Gutenberg Lab — Morning Paper

VP Vance Exposes Foreign Lobbying Campaign Against Iran Negotiations

Guests: Joe Rogan (podcast host) × JD Vance (Vice President of the United States)

Bottom line:

A sitting US Vice President publicly accused elements of the Israeli government of funding an influence campaign to sabotage American diplomacy—revealing how foreign actors distort US policy.

Three-Line Summary

Vice President JD Vance disclosed that a TIME magazine investigation (published the day before recording) exposed Israeli government-linked entities paying US influencers and former Trump campaign staffers to attack the Iran peace negotiations and Vance personally. He declared, "When I see Americans allow foreign influence to affect their judgment, my response is: Go to hell. I represent Americans first." Vance debunked the "$300 billion to Iran" claim (it's Gulf Arab private investment, not US taxpayer money) and defended the Memorandum of Understanding as the only viable path to keep the Strait of Hormuz open (20M barrels/day of oil). He also warned that AI-driven inequality, if left unchecked, will fuel communism—proposing stronger unions, immigration restriction, and manufacturing reshoring as solutions.

Key Takeaways

1. TIME Exposé: Israeli Government-Funded Influence Campaign

A TIME article published the day before recording revealed that Israeli government-linked sources paid US influencers and ex-Trump staffers to attack the Iran MOU and Vance. Vance: "When Americans allow foreign influence to affect their judgment, that's what bothers me. Go to hell. I represent Americans first." He positioned himself as defending US sovereignty against foreign lobbying.

2. The "$300B to Iran" Claim Is a Lie—It's Gulf Arab Private Investment

Conservative hawks (including Mike Pence) claim the US is "giving Iran $300 billion." Vance clarified: "Not a single dollar from the US. The MOU allows Gulf Arab states (UAE, Saudi Arabia) to invest in Iran *if* Iran changes behavior." He argued the Strait of Hormuz (25% of global oil supply) cannot be defended by military force alone—cheap drones can shut it down, so diplomacy is essential.

3. AI Inequality Will Breed Communism Unless Workers Gain Power

Vance warned that AI will mirror the Industrial Revolution: massive wealth creation, but if gains accrue only to elites, it will fuel socialism/communism (as robber barons did in the late 1800s). His solutions: (1) Strengthen labor unions (European-style flexibility), (2) Antitrust enforcement against AI monopolies, (3) Restrict immigration (low-wage migrants undercut native workers), (4) Reshore manufacturing. "If we don't give young people a stake—ownership, the American dream—they will become socialists."

Editorial Take

This episode reveals the raw mechanics of how foreign lobbying shapes—or attempts to shape—US policy. While Israeli influence in Washington has long been documented, a sitting VP publicly calling it a "foreign influence campaign" and citing a major news outlet is unprecedented. Equally significant is Vance's articulation of a "populist conservative" economic vision that breaks with Reagan-era GOP orthodoxy: he advocates for stronger unions, immigration restriction, and industrial policy—positions that would have been heretical in the Republican Party 20 years ago. His framing of AI inequality as an existential threat (leading to communism if unchecked) echoes the Progressive Era's response to robber barons, suggesting the Trump administration sees itself as navigating a similar inflection point. Whether this rhetoric translates into policy—and whether the GOP coalition can hold together under these tensions—remains the central question of the next two years.

Source: Joe Rogan Experience "#2526 - JD Vance" (2026-07-15)

(unknown)

• --

AI Disclosure (日本語): 本コンテンツはAIの支援を受けて制作し、編集部が事実確認・分析の監修を行っています。

AI Disclosure (English): Produced with AI assistance; facts and analysis reviewed by our editorial team.

• --

AI and the Future of Math — Connection Over Proof

Featuring: Dwarkesh Patel × Grant Sanderson (Creator of 3Blue1Brown)

Bottom line

AI's fastest progress in mathematics stems not from verifiability but from "grindability" — the ability to run thousands of parallel rollouts in containerized environments, a structural advantage that will spread to other knowledge work.

Three-line summary

Grant Sanderson of 3Blue1Brown argues that now AI has achieved IMO gold, the next frontier is how it solves problems like the Riemann Hypothesis. What matters is the *style* of solution: connecting existing fields (where LLMs should excel), building entirely new theoretical mountains (indicating AGI-level intelligence), or brute-forcing thousand-page proofs (solvable but alien to humans). Mathematicians will shift from proving theorems to curating what's worth exploring.

Three key points

1. Grindability trumps verifiability: Math and coding progress rapidly not just because they're verifiable, but because you can containerize them and run thousands of parallel rollouts, solving the credit assignment problem. Computer use lags despite verifiability because bot detectors prevent parallel farming. With sample efficiency unsolved, AI only learns in domains where you can repeatedly try [00:11:32].

