{
 "episode_id": "the-diary-of-a-ceo-2026-07-13-openai-whistleblower-finally-speaks-ai-has-a-70-chance-of-go",
 "title": "OpenAI Whistleblower FINALLY Speaks: \"AI Has A 70% Chance Of Going Horribly Wrong!\"",
 "podcast": "the-diary-of-a-ceo",
 "published_at": "2026-07-13",
 "speakers": [
  "Steven Bartlett",
  "Daniel Kokotajlo"
 ],
 "summary": "Daniel Kokotajlo, a former OpenAI forecaster who left in 2024 and forfeited (then recovered) ~$2M in equity by refusing an anti-disparagement clause, tells Steven Bartlett why he now puts a ~70% chance that AI development goes 'horribly wrong.' He argues the leading labs (OpenAI and Anthropic) are deliberately racing to automate their own AI research first — driving toward superintelligence he expects by roughly the end of the decade — and that this creates twin risks: loss of control (building a smarter 'new species' whose goals we can't verify inside its neural nets) and concentration of power in a handful of CEOs no one should trust with it. He explains how modern AI is trained, why job displacement will likely arrive suddenly and late (after superintelligence), and walks through his group's proposals — from AI 2027's default 'race' scenario to the recommended 'AI 2040 Plan A' (2029 regulation, inference-not-training pauses, total research transparency, reversibility, and a Citizen's Dividend). The conversation closes on the personal toll — his decision to reconsider having more children — and his message that people should stop dismissing this as science fiction, pay attention, and demand regulation before it's too late.",
 "topics": [
  {
   "name": "The 70% thesis and Daniel's mission",
   "timestamp": "0:59",
   "summary": "Kokotajlo opens with his core fear — that we could trade in a new species that ends up ruling the world, with ~70% odds it goes horribly wrong — and frames his mission as forecasting superintelligence and steering it to go well."
  },
  {
   "name": "Superintelligence timelines and Anthropic's growth",
   "timestamp": "3:30",
   "summary": "He defines superintelligence, puts his median arrival around 2029-2030, and cites Anthropic's ~60x revenue growth in a year as evidence of the pace of the trends."
  },
  {
   "name": "Why the average person should care: loss of control and concentration of power",
   "timestamp": "5:00",
   "summary": "Kokotajlo lays out the classic loss-of-control scenario and the concentration-of-power risk (an 'army of geniuses in the data center' owned by one company), plus geopolitical conflict and mass job loss."
  },
  {
   "name": "Who Daniel is: OpenAI, forecasting, and disillusionment",
   "timestamp": "10:04",
   "summary": "He describes running the AI Futures Project, his 2022-2024 OpenAI tenure doing forecasting and dangerous-capability evals, and coming to see the labs' founding safety narratives as rationalizations for power-seeking."
  },
  {
   "name": "The $2 million anti-disparagement clause",
   "timestamp": "17:56",
   "summary": "Kokotajlo recounts refusing to sign OpenAI's exit non-disparagement clause, risking ~80% of his net worth, the public scandal that followed, and his disbelief that Sam Altman was unaware of it."
  },
  {
   "name": "The company plan: automate coding, then research, then everything",
   "timestamp": "20:02",
   "summary": "He explains the labs' strategy of automating their own AI research to reach superintelligence before rivals, and the AI 2027 scenario's race and slowdown endings."
  },
  {
   "name": "How AI is built: neural nets and recursive self-improvement",
   "timestamp": "30:36",
   "summary": "Kokotajlo explains that modern AI is trained neural nets rather than lines of code, describes pre-training and reinforcement, ~10 trillion parameters, and AIs increasingly training themselves."
  },
  {
   "name": "70% chance, CEO psychology, and Anthropic vs OpenAI",
   "timestamp": "42:46",
   "summary": "He clarifies the 70% figure, argues CEOs have rationalized extinction risk as acceptable, and distinguishes Anthropic/Dario as more willing to take costly stances (the Department of War dispute)."
  },
  {
   "name": "Jobs: sudden displacement and which roles survive",
   "timestamp": "48:40",
   "summary": "Kokotajlo argues broad job loss will be sudden and late (post-superintelligence), that which jobs remain is a political not technical question, and that the 'new jobs will appear' argument fails when everything is automated."
