06-reference

moonshots ep191 balaji ai bitcoin collapse

Thu Aug 28 2025 20:00:00 GMT-0400 (Eastern Daylight Time) ·reference ·source: Moonshots Podcast ·by Peter Diamandis
balaji-srinivasanai-polytheismbitcoinchinanetwork-statesgeopoliticsself-improvement

Moonshots EP 191: Balaji on the State of AI/AGI, Bitcoin and America’s Incoming Collapse

Summary

Deep conversation with Balaji Srinivasan (former CTO of Coinbase, GP at a16z, author of The Network State), joined by Dave Blundin and Salim Ismail. Balaji presents a macro framework: the two growing powers in the world are China (everything physical — manufacturing, military, drones, ships) and the internet (everything digital — AI, crypto, media). Western currencies are collapsing in real terms while capital flows into RMB and BTC. He argues the US cannot out-China China and should instead “be more America than America” via the internet’s strengths: free speech, free markets, decentralization. On China specifically, he pushes back on the Japan comparison: China is genuinely sovereign (can conduct stealth operations, killed US spies in 2010), has introduced a K visa to recruit global tech talent as the US cuts H-1B, and is using demographic pressure to force automation rather than facing Japan-style stagnation. The Pentagon’s own $400M Govini study showed US weapons supply chains are made in China. The core AI thesis is from his paper “AI is polytheistic, not monotheistic” — there will be many AGIs, not one god-like winner. No hard takeoff has occurred; models leapfrog each other within narrow margins. AI currently operates “middle to middle” with bottlenecks at prompting (pointing the fast spaceship) and verification (checking output). Balaji sees AI as reversing 20th-century extreme specialization — “AI means everybody’s a CEO” — and argues the smarter the human, the smarter their AI becomes. Dave reveals that foundation model companies are diverting compute from consumer products to internal self-improvement, confirmed through research after the Kevin Weil interview. Dave estimates a 5-6 month window to determine whether true recursive self-improvement produces a single winner.

Key Segments

Notable Claims