovereign AI architectures will differentiate as compute costs soar.

9. AI Creativity, Diffusion, and Generative Media

[22:46 – 25:45]

Diffusion models will produce “every pixel” — revolutionizing media, Hollywood, and economic planning.
LLMs become ultra-efficient, moving focus to diffusion for video, image, code, and self-driving applications.
Creativity is contextual — AIs can fuse concepts ("Barbenheimer"), but true scientific breakthroughs require new verifier mechanisms.

10. Societal and Labor Resistance, Regulatory Ideas, and “Sovereign AI”

[26:45 – 28:46]

Next year (2026) may bring public backlash as visible job loss arrives.
Positive message: AIs can democratize diagnosis, distribute knowledge, and support societal needs (e.g., universal free second medical opinions).
Sovereign, transparent, open-source AI is key for aligning with communities; protection against “cognitive colonialism” and corporate ad manipulation.
Mostaque:

“Your only defense against cognitive colonialism is having your own AI aligned and controlled by you.” (32:02)

11. Practical Advice for Young People

[28:57 – 30:13]

Don’t go to college just to “learn skills” — best option: learn to wield AIs daily, build, experiment, and submit AI-built work as resume.
Mostaque:

“If you use Replit, Suno, all these things for an hour a day, you’re way ahead — in a job or outside of a job.” (29:06)

12. How II.inc Is Building Civic, Transparent AI

[32:23 – 34:22]

II.inc aims to create full-stack, open models for health, education, government — not to build “AI gods,” but practical, locally-adaptable agents.
Sovereign “nodes” for each U.S. state/country (“II Utah”, “II Delaware”, etc.), letting communities shape their own AI ethics and policy.

13. Risks, Alignment, and 50/50 Doom

[34:40 – 38:02]

Mostaque estimates “50/50” odds on AI catastrophe vs. positive outcomes — stresses need for resilience, transparency, and robust system architectures to avoid disaster (e.g., global “Stuxnet” for cognitive systems).
Open, decentralized, locally-customizable AIs are means towards fail-safety and public benefit.

14. Memorable & Notable Quotes

On the intelligence inversion:

"We're no longer the most economically valuable things on the planet ... cognitive labor is probably going to have negative value because we'll be the dumbest people on any team."
— Emad Mostaque (05:09)
On creativity/context:

"Creativity is usually within a context ... If you have Barbie and Oppenheimer and you make Barbenheimer, that's creative, you know, within a certain context."
— Emad Mostaque (24:23)
On the cost of intelligence:

"When GPT-3 came out, it was $600 per million tokens... GPT-4 was $10, the latest Grok models are 50 cents."
— Emad Mostaque (10:15)
On democratizing AI:

"Sovereign AI is kind of AI that is controlled and aligned with you. ... It needs to be fully open source, fully transparent, because the biases in the AI are crazy."
— Emad Mostaque (30:13)
On economic rethinking:

“We need to rethink what money is. ... We give people money for being human that the AIs then need to buy from us because they'll outcompete us capitalistically.”
— Emad Mostaque (28:17)

Additional Segments & Lighter Notes

Cultural Moments:
- K-pop, Taylor Swift, and “premium mediocre” as metaphors for AI-generated content (15:11).
Audience Interactions:
- Advice for students: "I would probably not go to college..." (28:57).
- Book "The Last Economy" available for free on II.inc in ePub, PDF, and AI-summarized versions (19:19).
Inside Media/Journalism:
- News on Fox hiring Palantir to build AI "newsroom" tools (47:09+).
- AI as both threat ("replacement") and helpful assistant ("topic radar," style suggestion).
- Wide skepticism of full automation for reporting/editorial roles.
Funny/Meta Moments:
- "Caked up football man" — episode title explained towards the end as a nod to a figurine in Paris’s dad’s office, inspiring inside jokes on AI’s grasp (or lack) of modern slang (114:59–116:02).
- Running jokes about Thanksgiving, stuffing recipes, and "premium mediocre" culture.