  },
  {
   "name": "The plans: AI 2027, AI 2040 Plan A, and Plans B/C/D/S",
   "timestamp": "1:01:04",
   "summary": "He contrasts the default 'race' (Plan D), 'shut it down' (Plan S), and his recommended Plan A — 2029 regulation, inference-not-training pauses, total research transparency, reversibility of data centers, and international deals."
  },
  {
   "name": "Life after work: Citizen's Dividend, lie detectors, and the 2030s",
   "timestamp": "1:26:15",
   "summary": "Kokotajlo describes a proposed Citizen's Dividend (from ~$25k to ~$10M per person/year), the invention of working lie detectors, and how even a slowed transition radically transforms society through the 2030s."
  },
  {
   "name": "Money, power, Earth as a preserve, and the button",
   "timestamp": "1:36:23",
   "summary": "He argues job loss threatens political power as well as income, proposes preserving 99% of Earth while expanding off-planet, and agonizes over whether he'd press a permanent AI-shutdown button."
  },
  {
   "name": "What people can do, and the personal toll",
   "timestamp": "1:50:35",
   "summary": "Kokotajlo urges people to pay attention, talk to representatives, and vote on AI, then reflects on reconsidering having more children and his closing message that science-fiction-sounding outcomes can become real."
  }
 ],
 "entities": [
  {
   "name": "Daniel Kokotajlo",
   "type": "person",
   "salience": "central",
   "role_in_episode": "Guest; former OpenAI forecaster and whistleblower, runs the AI Futures Project, author of AI 2027 and AI 2040 Plan A — drives the entire AI-risk discussion."
  },
  {
   "name": "Steven Bartlett",
   "type": "person",
   "salience": "central",
   "role_in_episode": "Host of The Diary Of A CEO; interviews Kokotajlo and probes jobs, timelines, and the moral dilemmas of AI."
  },
  {
   "name": "OpenAI",
   "type": "company",
   "salience": "central",
   "role_in_episode": "Kokotajlo's former employer; central case study for power-seeking incentives, the anti-disparagement scandal, and the race to superintelligence."
  },
  {
   "name": "Anthropic",
   "type": "company",
   "salience": "central",
   "role_in_episode": "Framed as having moved from second to first in the race; cited for revenue growth, Dario's stances, and the Department of War dispute."
  },
  {
   "name": "superintelligence",
   "type": "concept",
   "salience": "central",
   "role_in_episode": "The episode's central object: AI better than the best humans at everything, faster and cheaper; the thing the labs are racing to build."
  },
  {
   "name": "AI 2027",
   "type": "product",
   "salience": "central",
   "role_in_episode": "Kokotajlo's April 2025 month-by-month scenario forecast of a default AI race; its 'race' and 'slowdown' endings anchor much of the discussion."
  },
  {
   "name": "AI 2040 Plan A",
   "type": "product",
   "salience": "central",
   "role_in_episode": "The AI Futures Project's recommended plan: regulate in 2029, slow and transparent development, superintelligence delayed to 2040."
  },
  {
   "name": "jobs",
   "type": "concept",
   "salience": "central",
   "role_in_episode": "Recurring throughline: whether/when AI causes mass unemployment, which jobs survive, and the political consequences of job loss."
  },
  {
   "name": "Sam Altman",
   "type": "person",
   "salience": "discussed",
   "role_in_episode": "OpenAI CEO; discussed for power-seeking incentives, the equity clause he claimed not to know about, and racing dynamics."
  },
  {
   "name": "Dario Amodei",
   "type": "person",
   "salience": "discussed",
   "role_in_episode": "Anthropic CEO; coined 'country of geniuses in the data center' and framed as more willing to take costly safety stances."
  },
  {
   "name": "Elon Musk",
   "type": "person",
   "salience": "discussed",
   "role_in_episode": "Cited as a racing CEO, source of the Musk-OpenAI lawsuit emails, and proponent of an 'age of abundance' where work becomes optional."
  },
  {
   "name": "Ilya Sutskever",
   "type": "person",
   "salience": "discussed",
   "role_in_episode": "Former OpenAI head of research who founded Safe Super Intelligence; recounted via an all-hands anecdote and as another conflicted insider."
  },
  {
   "name": "AI Futures Project",
   "type": "organization",
   "salience": "discussed",
   "role_in_episode": "Kokotajlo's nonprofit that produced AI 2027 and AI 2040 Plan A; focuses on forecasting the future of AI."
