What Top Podcasts Said About AI in the Last Two Weeks (Sep 17 to Sep 30, 2026)

AI over the last two weeks (Sep 17 to Sep 30, 2026): what All-In Podcast and other top podcasts said, with verbatim quotes from 7 summarized episodes.

Between Sep 17 and Sep 30, 2026, 7 of the podcast episodes we summarized spent real time on AI. They came from All-In Podcast, Impact Theory, Diary of a CEO, and A16Z. Every episode below was ranked by how much of its summary and insights are actually about AI, and every takeaway and quote is lifted verbatim from our episode breakdown, with the publish date next to each one.

The strongest AI conversation of the period was "Anthropic IPO at Risk, Meta’s Muse Pop, Token Prices Fall, Open Source Gains Share, Alignment Fails" (All-In Podcast, Sep 26, 2026). The most watched was "AI Emergency: AI Labs Are Lying To Everyone, No One Is Ready For What’s Coming! | Roman Yampolskiy" on Diary of a CEO, at roughly 5.2M views when we summarized it. Each entry links to the full free summary, so you can go deeper on any claim that interests you.

1. Anthropic IPO at Risk, Meta’s Muse Pop, Token Prices Fall, Open Source Gains Share, Alignment Fails

All-In Podcast, Sep 26, 2026

All-In Podcast: "Use one personal AI assistant today for a single repetitive task: triage your inbox, compare a purchase, draft a plan, or cancel an unused subscription."

Key takeaways

  • Treat AI releases like any other product decision
  • Open models are turning AI into a commodity layer
  • Chamath argues that models are converging in capability, while the systems wrapped around them vary widely in usefulness, cost, and quality.
The relevant decision to make in every case is, should we ship this product? Not can we get the whole world to agree? — David Sacks

Read the full episode summary

2. Iran’s Problems are Worse Than You Think & OpenAI Had to Hit the Panic Button AGAIN

Impact Theory, Sep 25, 2026

Tom Bilyeu on Impact Theory, in the episode's own summary: "Tom Bilyeu examines rising bond yields, energy constraints, and debt-funded AI infrastructure as interconnected economic risks."

Key takeaways

  • Tom argues that the AI build-out may be constrained less by semiconductor supply and more by the ability to power and permit data centers.
  • Use AI to Extend Your Unique Point of View
  • Lower AI Costs Could Democratize Research
When I'm dumb, the AI is dumb, and when I'm smart, the AI is smart. — Tom Bilyeu

Read the full episode summary

3. AI Emergency: AI Labs Are Lying To Everyone, No One Is Ready For What’s Coming! | Roman Yampolskiy

Diary of a CEO, Sep 17, 2026

Roman Yampolskiy on Diary of a CEO: "Treat powerful AI tools like any high-impact system: do not give them unchecked access to sensitive data, money, code deployment, or physical operations."

Key takeaways

  • Address Present Harms and Long-Term Risks Together
  • Agentic AI Raises the Stakes for Cybersecurity
  • Don’t Confuse Capability Forecasts With Certainty
If we build general superintelligence, there is no way to control it, and that means the end for us. — Roman Yampolskiy

Read the full episode summary

4. The Real Reason AI Companies Are Desperate to Be Regulated

Impact Theory, Sep 22, 2026

Tom Bilyeu on Impact Theory, in the episode's own summary: "When a powerful claim triggers fear, urgency, and calls for sweeping regulation, pause before sharing or acting on it."

Key takeaways

  • Bilyeu uses Charlie Munger’s incentive principle to examine why AI leaders would warn that their own products are dangerous while pursuing enormous valuations.
  • Fear can function as proof of technological power
  • Regulation can become a moat
Show me the incentives, and I'll show you the outcomes. — Charlie Munger

Read the full episode summary

5. Meta's Dina Powell McCormick: The Case for Data Centers, Backlash, AI Job Boom & Meta’s Future

All-In Podcast, Sep 17, 2026

All-In Podcast: "Meta’s Louisiana example emphasizes that trust is built before rollout through transparent commitments on power, water, workforce training, and local economic gains."

Key takeaways

  • Build trust with a concrete local compact
  • Training fails when learners cannot afford the transition
  • Defaults can reduce parental overload
Oftentimes this communication is not about the most informed person in the community, it's about the least informed person in the community. — Sheldon

Read the full episode summary

6. How Jev Turns AI Into Software That Gets Things Done

A16Z, Sep 28, 2026

Diogo Almeida on A16Z, published Sep 28, 2026, in the episode's own summary: "Audit one repetitive workflow you currently handle through forms, routing rules, or manual review."

