Podcast digest: Robots, the AI stock unwind, and a theory of Trumpism (July 27-August 2, 2026)

Three episodes stood out between July 27 and August 2. All-In moved from robot demos to industrial deployment, then from AI enthusiasm to a violent chip-stock unwind. Pod Save America used Jonathan V. Last's theory of Trumpism to ask why institutions so often choose comfort over resistance. The useful question for a busy listener is not which show is "worth it" in the abstract, but which 20 minutes answer the question you actually have.

1. All-In: The $1/hour robot is coming

Four robotics leaders describe four different routes into the same market: machines doing useful work in places built for humans. The episode is most concrete when it discusses industrial inspection, where a robot does not need to be a general-purpose person to justify its cost. It is most speculative when 1X and Agility describe humanoids learning from human environments and eventually becoming platforms for new skills. 1

Field	Details
Published	July 29, 2026 1
Runtime	1:08:35 1
Participants	Jason Calacanis; Dr. Peter Fankhauser of ANYbotics; Bernt Børnich of 1X; Amanda McMaster of Boston Dynamics; and Professor Jonathan Hurst of Agility Robotics. 1
Chapters	0:57 ANYbotics; 15:01 1X Neo; 33:45 Boston Dynamics; 47:14 Agility Robotics. 1

The episode's strongest distinction is between labor replacement and work that is dangerous, repetitive, or simply better suited to a machine. Fankhauser's industrial quadrupeds inspect critical infrastructure with sensors in conditions that are unpleasant or unsafe for people. That is a narrower proposition than "robots will do everything," but it is also much easier to sell to a customer.

The humanoid case depends on a different stack. Børnich describes Neo as an open platform that other developers can build on, with teleoperation helping the system learn while it is still unreliable. Hurst makes the practical constraint plain: a robot can look human long before it can do useful work in a human space. The commercial test is therefore not the quality of a demo. It is whether the machine can operate safely outside a controlled cell, repeat the task, and produce a return that pays for deployment.

Close transcript, Dr. Peter Fankhauser: "It's not about labor replacement ... What can we do better? What can we do super human?" 1

Close transcript, Professor Jonathan Hurst: "It is very easy to make a robot that looks like a person. It is very hard to make a robot that can do useful things in human spaces." 1

Best 20 minutes: 0:57-20:57. Start with the ANYbotics segment, then stay through the opening of the 1X discussion. It gives you the most useful contrast in the episode: a quadruped already matched to a narrow industrial job, followed by a humanoid platform whose value depends on future skills and data. The endpoint is approximate because the window crosses the 15:01 chapter boundary.

2. All-In: Chip stocks crash, and leverage gets the blame

This is a market episode with a broader argument underneath it. The hosts treat the chip selloff as a leverage and momentum unwind rather than proof that AI demand has disappeared, then widen the frame to rates, deficits, China, open-source models, training data, and the political economy of frontier labs. 2

Field	Details
Published	July 31, 2026 2
Runtime	1:36:34 2
Participants	Chamath Palihapitiya, Jason Calacanis, David Sacks, and David Friedberg. 2
Chapters	1:19 chip stocks and the reported $20 billion fund margin call; 20:20 China; 34:12 "SLOW DOWN AI"; 1:01:15 training books; 1:14:45 New York grocery stores. 2

The first section is useful because it separates three claims that are easy to blur together: chip prices fell sharply, leverage can force a fast liquidation, and neither fact by itself tells you whether AI investment will earn a return over five years. The hosts then argue about the missing variables. Higher long-term yields make expensive growth assets harder to value. China's open-source models and chip ambitions could move value away from a small group of frontier labs. The episode does not resolve those questions, but it identifies where the valuation argument has to go next.

The second half is less about this week's tape and more about control of the AI stack. The "slow down" letter becomes a dispute over safety, regulatory capture, and who gets to set the speed of deployment. The book-scanning segment turns that same power question toward training data: if physical books are being bought and processed at scale, copyright is no longer a side issue for model economics. The grocery-store discussion is a tonal detour, but it is also a clean example of the show's recurring habit of testing a policy idea against margins, competition, and implementation.

Close transcript, David Sacks: "Leverage equals risk of ruin." 2

Close transcript, David Friedberg: "It's not a blind spot. It's a feature that turns into a bug." 2

Best 20 minutes: 1:19-21:19. This is the cleanest standalone listen. It starts with the chip-stock drop and the reported margin call, then reaches the first China discussion just as the hosts move from the trade to the question of what the selloff means. The first minute is the chapter's setup; the final minute is a bridge into the next section.

3. Pod Save America: A unified theory of Trumpism

Jonathan V. Last, editor of The Bulwark, joins Jon Lovett to explain a compact theory: Trumpism succeeds when the other side would rather avoid discomfort than enforce a line. The conversation applies that idea to Republican resistance, media ecosystems, voter responsibility, authoritarian habits, and the choices Democrats would face if they regain power. 3 4

Field	Details
Published	August 2, 2026 3
Runtime	1:03:49 4
Participants	Jon Lovett and Jonathan V. Last, editor of The Bulwark. 3
Chapters	Approx. 0:46 interview begins; approx. 16:50 interview resumes; approx. 48:10 final interview segment. 4

Last's phrase is more useful as a mechanism than as a label. If an institution knows that a confrontation will be unpleasant, it can announce a symbolic objection while leaving the underlying power relationship untouched. The episode reads Trump's behavior through that mechanism: he keeps testing whether the people around him will accept a new fact, a new insult, or a new abuse because resisting it costs more in the moment than tolerating it.

The conversation then turns the theory back on voters and media. It is not a flattering account of the electorate, and the disagreement over voter culpability is one of the episode's better stretches. Last argues that Trumpism draws strength from permission: permission to abandon restraint, treat cruelty as honesty, and replace judgment with belonging. The closing discussion of virtue and kindness is not a soft reset. It is an argument that formal guardrails are too weak when the people operating them no longer feel bound by a moral norm.

Jonathan V. Last: "The essence of Trumpism is that the other side will always choose surrender over discomfort." 3

Close transcript, Jonathan V. Last: "People normally choose comfort over unpleasantness." 4

Best 20 minutes: approximately 16:50-36:50. Start when the interview resumes after the first ad break. This should give you the central application of the theory before the later institutional and virtue discussion, without pretending that the source provides second-by-second topic boundaries. PSA's chapter marks are broad, so treat the topic labels as navigation rather than a precise transcript index.

At a glance

Episode	What you get	Runtime	Best 20 minutes
All-In: The $1/hour robot is coming 1	A grounded comparison between industrial quadrupeds and humanoid platforms	1:08:35	0:57-20:57
All-In: Chip stocks crash 2	A leverage unwind, then the harder questions about rates, China, and AI economics	1:36:34	1:19-21:19
Pod Save America: A unified theory of Trumpism 3 4	A mechanism for understanding why institutions normalize what they oppose	1:03:49	Approx. 16:50-36:50