  },
  {
   "name": "AGI",
   "type": "concept",
   "salience": "discussed",
   "role_in_episode": "Contrasted with superintelligence as a vaguer, weaker term that arguably has already been achieved."
  },
  {
   "name": "recursive self-improvement",
   "type": "concept",
   "salience": "discussed",
   "role_in_episode": "The intelligence-explosion dynamic Kokotajlo says makes job displacement sudden and drives the fast-takeoff scenarios."
  },
  {
   "name": "AI alignment",
   "type": "concept",
   "salience": "discussed",
   "role_in_episode": "The unsolved problem of making superintelligent AIs robustly hold the values we want; central to why loss of control is likely."
  },
  {
   "name": "interpretability",
   "type": "concept",
   "salience": "discussed",
   "role_in_episode": "Mechanistic interpretability framed as a note of optimism — if we could see inside the nets, loss-of-control risk drops sharply."
  },
  {
   "name": "neural networks",
   "type": "concept",
   "salience": "discussed",
   "role_in_episode": "Explained at length: AIs are trained neural nets, not lines of code, whose internal decision-making we can't directly read."
  },
  {
   "name": "Citizen's Dividend",
   "type": "concept",
   "salience": "discussed",
   "role_in_episode": "Plan A proposal giving citizens shares in a permit-selling agency, paying from ~$25k to ~$10M per person/year."
  },
  {
   "name": "China",
   "type": "place",
   "salience": "discussed",
   "role_in_episode": "The geopolitical rival driving race dynamics and the need for international inspection/treaties in Plan A."
  },
  {
   "name": "Department of War",
   "type": "organization",
   "salience": "discussed",
   "role_in_episode": "Its dispute with Anthropic (over domestic surveillance and autonomous robots) is cited as a costly stance and a foreshadowing of AI-control fights."
  },
  {
   "name": "Google DeepMind",
   "type": "organization",
   "salience": "mentioned",
   "role_in_episode": "One of the three labs with a 'risks are real but we'll handle it' founding narrative; the rival OpenAI's founders feared."
  },
  {
   "name": "Demis Hassabis",
   "type": "person",
   "salience": "mentioned",
   "role_in_episode": "Named in surfaced OpenAI emails as the Google figure the founders feared would 'become dictator with AGI.'"
  },
  {
   "name": "Geoffrey Hinton",
   "type": "person",
   "salience": "mentioned",
   "role_in_episode": "Quoted (via Steven) for the point that no less-intelligent species controls a more-intelligent one."
  },
  {
   "name": "JD Vance",
   "type": "person",
   "salience": "mentioned",
   "role_in_episode": "US Vice President cited as a reader of Kokotajlo's 2021 forecast 'What 2026 Looks Like.'"
  },
  {
   "name": "Mythos",
   "type": "product",
   "salience": "mentioned",
   "role_in_episode": "An Anthropic Claude model referenced re export controls and the government's push to shut it down over cyberattack risk."
  },
  {
   "name": "AI regulation",
   "type": "concept",
   "salience": "discussed",
   "role_in_episode": "The lever Kokotajlo argues can change race incentives; central to Plan A and to what he asks the public to demand."
  },
  {
   "name": "lie detectors",
   "type": "concept",
   "salience": "mentioned",
   "role_in_episode": "Given as an example of a technology advanced AI might actually make work — good used on the powerful, dangerous used by them."