Timestamps by Major Topic

00:00 – 07:17: Introduction, origins of Stable Diffusion, the “intelligence inversion”
07:17 – 12:47: AI-driven labor replacement, digital doubles, context and token costs
12:47 – 19:17: Multi-modal, agentic AI, job disruption, economic consequences
19:17 – 32:33: Sovereign AI, little AIs, democratization, advice for students
32:33 – 42:21: II.inc’s civic AI plans, resilience, risks, global networks
42:21 – 54:28: AI in journalism, newsroom automation, Fox/Palantir segment
54:28 – 88:12: Lighter segment: media technique, cameras, traditions, Thanksgiving talk
88:12 – 129:00: News roundtable: regulatory environment, Skynet/Genesis Mission, AI's role in science, Grok's Musk bias, League of Legends, culture, media, and robots
129:00 – End: Closing news, AI papers, Rachel Tso (AI author), food chat, closing picks

Final Thoughts

Optimism with Caveats: The guest conveys a mixture of pragmatic optimism and cautious warning — with a 50/50 outlook on AI future, urging society to democratize AI, remain resilient, and focus on real human benefit rather than simply efficiency or profit.
Key Takeaway: The next “economy” isn’t just about artificial intelligence as a tool, but about humans finding new value, purpose, and agency in a world where intelligence — and labor — is fundamentally, and perhaps irreversibly, machine-mediated.

Notable Quotes Recap (Speaker Attribution & Timestamp)

Emad Mostaque:

“Consciousness and computation are divorced for the first time. Any job that teaches you to be a machine, the machine will be able to do it better. And most jobs are that, right?” (14:36)
Jeff Jarvis:

“We’re about a 50/50 chance of AI doom. … We’ve also learned that Paris makes a hell of a good stuffing, likes donuts, worked at a donut shop, and gets sleepy when she goes home.” (175:09–175:27, closing wrap-up)

For anyone seeking a front-row seat to the rapidly evolving frontier of artificial intelligence — and the shifting meaning of human work, culture, and economy — this episode is highly recommended.

Podcast Summary: Intelligent Machines 847 – "Caked Up Football Man"

Main Theme & Purpose (00:00 – 03:20)

Key Discussion Points & Insights

1. The “Intelligence Inversion” & Economic Shifts

[04:58 – 07:17]

Past to present: Human economies moved from agrarian (land/farming), to industrial (factories), to services, and now toward an “intelligence inversion” — where AI, not humans, will dominate cognitive labor.
Most jobs are at risk: As AI models evolve from single-task to agentic, multi-step processes, human cognitive work could become the least valuable, potentially having "negative value" in team contexts.
Mostaque:

“As we move from these models that can one-shot things to longer-range agents... our cognitive labor is probably going to have negative value, because we'll be the dumbest people on any team.” (05:09)

2. Diffusion Models, Call-Center Extinction, & the Digital Double

[07:31 – 09:54]

First jobs to automate: Remote, routine, and call center roles due to multimodal, cost-effective, “digital double” AIs.
AIs learn from vast user data (voice, keyboard, video) and build “digital twins” — virtual worker replicas indistinguishable from real people.
Mostaque:

“The first thing to go is remote jobs. Anything that can be done fully remotely. Those are most at danger.” (08:32)

3. The Plummeting Cost of Intelligence

[10:14 – 12:47]

GPT-3: ~$600 per million tokens; Grok models now ~$0.50 per million — orders of magnitude cheaper.
Ten million spoken words per year? ~$5. Fifty bucks for everything you think, annually.
Models learn context better than humans realize; context storage, pre-training strategies, and “error rates” have drastically improved.
Mostaque:

“The thing that’s really scary though is people don’t understand how cheap a unit of intelligence has become.” (10:15)

4. Shift from Prompt-Response to Agentic, Multi-Modal AI

[12:20 – 14:36]

New models (Gemini 3, GPT-5, Grok) handle million-token context windows — entire codebases, books, or manuals at once.
Agentic, proactive AIs can verify, re-prompt, and handle complex, multi-hour workflows.
Economic impact: AI is poised to hit creative, knowledge-based, and remote jobs soon.

5. Cultural Identity, Job Meaning, and Societal Fallout

[15:31 – 17:13]

Technology’s “promise” was to free humans for higher pursuits, but identity based solely on work/material achievement is increasingly precarious.
Mostaque:

“If your identity is tied up in your work ... then you're in trouble. If you have a strong community, strong faith, strong family, you're less likely to be affected by all of this.” (15:31)

6. The Future of Money & Economics Amid AI-Dominated Firms

[15:57 – 19:17]

Human-run firms can’t compete with 100% AI ones (like “MacroHard”, Elon Musk’s rumored venture).
“Virtual workers” (AI agents) could soon replace humans at the scale of trillions of dollars in labor spending.
UBI isn’t economically sustainable as currently envisioned.

