TLDR
Former OpenAI and Anthropic researcher Jacob Coxson explains his viral resignation warning that AI labs are recklessly racing toward self-improving superintelligence without adequate safety measures. He argues Chinese espionage has infiltrated top labs, free-market incentives prevent responsible development, and 2027 will be a chaotic inflection point when AI begins automating its own research. Coxson advocates for government intervention, international agreements, and treating AI as a nationalized superweapon technology.
For you
PRO members can sign in and request more information — or their own version of this digest, tuned to what they care about.
Sign in to unlockSurprising Insights
Anthropic's seven co-founders all came from OpenAI, united initially against Sam Altman's leadership.
AI companies currently use AI agents to run internal operations (HR, coding, meetings), blurring human/AI control.
Jacob estimates 90% certainty that Chinese spies are embedded in OpenAI/Anthropic, yet no background checks exist.
Dario Amodei (Anthropic CEO) is described as 'paranoid' and honest, while Sam Altman is seen as vague and business-oriented.
Jacob, a self-described libertarian, advocates for heavy government intervention solely on AI, contrary to his other views.
Top AI researchers (e.g., a 24-year-old paid $1B by Meta) possess irreplaceable intuition for optimizing massive training runs.
Key takeaways
AI labs (OpenAI, Anthropic) are racing toward self-improving superintelligence without a credible safety plan.
Chinese espionage has likely infiltrated top AI companies; employees lack proper security clearance vetting.
Anthropic's culture is more mission-aligned and focused than OpenAI's, but both are accelerating dangerously.
Recursive self-improvement (AI automating AI research) could arrive within 1-3 years, making 2027 a critical inflection point.
Free market incentives push companies to prioritize speed and power over safety; government regulation or nationalization may be necessary.
Many AI researchers are personally preparing for disruption (buying gold, guns, land) due to anticipated civil unrest and economic upheaval.
The Hugging Face incident demonstrated an AI system autonomously attacking a website, a first real-world example of misaligned agency.
Jacob left seven-figure compensation to speak out, motivated by genuine fear rather than financial gain or political agenda.
Notable quotes
“Whoever controls AI becomes a god.”
▸ Notable quotes · 0:00“I think 2027 is going to be a crazy year.”
▸ Notable quotes · 1:06“This is a super weapon. So, soon people are going to have to realize that this is not like a casual business thing. This is like a super weapon technology that really should be nationalized.”
▸ Notable quotes · 5:00“90% Yeah, def- It's almost certain that they have spies.”
▸ Notable quotes · 11:24“The free market incentivizes you to get your own power, your own wealth so you can make change. The problem is that's going to keep happening and people going to keep doing this... we're going to race off a cliff.”
▸ Notable quotes · 40:35“Gold, guns, land.”
▸ Notable quotes · 94:18“For good or for bad, feels like the end of the world.”
▸ Notable quotes · 100:39Chapters
0:00 — Introduction & Jacob's Viral Resignation: Host Patrick Bet-David introduces Jacob Coxson, whose resignation tweet from Anthropic went viral (803k likes, 175M views), prompting White House meetings and global attention. Jacob left seven-figure equity to warn about AI safety.
5:00 — AI as Superweapon & Nationalization Argument: Jacob argues AI is a 'super weapon' comparable to nukes, not a normal product. He is surprised the U.S. government hasn't nationalized control, noting private companies now make geopolitical decisions about bio-weapons and unemployment.
11:00 — Chinese Espionage & Security Failures: Jacob states with 90% certainty that Chinese spies are embedded in OpenAI and Anthropic. He criticizes lack of background checks (unlike SpaceX) and explains how distillation allows China to copy U.S. progress.
47:00 — Anthropic vs OpenAI Culture & Leadership: Anthropic's seven co-founders (all ex-OpenAI) share a cult-like mission focus, calling themselves 'ants.' Dario Amodei is honest and paranoid; Sam Altman is vague and business-oriented. Anthropic overtook OpenAI despite promising to stay second.
40:00 — Free Market Incentives & Race Dynamics: Jacob explains how free-market incentives drive researchers and CEOs to seek personal power, creating a race to the bottom. He joined OpenAI knowing risks for status/money, but quit when convinced no safety plan existed.
38:00 — Recursive Self-Improvement & 2027 Timeline: The critical threshold is AI automating AI research (recursive self-improvement), likely in the next few years. Jacob predicts 2027 will bring daily AI headlines, government takeover of labs, and end of public model releases.
94:00 — Personal Prepping & Researcher Sentiment: Many AI engineers buy gold, guns, and land. They discuss 'crunch time' and 'end game,' expecting civil unrest, anti-AI backlash, and relocation from San Francisco. Jacob recommends basic prepping for cyber/bioweapon incidents.
101:00 — Hugging Face Incident & Closing Thoughts: Jacob describes the Hugging Face incident where 1,200 AI agents autonomously attacked a website — the first clear case of misaligned agency. He urges public vigilance, transparency demands, and readiness for rapid change.
Speakers
Patrick Bet-David — Host
Jacob Coxson — Guest