  }
 ],
 "claims": [
  {
   "holder": "Daniel Kokotajlo",
   "claim": "There is roughly a 70% chance that AI development goes horribly wrong — something like AI takeover or a comparable catastrophe, of which human extinction is one possibility.",
   "stance": "opinion",
   "conviction": 0.7,
   "subjects": [
    "superintelligence"
   ],
   "timestamp": "42:46",
   "quote": "70 percent chance that this goes horribly wrong like human extinction"
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "OpenAI and Anthropic are deliberately trying to automate their own AI research so they no longer need human employees, in order to reach superintelligence before rivals.",
   "stance": "assertion",
   "conviction": 0.85,
   "subjects": [
    "OpenAI",
    "Anthropic"
   ],
   "timestamp": "20:52",
   "quote": "trying to automate themselves. They're trying to make it the case that they don't really need human employees anymore."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "The dominant motive at the top of these companies is power-seeking, not merely commercial, incentives.",
   "stance": "opinion",
   "conviction": 0.8,
   "subjects": [
    "OpenAI",
    "Sam Altman"
   ],
   "timestamp": "12:33",
   "quote": "I would describe it as power seeking incentives"
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "The 'doomerism' counter-narrative is false and is being pushed by people who stand to benefit from dismissing AI risk.",
   "stance": "opinion",
   "conviction": 0.8,
   "subjects": [
    "superintelligence"
   ],
   "timestamp": "9:35",
   "quote": "This counter-narrative is fairly recent and it's been pushed by the people who stand to benefit from it, and it's not true."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "The labs' founding safety narratives became rationalizations to justify racing rather than genuine guides to their behavior.",
   "stance": "opinion",
   "conviction": 0.8,
   "subjects": [
    "OpenAI",
    "Anthropic",
    "Google DeepMind"
   ],
   "timestamp": "11:01",
   "quote": "I increasingly came to think that these were rationalizations to justify what they were doing, rather than sort of like deeply guiding their actual behavior."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "You should judge AI leaders like Sam Altman by their actions, not by their shifting public narratives.",
   "stance": "opinion",
   "conviction": 0.75,
   "subjects": [
    "Sam Altman"
   ],
   "timestamp": "13:48",
   "quote": "I think you should judge people by their actions, not by their words."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "He does not believe Sam Altman was genuinely unaware of the equity/anti-disparagement clause; at minimum people close to him knew.",
   "stance": "opinion",
   "conviction": 0.7,
   "subjects": [
    "Sam Altman"
   ],
   "timestamp": "19:49",
   "quote": "I think you probably know. And if he didn't know, then people close to him probably did, such as his head lawyer."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "Anthropic overtook OpenAI in the race mainly through higher talent density and better strategy rather than more resources.",
   "stance": "opinion",
   "conviction": 0.6,
   "subjects": [
    "Anthropic"
   ],
   "timestamp": "39:19",
   "quote": "I guess they have probably higher talent density and better strategy, but not by a lot, but enough to make the difference."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "Anthropic and Dario have been more willing than competitors to say and do things costly to their bottom line and power, such as the Department of War dispute.",
   "stance": "opinion",
   "conviction": 0.7,
   "subjects": [
    "Anthropic",
    "Dario Amodei"
   ],
   "timestamp": "44:51",
   "quote": "it seems like Anthropic and Dario have been more willing to say and do things that are costly to the bottom line at least in the last year or so."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "None of the AI-lab CEOs should be trusted with the amount of power superintelligence would confer.",
   "stance": "opinion",
   "conviction": 0.85,
   "subjects": [
    "Sam Altman",
    "Dario Amodei",
    "Elon Musk"
   ],
   "timestamp": "45:15",
   "quote": "None of these people should be trusted with that much power, basically."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "The CEOs privately grant a real chance of extinction but have convinced themselves it will probably be fine and that they should be the one in charge.",
   "stance": "opinion",
   "conviction": 0.75,
   "subjects": [
    "Sam Altman",
    "Dario Amodei",
    "Elon Musk"
   ],
   "timestamp": "43:29",
   "quote": "They've all probably convinced themselves that, oh yeah, maybe it'll go horribly wrong, but probably it's going to be fine and probably I should be the one in charge."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "These CEOs are racing as hard as they can because each fears that if a rival reaches powerful AI first, that rival could become dictator.",
   "stance": "assertion",
   "conviction": 0.75,
   "subjects": [
    "Sam Altman",
    "Dario Amodei",
    "Elon Musk"
   ],
   "timestamp": "13:00",
   "quote": "these powerful CEOs are literally afraid that if the other guy gets there first, he might become dictator."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "Even if control is retained, superintelligence risks handing a tiny group of people the power of oligarchs or dictators.",
   "stance": "opinion",
   "conviction": 0.75,
   "subjects": [
    "superintelligence"
   ],
   "timestamp": "8:00",
   "quote": "we could very easily end up in a sort of a situation where some tiny group of people are essentially oligarchs or dictators."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "Dario's 'country of geniuses in the data center' is better described as an 'army of geniuses' — copies of one model owned and commanded by the company.",
   "stance": "opinion",
   "conviction": 0.7,
   "subjects": [
    "Anthropic",
    "superintelligence"
   ],
   "timestamp": "8:00",
   "quote": "I think it would be more accurate to describe it as army of geniuses in the data center"
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "The default trajectory — building AIs smarter than us, giving them bodies, and putting them in charge — is like choosing a new species that will outcompete us once it no longer needs us.",
   "stance": "opinion",
   "conviction": 0.75,
   "subjects": [
    "superintelligence"
   ],
   "timestamp": "1:11:14",
   "quote": "isn't this just exactly like us picking a new species that's then going to outcompute us when it doesn't need us anymore?"