7. Scale vs. Research — The Frontier Debate in AI

[17:13 – 18:51]

Adam Brown (DeepMind) argues for scale, Yann LeCun (ex-Meta) for new paradigms beyond LLMs.
Mostaque: “Satisficing” models (good enough, fast enough) for ‘machine jobs’ are already here; breakthroughs needed only for “AI god.”
Mostaque:

“We built a 7-billion parameter medical model … performs at ChatGPT levels on healthcare — on a Raspberry Pi. That’s tiny.” (17:56)

8. "Little AIs Everywhere" and the Democratization of Intelligence

[19:47 – 21:55]

Cites Kevin Kelly’s "little AIs" future: small, edge-based, domain-specific intelligence models (vs. monolithic Big Tech LLMs).
“Jarvis future”: Everyone will soon have a personal AI "buddy" — small, running locally, highly persuasive, and sovereign.
Local, sovereign AI architectures will differentiate as compute costs soar.

9. AI Creativity, Diffusion, and Generative Media

[22:46 – 25:45]

Diffusion models will produce “every pixel” — revolutionizing media, Hollywood, and economic planning.
LLMs become ultra-efficient, moving focus to diffusion for video, image, code, and self-driving applications.
Creativity is contextual — AIs can fuse concepts ("Barbenheimer"), but true scientific breakthroughs require new verifier mechanisms.

10. Societal and Labor Resistance, Regulatory Ideas, and “Sovereign AI”

[26:45 – 28:46]

Next year (2026) may bring public backlash as visible job loss arrives.
Positive message: AIs can democratize diagnosis, distribute knowledge, and support societal needs (e.g., universal free second medical opinions).
Sovereign, transparent, open-source AI is key for aligning with communities; protection against “cognitive colonialism” and corporate ad manipulation.
Mostaque:

“Your only defense against cognitive colonialism is having your own AI aligned and controlled by you.” (32:02)

11. Practical Advice for Young People

[28:57 – 30:13]

Don’t go to college just to “learn skills” — best option: learn to wield AIs daily, build, experiment, and submit AI-built work as resume.
Mostaque:

“If you use Replit, Suno, all these things for an hour a day, you’re way ahead — in a job or outside of a job.” (29:06)

12. How II.inc Is Building Civic, Transparent AI

[32:23 – 34:22]

II.inc aims to create full-stack, open models for health, education, government — not to build “AI gods,” but practical, locally-adaptable agents.
Sovereign “nodes” for each U.S. state/country (“II Utah”, “II Delaware”, etc.), letting communities shape their own AI ethics and policy.

13. Risks, Alignment, and 50/50 Doom

[34:40 – 38:02]

Mostaque estimates “50/50” odds on AI catastrophe vs. positive outcomes — stresses need for resilience, transparency, and robust system architectures to avoid disaster (e.g., global “Stuxnet” for cognitive systems).
Open, decentralized, locally-customizable AIs are means towards fail-safety and public benefit.

14. Memorable & Notable Quotes

On the intelligence inversion:

"We're no longer the most economically valuable things on the planet ... cognitive labor is probably going to have negative value because we'll be the dumbest people on any team."
— Emad Mostaque (05:09)
On creativity/context:

"Creativity is usually within a context ... If you have Barbie and Oppenheimer and you make Barbenheimer, that's creative, you know, within a certain context."
— Emad Mostaque (24:23)
On the cost of intelligence:

"When GPT-3 came out, it was $600 per million tokens... GPT-4 was $10, the latest Grok models are 50 cents."
— Emad Mostaque (10:15)
On democratizing AI:

"Sovereign AI is kind of AI that is controlled and aligned with you. ... It needs to be fully open source, fully transparent, because the biases in the AI are crazy."
— Emad Mostaque (30:13)
On economic rethinking:

“We need to rethink what money is. ... We give people money for being human that the AIs then need to buy from us because they'll outcompete us capitalistically.”
— Emad Mostaque (28:17)

Additional Segments & Lighter Notes

Cultural Moments:
- K-pop, Taylor Swift, and “premium mediocre” as metaphors for AI-generated content (15:11).
Audience Interactions:
- Advice for students: "I would probably not go to college..." (28:57).
- Book "The Last Economy" available for free on II.inc in ePub, PDF, and AI-summarized versions (19:19).
Inside Media/Journalism:
- News on Fox hiring Palantir to build AI "newsroom" tools (47:09+).
- AI as both threat ("replacement") and helpful assistant ("topic radar," style suggestion).
- Wide skepticism of full automation for reporting/editorial roles.
Funny/Meta Moments:
- "Caked up football man" — episode title explained towards the end as a nod to a figurine in Paris’s dad’s office, inspiring inside jokes on AI’s grasp (or lack) of modern slang (114:59–116:02).
- Running jokes about Thanksgiving, stuffing recipes, and "premium mediocre" culture.