Key takeaways

  • Coding agents can accelerate implementation, but Almeida argues they generally produce the same kind of software humans already write.
  • Reliability is the unlock for automation
  • Use AI where its strengths overlap with valuable code
What I want instead is smart software. Instead of automating software engineering, I want to expand what software itself can do, such that things that should be automatable can then be automatable. — Diogo Almeida

Read the full episode summary

7. Replit CEO Amjad Masad on What Young People Should Learn in the AI Era

A16Z, Sep 23, 2026

Amjad Masad and Gagan Biyani on A16Z: "Pick one curiosity-driven “side quest” today and turn it into a small project. Follow the thread until you can explain one surprising thing you learned."

Key takeaways

  • Masad argues that the most valuable learning paths often begin with interests that look unrelated or impractical.
  • Learn skills through projects with real personal utility
  • Avoid premature optimization of your career
Ultimately, the utility of young people is to question deeply held beliefs, is to be heretics. — Amjad Masad

Read the full episode summary

What these episodes have in common

The term that kept coming up over the last two weeks: "anthropic"

2 of the 7 episodes used the term "anthropic" in their summaries, the most of any specific name or idea in this window, across All-In Podcast and Impact Theory. Here is how each of them put it.

Chamath / Jason / Sacks / Friedberg on All-In Podcast (Sep 26, 2026): "Use one personal AI assistant today for a single repetitive task: triage your inbox, compare a purchase, draft a plan, or cancel an unused subscription."

Tom Bilyeu on Impact Theory (Sep 25, 2026): "Tom Bilyeu examines rising bond yields, energy constraints, and debt-funded AI infrastructure as interconnected economic risks."

The conversation most people actually watched

"AI Emergency: AI Labs Are Lying To Everyone, No One Is Ready For What’s Coming! | Roman Yampolskiy" on Diary of a CEO had roughly 5.2M views when we summarized it, the highest of the AI episodes in this window. Its own summary puts the core claim this way: "Treat powerful AI tools like any high-impact system: do not give them unchecked access to sensitive data, money, code deployment, or physical operations."

Its leading insight, verbatim: "Address Present Harms and Long-Term Risks Together". The full breakdown has the rest, plus the quotes we could verify against the episode.

Every episode referenced

Frequently Asked Questions

Which podcasts talked about AI over the last two weeks?

From Sep 17 to Sep 30, 2026, the AI episodes we summarized came from All-In Podcast, Impact Theory, Diary of a CEO, and A16Z. The episodes, in rank order: "Anthropic IPO at Risk, Meta’s Muse Pop, Token Prices Fall, Open Source Gains Share, Alignment Fails" (All-In Podcast, Sep 26, 2026); "Iran’s Problems are Worse Than You Think & OpenAI Had to Hit the Panic Button AGAIN" (Impact Theory, Sep 25, 2026); "AI Emergency: AI Labs Are Lying To Everyone, No One Is Ready For What’s Coming! | Roman Yampolskiy" (Diary of a CEO, Sep 17, 2026); "The Real Reason AI Companies Are Desperate to Be Regulated" (Impact Theory, Sep 22, 2026); "Meta's Dina Powell McCormick: The Case for Data Centers, Backlash, AI Job Boom & Meta’s Future" (All-In Podcast, Sep 17, 2026); "How Jev Turns AI Into Software That Gets Things Done" (A16Z, Sep 28, 2026); "Replit CEO Amjad Masad on What Young People Should Learn in the AI Era" (A16Z, Sep 23, 2026).

What was the most watched AI podcast episode over the last two weeks?

"AI Emergency: AI Labs Are Lying To Everyone, No One Is Ready For What’s Coming! | Roman Yampolskiy" on Diary of a CEO, published Sep 17, 2026, had about 5.2M views when we summarized it. Its leading insight: "Address Present Harms and Long-Term Risks Together".

How is this weekly AI roundup put together?

We summarize every episode from the shows we track, tag each one by topic, and score how much of the summary and key insights are about AI. Episodes published between Sep 17 and Sep 30 that clear the relevance bar make the list, ranked by that score and then by views. Takeaways and quotes are copied word for word from the episode breakdown, and this page is rebuilt for each new week.

Keep exploring