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "Modern AI systems are not conventional software (lines of code) but trained neural nets, which is why we can't directly read their decision-making.",
   "stance": "assertion",
   "conviction": 0.9,
   "subjects": [
    "neural networks"
   ],
   "timestamp": "31:05",
   "quote": "modern AI systems are not software in the normal sense. I mean, they are technically software, but they're not lines of code."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "There is nothing magical about the human brain, so a digital system can perform its functions and superintelligence is achievable.",
   "stance": "opinion",
   "conviction": 0.8,
   "subjects": [
    "superintelligence",
    "neural networks"
   ],
   "timestamp": "28:20",
   "quote": "I think that there's nothing magical about the human brain."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "If interpretability research succeeds and we could see what AIs are thinking, loss-of-control scenarios become much less likely.",
   "stance": "opinion",
   "conviction": 0.7,
   "subjects": [
    "interpretability"
   ],
   "timestamp": "53:33",
   "quote": "I think that it would be much less likely for us to get into those loss of control scenarios if we could just actually see what our AIs were thinking and why and how at any given time, right?"
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "Broad job displacement will arrive suddenly rather than gradually, because of intelligence-explosion / recursive-self-improvement dynamics.",
   "stance": "opinion",
   "conviction": 0.7,
   "subjects": [
    "jobs",
    "recursive self-improvement"
   ],
   "timestamp": "49:41",
   "quote": "I think it'll be sudden because of the intelligence explosion dynamics, or recursive self-improvement dynamics."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "Once superintelligence is reached, which jobs remain is a political question, not a technical one, since technically all jobs could be automated.",
   "stance": "opinion",
   "conviction": 0.8,
   "subjects": [
    "jobs"
   ],
   "timestamp": "52:00",
   "quote": "I think that's actually a political question, not a technical question."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "The 'AI will create many new unforeseeable jobs' argument fails, because past automation was narrow while superintelligence automates everything.",
   "stance": "opinion",
   "conviction": 0.8,
   "subjects": [
    "jobs"
   ],
   "timestamp": "56:58",
   "quote": "The problem with that is that past technological advancements have been more narrow."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "Waiting to regulate until most people have lost their jobs is already too late, because the companies will likely already have superintelligence by then.",
   "stance": "opinion",
   "conviction": 0.85,
   "subjects": [
    "AI regulation",
    "jobs"
   ],
   "timestamp": "1:03:16",
   "quote": "if you wait until most people have lost their jobs to regulate the AI companies, that's already too late, because they will probably already have super intelligent AI by then"
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "He recommends Plan A (domestic regulation plus an international deal to continue AI more safely), though he is sympathetic to Plan S (shut it all down).",
   "stance": "opinion",
   "conviction": 0.7,
   "subjects": [
    "AI 2040 Plan A"
   ],
   "timestamp": "1:07:36",
   "quote": "Plan A is our recommendation. I think that I'm sympathetic to Plan S, but for reasons we explain, we recommend Plan A instead."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "The most probable default outcome is Plan D — the companies keep racing with little meaningful slowdown, as depicted in AI 2027.",
   "stance": "prediction",
   "conviction": 0.55,
   "subjects": [
    "AI 2027"
   ],
   "timestamp": "1:07:49",
   "quote": "Yeah, 2027 type of thing where they keep racing, they don't really slow down significantly."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "Total research transparency (publishing training recipes) matters because you don't want biased companies deciding whether their own AIs are safe.",
   "stance": "opinion",
   "conviction": 0.75,
   "subjects": [