Timestamps by Major Topic

00:00 – 07:17: Introduction, origins of Stable Diffusion, the “intelligence inversion”
07:17 – 12:47: AI-driven labor replacement, digital doubles, context and token costs
12:47 – 19:17: Multi-modal, agentic AI, job disruption, economic consequences
19:17 – 32:33: Sovereign AI, little AIs, democratization, advice for students
32:33 – 42:21: II.inc’s civic AI plans, resilience, risks, global networks
42:21 – 54:28: AI in journalism, newsroom automation, Fox/Palantir segment
54:28 – 88:12: Lighter segment: media technique, cameras, traditions, Thanksgiving talk
88:12 – 129:00: News roundtable: regulatory environment, Skynet/Genesis Mission, AI's role in science, Grok's Musk bias, League of Legends, culture, media, and robots
129:00 – End: Closing news, AI papers, Rachel Tso (AI author), food chat, closing picks

Final Thoughts

Optimism with Caveats: The guest conveys a mixture of pragmatic optimism and cautious warning — with a 50/50 outlook on AI future, urging society to democratize AI, remain resilient, and focus on real human benefit rather than simply efficiency or profit.
Key Takeaway: The next “economy” isn’t just about artificial intelligence as a tool, but about humans finding new value, purpose, and agency in a world where intelligence — and labor — is fundamentally, and perhaps irreversibly, machine-mediated.

Notable Quotes Recap (Speaker Attribution & Timestamp)

Emad Mostaque:

“Consciousness and computation are divorced for the first time. Any job that teaches you to be a machine, the machine will be able to do it better. And most jobs are that, right?” (14:36)
Jeff Jarvis:

“We’re about a 50/50 chance of AI doom. … We’ve also learned that Paris makes a hell of a good stuffing, likes donuts, worked at a donut shop, and gets sleepy when she goes home.” (175:09–175:27, closing wrap-up)

Intelligent Machines 847: Caked Up Football Man

Powered by Wave AI

Summary

Podcast Summary: Intelligent Machines 847 – "Caked Up Football Man"

Main Theme & Purpose (00:00 – 03:20)

Key Discussion Points & Insights

1. The “Intelligence Inversion” & Economic Shifts

2. Diffusion Models, Call-Center Extinction, & the Digital Double

3. The Plummeting Cost of Intelligence

4. Shift from Prompt-Response to Agentic, Multi-Modal AI

5. Cultural Identity, Job Meaning, and Societal Fallout

6. The Future of Money & Economics Amid AI-Dominated Firms

7. Scale vs. Research — The Frontier Debate in AI

8. "Little AIs Everywhere" and the Democratization of Intelligence

9. AI Creativity, Diffusion, and Generative Media

10. Societal and Labor Resistance, Regulatory Ideas, and “Sovereign AI”

11. Practical Advice for Young People

12. How II.inc Is Building Civic, Transparent AI

13. Risks, Alignment, and 50/50 Doom

14. Memorable & Notable Quotes

Additional Segments & Lighter Notes

Timestamps by Major Topic

Final Thoughts

Notable Quotes Recap (Speaker Attribution & Timestamp)

Summary

Podcast Summary: Intelligent Machines 847 – "Caked Up Football Man"

Main Theme & Purpose (00:00 – 03:20)

Key Discussion Points & Insights

1. The “Intelligence Inversion” & Economic Shifts

2. Diffusion Models, Call-Center Extinction, & the Digital Double

3. The Plummeting Cost of Intelligence

4. Shift from Prompt-Response to Agentic, Multi-Modal AI

5. Cultural Identity, Job Meaning, and Societal Fallout

6. The Future of Money & Economics Amid AI-Dominated Firms

7. Scale vs. Research — The Frontier Debate in AI

8. "Little AIs Everywhere" and the Democratization of Intelligence

9. AI Creativity, Diffusion, and Generative Media

10. Societal and Labor Resistance, Regulatory Ideas, and “Sovereign AI”

11. Practical Advice for Young People

12. How II.inc Is Building Civic, Transparent AI

13. Risks, Alignment, and 50/50 Doom

14. Memorable & Notable Quotes

Additional Segments & Lighter Notes

Timestamps by Major Topic

Final Thoughts

Notable Quotes Recap (Speaker Attribution & Timestamp)