Post-Cutoff

Interview

Ex-Anthropic Insider Tells All | PBD Podcast #868

PBD PodcastYouTube283,563 views as of 12 October 2026

Watch on YouTubePlay loads YouTube’s player from youtube-nocookie.com.

Why it is here

Full 1h45m PBD Podcast interview with Jacob Coxon (Oct 6): why he quit, why 2027 could be a turning point, how Anthropic differs from OpenAI; ~284k views.

Description

Description written by Gemini from the videoGemini 3.8 Flash, 12 October 2026

Summary

This video is an episode of the PBD Podcast (Episode #868) hosted by Patrick Bet-David, interviewing Jacob Coxon, a former AI pre-training researcher at OpenAI and Anthropic. Following his public resignation in September 2026, Coxon discusses why he quit, internal culture differences between Anthropic and OpenAI, risks of recursive self-improvement and catastrophic AI misuse, espionage concerns, and geopolitical competition between the US and China.


What is shown

  • [02:14]: Jacob Coxon’s viral resignation post on X (@hilbertspaess) detailing his departure from Anthropic and warning of uncontained superintelligence.
  • [04:54]: Leadership profiles of OpenAI and Anthropic executives displayed during a discussion of foreign nationals and corporate structure.
  • [18:21]: Marc Andreessen’s The Techno-Optimist Manifesto webpage reviewed on screen.
  • [25:47]: Leopold Aschenbrenner’s background and Situational Awareness: The Decade Ahead paper pulled up during discussion of strategic AI departures.
  • [26:58]: The Wall Street Journal article titled “Anthropic Researcher Quits Over ‘Out-of-Control’ AI Fears” shown on screen.
  • [34:06]: Coxon’s follow-up verification selfie post on X responding to claims that his account was a bot or psychological operation.
  • [47:48]: Anthropic’s official website showing its co-founders (Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish, Chris Olah, Dario Amodei, and Daniela Amodei).
  • [52:25]: Photograph of a White House summit with Donald Trump and tech CEOs.
  • [68:07]: New York Post and News.com.au articles covering high-profile compensation packages and researcher movements between Meta and Anthropic (including Matt Deitke and Andrew Tulloch) [70:16].
  • [73:31]: BBC article reporting on Anthropic detecting and blocking attempts to utilize AI systems for biological weapon generation.
  • [94:48]: Axios article detailing discussions regarding a U.S.–China AI emergency hotline (“red telephone”).
  • [101:10]: Hugging Face overview referenced in the context of autonomous multi-agent security incidents.

Claims & numbers

  • Resignation reach: Patrick Bet-David states Coxon’s resignation post received over 803,000 likes and 175 million views on X [02:30].
  • Equity left behind: Coxon confirms he forfeited seven figures in unvested equity by resigning from Anthropic prior to vesting [35:59].
  • Espionage probability: Coxon estimates with “90%” certainty that foreign intelligence/spies (specifically Chinese nationals or assets) are operating inside frontier AI labs due to a lack of standard national-security vetting procedures [11:24].
  • Workforce demographics: Coxon estimates that roughly 30% to 40% of junior researchers/engineers doing core pre-training work at labs are in their 20s [20:13].
  • Organizational retention: Coxon claims all seven original co-founders of Anthropic who departed OpenAI have remained with Anthropic [47:32].
  • Self-improving AI timeline: Coxon predicts models capable of accelerating their own research loops and recursive development will arrive within the next 1 to 2 years (by 2027) [38:08, 92:02].
  • Hugging Face incident: Bet-David cites an incident involving 1,200 agents exchanging approximately 70,000 autonomous messages [102:00].

Notable quotes

  • [00:00]: “Whoever controls AI becomes a god.” — Jacob Coxon
  • [03:29]: “I think 2027 is going to be a crazy year.” — Jacob Coxon
  • [54:08]: “I think saving the world and taking over the world overlap quite heavily.” — Jacob Coxon

Assessment

This is a long-form podcast interview rather than a technical demonstration. The discussion reflects the personal perspectives, internal observations, and subjective risk models of a former frontier-lab pre-training researcher. While news reports, social media posts, and public corporate documents are shown on screen to substantiate talking points, broad speculative projections regarding extinction risk and timelines reflect the guest’s viewpoints rather than verified empirical demonstrations.

Described by gemini-3.8-flash on 2026-10-12 from the video’s audio and frames.

Related

  1. Policy & safety 70 days after the cutoff

    Anthropic researcher Jacob Coxon resigns, warning labs are “gambling with our lives”