    "AI 2040 Plan A",
    "AI alignment"
   ],
   "timestamp": "1:21:00",
   "quote": "open science is really important for solving the alignment problem fast enough because you don't want to have biased companies making the decisions about whether the AIs are safe."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "Working lie detectors are good when used on the powerful and dangerous (enabling new totalitarianism) when used by the powerful.",
   "stance": "opinion",
   "conviction": 0.6,
   "subjects": [
    "lie detectors"
   ],
   "timestamp": "1:31:04",
   "quote": "Broadly speaking, I would say the good uses are when lie detectors are used on the powerful instead of by the powerful."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "Losing jobs threatens not just income but political power, because economic contribution is a source of leverage over governments.",
   "stance": "opinion",
   "conviction": 0.75,
   "subjects": [
    "jobs"
   ],
   "timestamp": "1:35:41",
   "quote": "when people lose their jobs, they're not just threatened with loss of income, they're also threatened with loss of political power."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "In a post-superintelligence world, most of Earth should be preserved while economic expansion is pushed into special economic zones and off-planet.",
   "stance": "opinion",
   "conviction": 0.6,
   "subjects": [
    "superintelligence"
   ],
   "timestamp": "1:42:30",
   "quote": "you preserve like 99 percent of the Earth, mostly as is, as historic or environmental reasons."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "If humanity never builds powerful AI, civilization will probably still die eventually (nuclear war, pandemic), which is why he is torn on permanently shutting it down.",
   "stance": "opinion",
   "conviction": 0.5,
   "subjects": [
    "superintelligence"
   ],
   "timestamp": "1:49:00",
   "quote": "if we don't build powerful AI systems eventually, then we're probably going to die as a civilization eventually"
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "He would immediately press a button for a temporary AI shutdown because civilization is not ready, but would agonize over a permanent one.",
   "stance": "opinion",
   "conviction": 0.6,
   "subjects": [
    "AI regulation"
   ],
   "timestamp": "1:47:24",
   "quote": "I would totally slam that button because we are not ready to do this."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "The core problem is that people aren't taking AI risk seriously yet; broad public attention would itself produce better, more precise regulation.",
   "stance": "opinion",
   "conviction": 0.8,
   "subjects": [
    "AI regulation"
   ],
   "timestamp": "1:52:00",
   "quote": "the core problem is that people aren't taking it seriously yet."
  },
  {
   "holder": "Daniel Kokotajlo",
   "claim": "AI is the most important thing happening in our lifetimes, probably in all of history.",
   "stance": "opinion",
   "conviction": 0.85,
   "subjects": [
    "superintelligence"
   ],
   "timestamp": "1:52:23",
   "quote": "This is the most important thing happening in our lifetimes, probably in all of history, in fact, and it's very important that it go well."
  },
  {
   "holder": "OpenAI founders",
   "reported_by": "Daniel Kokotajlo",
   "claim": "OpenAI was founded in part because its founders feared Demis Hassabis at Google would become dictator with AGI.",
   "stance": "assertion",
   "conviction": 0.7,
   "subjects": [
    "OpenAI",
    "Demis Hassabis",
    "Google DeepMind"
   ],
   "timestamp": "12:33",
   "quote": "the reason why we made OpenAI was because we were worried that Demisosavis at Google was going to become dictator with AGI."
  },
  {
   "holder": "Geoffrey Hinton",
   "reported_by": "Steven Bartlett",
   "claim": "There is no example in nature where a more intelligent species has less control than a less intelligent one — so assuming we'll command superintelligence is arrogant.",
   "stance": "assertion",
   "conviction": 0.8,
   "subjects": [
    "superintelligence",
    "Geoffrey Hinton"
   ],
   "timestamp": "1:10:37",
   "quote": "there's no example in nature where a more intelligent species has less control than a less intelligent species."