2. The 100-year verification loop: Galois invented group theory at 21, died in a duel, and his papers were rejected as incoherent. It took 100+ years for the value to be recognized, with physics applications (predicting quarks) coming in the 20th century. Truly novel conceptual frameworks can't be verified by contemporaries — if AI creates "alien mathematics," humans risk climbing wrong mountains for years [00:14:09].

3. Mathematicians as curators, not provers: Even if AI proves theorems and explains them well, humans will curate what's worth engaging with. Learning is a social phenomenon; trust in human judgment for navigating infinite space won't disappear. Teaching will be one of the most stable post-AGI careers [00:36:36].

One-line take

Math's lead in AI progress reveals a structural truth: "grindability" — the ability to parallelize containerized trials — matters more than verifiability alone. Whether the Riemann Hypothesis falls via connecting fields or building new theory will determine AGI's arrival timeline and economic spread. Galois's 100-year recognition lag warns us: if AI builds alien mathematics, will humans understand it, or spend years lost in conceptual wilderness?

Source: Dwarkesh Podcast "Grant Sanderson – AI and the future of math" (2026-06-30)

https://www.dwarkesh.com/p/grant-sanderson-2

• --

Palantir-NVIDIA Deal Signals Structural Shift to "AI Infrastructure Sovereignty"

Featuring: Jason Calacanis (host) × David Sacks (former AI Czar, Trump administration) × Chamath Palihapitiya (CEO, 8090) × David Friedberg (CEO, OHALO)

Bottom line

Enterprises have reached a tipping point in recognizing that sharing proprietary data with frontier labs (Anthropic, OpenAI) is tantamount to funding their own competitors.

Three-line summary

The Palantir-NVIDIA partnership crystallizes the "AI infrastructure sovereignty" paradigm. Anthropic blindsided Figma by launching competing app "Claude Design" while its CPO sat on Figma's board—Figma stock fell 50%. Real-world testing by 8090 shows open-source models + control plane are 16.4x cheaper than Claude alone (3x slower, but "a couple extra hours to save 16.4x"). Life sciences companies nearly universally rejected Anthropic's pitch to contribute proprietary datasets—refusing to commoditize billions in R&D assets.

Three key points

1. Anthropic's "Figma betrayal" proves structural risk: Anthropic's CPO served on Figma's board, resigned three days before launching competing product Claude Design. Figma stock fell 50%; Anthropic's valuation surged. Sacks: "They're watching where value is created on top of their models, then moving in directly." Anthropic has launched vertical apps (Claude Science, Security, Legal, Financial, Code) competing with its own customers.

2. Economic case for open-source + harness: 8090's real-world data on code migration task—8090 control plane + open-source model was 16.4x cheaper than Claude alone (3x slower, but "a couple extra hours to save 16.4x"). 8090 + Claude was 1.4x cheaper and 1.5x faster than Claude alone. Chamath: "You can't rent intelligence from the same place that rents it to your competitor."

3. Life sciences industry's collective rejection signals turning point: Friedberg reports Anthropic approached life sciences companies to contribute proprietary datasets in exchange for early access to new model. "Nearly everyone said no," realizing they'd be "commoditizing the asset" they'd spent billions developing. This marks enterprises waking up to the asset value of their data.

One-line take

This is a replay of 1980s Microsoft (swallowing Lotus 1-2-3, WordPerfect) and 2000s Facebook platform (building up Zynga, then crushing it). The critical difference: open-source provides an "exit door." As Sacks put it, "Once a model is open-source, it stops being Chinese—you fork it, run it on US hardware." For the first time, enterprises have a real choice to escape platform lock-in. The Palantir-NVIDIA partnership symbolizes this structural shift.

Additional context: The episode also covered SCOTUS striking down Trump's birthright citizenship EO (6-3), California's fiscal crisis ($1.5-2T in unaccounted liabilities, top 1% paying 33% of state revenue), and debate over AI job displacement (Ramp/Revelio study of 21,000 firms shows AI adopters grew headcount ~10% in two years).

Source: All-In Podcast "AI Sovereignty Wars, Palantir-Nvidia Deal, SCOTUS Birthright Ruling, Newsom's CA Budget Lie" (2026-07-03)

https://allinchamathjason.libsyn.com/ai-sovereignty-wars-palantir-nvidia-deal-scotus-birthright-ruling-newsoms-ca-budget-lie

• --

AI開示文 / AI Disclosure

本コンテンツはAIの支援を受けて制作し、編集部が事実確認・分析の監修を行っています。 / Produced with AI assistance; facts and analysis reviewed by our editorial team.

• --