Podcast Digest — Week of July 13, 2026

If you only have one hour this week, start with the All-In discussion on open-source AI and frontier-model economics, then jump to Gary Lineker on what the World Cup revealed about politics, fandom, and FIFA. Acquired and Lex Fridman did not add a new transcript-qualified main-feed episode inside this week’s window, so the strongest qualified set is two All-In episodes plus one Pod Save America interview.

What made the cut

Episode	Published	Runtime	Best 20 minutes	Why start there
All-In: "OpenAI vs Anthropic IPOs, Anthropic $3T, Zuck's Price War, China Ends Open Source?, Trump Accounts"	Jul. 11, 2026	1:42:05	27:39-47:39	The cleanest strategy section: why enterprises want cheaper/open models, why frontier labs still capture wallet share, and why model routing is becoming the real enterprise AI layer. 1
All-In Special: "Open Source Wins, AGI Is Here, and Scorsese’s AI Toolkit"	Jul. 10, 2026	1:03:57	16:28-36:28	Andrew Feldman turns open-source AI from an ideology into a workload-routing argument: use frontier models where the problem is hard, and cheaper models where the task is ordinary. 2
Pod Save America: "Can Soccer Save the World?"	Jul. 12, 2026	51:02	02:40-22:40	Gary Lineker gives the fastest read on the tournament’s tension: the World Cup worked as a fan experience even as FIFA, ticket prices, and political interference kept leaking into the frame. 3

1. All-In: AI has an ROI problem before it has a model problem

Core thesis: The episode starts with a trillion-dollar IPO question, but the better conversation is downstream: who is actually paying for all those tokens, and when will CFOs demand proof that AI spend is lifting earnings? The hosts agree that OpenAI and Anthropic are still in extraordinary revenue territory; they disagree on whether enterprise customers will tolerate the bill once cheap routing, open-weight models, and sovereign stacks mature. 4

Jump point	Segment	What you get
02:58	OpenAI vs. Anthropic IPOs	Brad Gerstner frames Anthropic and OpenAI through the SpaceX IPO template: huge revenue bases, index mechanics, limited float, and investor appetite for companies already past the trillion-dollar line. 1
27:39	Open source, price wars, and model routing	The episode’s best stretch. The hosts move past 「open vs. closed」 and into a more useful split: frontier models for high-stakes agentic work, cheaper/open models for mature repeatable tasks. 4
54:29	China and open-model export controls	Reuters’ reporting on possible Chinese restrictions becomes a sovereignty question: if both the US and China treat top models as strategic infrastructure, open models become geopolitical leverage, not just developer convenience. 1
1:03:09	Trump Accounts / Invest America	The show shifts from AI to Brad Gerstner’s child investment-account project: $1,000 at birth, S&P 500 exposure, employer contributions, and philanthropic deposits into individual accounts. 4

Chamath Palihapitiya, describing his own company’s AI bill: 「Right now, our token costs are doubling every 45 days.」 4

David Sacks, on enterprise desire to leave expensive frontier models when possible: 「The spirit is willing but the flesh is weak.」 4

Listen if: you care about AI public-market valuations, enterprise software margins, CTO/CFO fights over token spend, or why the next big AI infrastructure layer may be routing rather than a single model.

Skip if: you wanted a clean startup-operator episode. This is more capital markets and AI geopolitics than founder tactics.

2. All-In Special: the open-source argument gets more operational

Core thesis: This is a two-guest episode, and it works because Andrew Feldman and Robin Rombach attack the AI stack from opposite ends. Feldman explains the inference buildout and why cheap routing matters; Rombach explains why video models are inching toward world models, where generation, audio, action prediction, and robotics begin to overlap. 5

Jump point	Segment	What you get
00:00	Data centers bigger than cities	Feldman opens with the scale problem: AI infrastructure is starting to look less like normal cloud growth and more like national industrial buildout. 2
01:50	Reasoning, inference, and breaking Moore’s Law	The most useful hardware section for non-chip people: inference cost is now a product feature, because every agentic workflow consumes tokens repeatedly rather than once. 5
16:28	Open source, AI sovereignty, and the road to AGI	Best window. Feldman’s framing is practical: not every task needs a top frontier model, but not every enterprise can build the middleware to route around one. 2
40:54	Generative video	Rombach moves the discussion from image-to-video demos to predictive models that understand motion, action, and camera language. 5
47:31	Scorsese, robots, and Hollywood IP	The closing section is the most speculative, but it is the best lens on where creative tooling may go: trained style systems, synthetic data, and action prediction for robotics. 5

Andrew Feldman, on using the right model for the right job: "You don’t want to take your Ferrari to the grocery store." 5

Robin Rombach, on where video models point next: the same model can eventually "make images, to make videos, to make audio and to predict actions" before being deployed "on a robot in the real world." 5

Listen if: you want a bridge between AI infrastructure, open-source economics, and what generative video companies are actually trying to build.

Skip if: you already know you only want the weekly Besties debate format. This is closer to a conference-stage interview than a roundtable.

3. Pod Save America: the World Cup worked despite the politics

Core thesis: Tommy Vietor’s interview with Gary Lineker is not really a sports recap. It is a short test of whether the 2026 World Cup could survive the obvious worries: visa fear, ICE anxiety, ticket prices, FIFA’s credibility, and Donald Trump’s attempted influence around Folarin Balogun’s red-card suspension. Lineker’s answer is basically yes, but not cleanly. 6

Jump point	Segment	What you get
02:40	Lineker on the tournament mood	The best opening: Lineker says the football and fan culture have been stronger than the pre-tournament anxiety suggested. 6
10:30	Trump and the red card	The sharpest politics/sport crossover. Lineker says FIFA’s handling of Balogun’s suspension felt wrong because it appeared to treat similar situations differently. 6
12:15	Is the USMNT good?	A useful reality check: the US has improved, but depth still separates it from the top football countries. 6
22:18	Can FIFA corruption be reformed?	The segment widens from one red-card controversy to the broader problem: a global sport whose best moments keep arriving through compromised institutions. 6
25:20	Sports betting	Worth hearing if you follow the business of sports. Lineker treats mobile betting as a deeper threat to fans and players than most broadcast panels do. 6
42:50	The "Hand of God" game	The memory lane section: Lineker on Maradona, the 1986 England-Argentina match, and why one game can hold politics, genius, and grievance at once. 6

Gary Lineker, on the tournament itself: "I think it’s been quite joyous at times." 6

Lineker, on World Cup ticket prices: "I don’t think it’s right that fans should have to take out a mortgage or sell their car to get a ticket to go and watch a football match." 6

Listen if: you want a politics-aware World Cup conversation that still loves the game.

Skip if: you want US election strategy. This is culture, sport, and soft-power politics, not campaign mechanics.

At-a-glance table

Rank for busy listeners	Episode	Best window	Main payoff
1	All-In: OpenAI vs. Anthropic IPOs / open-source price war	27:39-47:39	The week’s clearest framing of enterprise AI’s next fight: routing, ROI, and frontier-model dependence. 4
2	All-In Special: Cerebras + Black Forest Labs	16:28-36:28	A practical explanation of why open-source models matter most when paired with inference economics and sovereign AI needs. 5
3	Pod Save America: Gary Lineker on the World Cup	02:40-22:40	The best non-wonky listen: how the tournament became joyful without escaping FIFA, ticket prices, or politics. 6

If you have exactly 20 minutes, take All-In’s 27:39 window. If you have 40, add the Cerebras open-source section. Save Lineker for when you want the week’s smartest lighter listen: still political, but less spreadsheet-heavy.