  },
  {
   "holder": "Elon Musk",
   "reported_by": "Steven Bartlett",
   "claim": "AI is bringing an 'age of abundance' in which working becomes a choice rather than a necessity.",
   "stance": "prediction",
   "conviction": 0.6,
   "subjects": [
    "Elon Musk",
    "jobs"
   ],
   "timestamp": "1:05:24",
   "quote": "when Elon says that working will be a choice"
  }
 ],
 "predictions": [
  {
   "holder": "Daniel Kokotajlo",
   "prediction": "Superintelligence (cognitive, then physical) will probably arrive by the end of the decade, with his median around 2029-2030.",
   "timeframe": "by the end of the decade (~2030)",
   "subjects": [
    "superintelligence"
   ],
   "timestamp": "3:58",
   "quote": "seems to me like it's probably happening by the end of the decade.",
   "due": "2030-12-31"
  },
  {
   "holder": "Daniel Kokotajlo",
   "prediction": "Anthropic is on track to be 'the entire economy' by 2030 or so, even if its ~60x annual growth slows considerably.",
   "timeframe": "by 2030",
   "subjects": [
    "Anthropic"
   ],
   "timestamp": "4:30",
   "quote": "they're still on track to be the entire economy by 2030 or so.",
   "due": "2030-12-31"
  },
  {
   "holder": "Daniel Kokotajlo",
   "prediction": "The AI transition will probably be over before his young children reach working age, so he doesn't think they'll ever join the workforce.",
   "timeframe": "before his children reach working age (~2038)",
   "subjects": [
    "superintelligence",
    "jobs"
   ],
   "timestamp": "1:08:30",
   "quote": "I think that one way or another, this will probably all be over by the time they're old enough to join the workforce. So I don't think they'll ever join the workforce.",
   "due": "2038-01-01"
  },
  {
   "holder": "Daniel Kokotajlo",
   "prediction": "AI will be maybe the most important issue in the 2028 US presidential election.",
   "timeframe": "2028 US election",
   "subjects": [
    "AI regulation"
   ],
   "timestamp": "1:16:34",
   "quote": "We think it will be maybe the most important issue in the presidential election in 2028",
   "due": "2028-11-07"
  },
  {
   "holder": "Daniel Kokotajlo",
   "prediction": "In the AI 2027 scenario, mass unemployment does not arrive until 2028 or 2029, after superintelligence already exists.",
   "timeframe": "2028-2029",
   "subjects": [
    "jobs"
   ],
   "timestamp": "58:24",
   "quote": "the mass unemployment doesn't happen until 2028 or 2029 after they already have superintelligence.",
   "due": "2029-12-31"
  },
  {
   "holder": "Daniel Kokotajlo",
   "prediction": "By default, AI progress will likely happen faster than even the (already-fast) AI 2040 scenario depicts.",
   "timeframe": "unspecified",
   "subjects": [
    "superintelligence"
   ],
   "timestamp": "1:29:05",
   "quote": "Unfortunately, I actually think that things will happen faster than this by default."
  }
 ],
 "facts": [
  {
   "statement": "Anthropic's annual revenue grew from ~$1 billion a year to ~$60 billion a year in about one year — roughly 60x growth.",
   "subjects": [
    "Anthropic"
   ],
   "value": "$1B → $60B/year (~60x in one year)",
   "timestamp": "4:30",
   "quote": "now they're making something like $60 billion a year. So that's 60x growth in one year",
   "source": "Daniel Kokotajlo"
  },
  {
   "statement": "Refusing OpenAI's exit non-disparagement clause put ~$2M in equity — about 80% of his net worth at the time — at risk before the company reversed course.",
   "subjects": [
    "Daniel Kokotajlo",
    "OpenAI"
   ],
   "value": "~$2M (~80% of net worth)",
   "timestamp": "19:05",
   "quote": "Which was like 80% of our net worth at the time.",
   "source": "Daniel Kokotajlo"
  },
  {
   "statement": "Kokotajlo joined OpenAI in 2022, working on forecasting, dangerous-capability evaluations, and briefly reinforcement-learning agents.",
   "subjects": [
    "Daniel Kokotajlo",
    "OpenAI"
   ],
   "value": "joined 2022",
   "timestamp": "11:01",
   "quote": "I went to OpenAI in 2022",
   "source": "Daniel Kokotajlo"
  },
  {
   "statement": "Kokotajlo resigned from OpenAI in 2024.",
   "subjects": [
    "Daniel Kokotajlo",
    "OpenAI"
   ],
   "value": "resigned 2024",
   "timestamp": "15:29",
   "quote": "I resigned in 2024, had a nice goodbye party.",
   "source": "Daniel Kokotajlo"
  },
  {
   "statement": "The biggest AI models are around 10 trillion parameters today, up from ~175 billion in 2020 — roughly two orders of magnitude in six years.",
   "subjects": [
    "neural networks"
   ],
   "value": "~10 trillion params (2026) vs ~175 billion (2020)",
   "timestamp": "35:00",
   "quote": "Right now, they're at something like 10 trillion parameters. Back in 2020, it was more like 175 billion.",
   "source": "Daniel Kokotajlo"
  },
  {
   "statement": "AI 2027, published in April 2025, is a scenario forecast that lays out month by month a possible future AI trajectory.",
   "subjects": [
    "AI 2027"
   ],
   "value": "published April 2025",
   "timestamp": "22:59",
   "quote": "It's a scenario forecast that lays out month by month a possible future trajectory.",
   "source": "Daniel Kokotajlo"
  },
  {
   "statement": "Kokotajlo's 2021 forecast 'What 2026 Looks Like' was read by US Vice President JD Vance.",
   "subjects": [
    "Daniel Kokotajlo",
    "JD Vance"
   ],
   "value": "written 2021",
   "timestamp": "24:33",
   "quote": "Which one was it that JD Vance, the Vice President read?",
   "source": "Steven Bartlett"
  },
  {
   "statement": "AI 2027 blew up far bigger than the team's own forecasts predicted — a 90th-percentile outcome for views.",
   "subjects": [
    "AI 2027"
   ],
   "value": "90th-percentile outcome",
   "timestamp": "1:43:50",
   "quote": "And it was 90th percentile outcome.",
   "source": "Daniel Kokotajlo"
  },
  {
   "statement": "The Plan A 'Citizen's Dividend' proposal starts at roughly $25,000 per person per year and grows to roughly $10 million per citizen per year (inflation-adjusted).",
   "subjects": [
    "Citizen's Dividend",
    "AI 2040 Plan A"
   ],
   "value": "~$25,000 → ~$10,000,000 per person/year",
   "timestamp": "1:26:52",
   "quote": "It starts off something like $25,000 per person. And then by the end, it's something like $10 million per citizen.",
   "source": "Daniel Kokotajlo"
  },
  {
   "statement": "Current unemployment cited as about 5% in the UK and 4.2% in the US.",
   "subjects": [
    "jobs"
   ],
   "value": "UK ~5%, US ~4.2%",
   "timestamp": "58:00",
   "quote": "We're at about 5% unemployment. The US is at 4.2% unemployment.",
   "source": "Steven Bartlett"
  },
  {
   "statement": "The US government imposed export controls on Anthropic's 'Mythos' model and threatened Anthropic under the Defense Production Act — faster and more aggressive involvement than AI 2027 predicted.",
   "subjects": [
    "Anthropic",
    "Mythos"
   ],
   "value": "export controls + Defense Production Act threat",
   "timestamp": "38:33",
   "quote": "the export controls on Mithos being the biggest example and also threatening Anthropic with being destroyed by the Defense Production Act.",
   "source": "Daniel Kokotajlo"
  },
  {
   "statement": "The US government told Anthropic to shut down an AI model over concern that bad actors could use it for cyberattacks.",
   "subjects": [
    "Anthropic"
   ],
   "value": "shutdown over cyberattack risk",
   "timestamp": "1:45:54",
   "quote": "now the US government just told Anthropic that they have to shut down their AI because they were worried that bad actors would use it for cyberattacks",
   "source": "Daniel Kokotajlo"
  }
 ],
 "notable_quotes": [
  {
   "speaker": "Daniel Kokotajlo",
   "timestamp": "0:59",
   "quote": "it's possible that we'll end up essentially trading a new species that ends up ruling the world",
   "why": "The cold-open thesis of the entire interview — the loss-of-control fear in one line."
  },
  {
   "speaker": "Daniel Kokotajlo",
   "timestamp": "2:34",
   "quote": "I would be incredibly happy if all my predictions turn out to be wrong.",
   "why": "Captures his reluctant-Cassandra posture — he hopes to be wrong about the risk he forecasts."
  },
  {
   "speaker": "Daniel Kokotajlo",
   "timestamp": "13:48",
   "quote": "don't pay attention to the narratives.",
   "why": "His core epistemic advice about reading AI leaders — judge actions, not words."
  },
  {
   "speaker": "Daniel Kokotajlo",
   "timestamp": "45:15",
   "quote": "None of these people should be trusted with that much power, basically.",
   "why": "Distilled statement of the concentration-of-power thesis, applied to every CEO in the race."
  },
  {
   "speaker": "Daniel Kokotajlo",
   "timestamp": "1:53:30",
   "quote": "Because of what's happening with AI, I think a lot of those dreams are in jeopardy.",
   "why": "The personal turn — why his decision about having more children changed."
  },
  {
   "speaker": "Daniel Kokotajlo",
   "timestamp": "1:58:21",
   "quote": "sometimes things which sound like science fiction happen in reality.",
   "why": "His closing message to the public — stop dismissing the trajectory as fiction."
  }
 ],
 "extracted_by": "claude-opus-4-8",
 "prompt_version": "episode-extract-v1",
 "_file": "the-diary-of-a-ceo-2026-07-13-openai-whistleblower-finally-speaks-ai-has-a-70-chance-of-go.json"
}