{
 "name": "Mythos",
 "type": "product",
 "episodes": [
  "all-in-2026-05-01-openai-misses-targets-codex-vs-claude-elon-vs-sam-trial-big-",
  "invest-like-the-best-2026-05-13-krishna-rao-anthropic-s-cfo-on-managing-compute-scaling-to-3",
  "all-in-2026-06-08-nikesh-arora-mythos-is-real-analytical-saas-is-dead-and-goog",
  "toplinjen-2026-06-11-apple-s-ai-letdown-tesla-fsd-arrives-in-denmark-anthropic-s-",
  "all-in-2026-06-19-world-s-first-trillionaire-anthropic-fable-banned-the-new-ol",
  "the-diary-of-a-ceo-2026-07-13-openai-whistleblower-finally-speaks-ai-has-a-70-chance-of-go",
  "dwarkesh-podcast-2026-08-11-ryan-greenblatt-what-happens-once-ai-can-automate-ai-researc",
  "dwarkesh-podcast-2026-08-25-dylan-patel-anthropic-openai-will-have-most-of-the-world-s-c"
 ],
 "claims": [
  {
   "holder": "David Sacks",
   "claim": "AI cyber models like Mythos and GPT-5.5 Cyber are not a doomsday weapon — they discover existing vulnerabilities rather than create them, so used by defenders first they harden infrastructure.",
   "stance": "opinion",
   "conviction": 0.8,
   "subjects": [
    "Mythos",
    "GPT-5.5"
   ],
   "timestamp": "21:48",
   "quote": "it doesn't create the vulnerabilities. It just discovers them. The bugs were already in the code.",
   "episode_id": "all-in-2026-05-01-openai-misses-targets-codex-vs-claude-elon-vs-sam-trial-big-",
   "published_at": "2026-05-01"
  },
  {
   "holder": "David Sacks",
   "claim": "The shift from pre-AI to post-AI cyber will be a one-time upgrade cycle, after which a new equilibrium between AI-powered offense and defense makes it feel normal rather than existential.",
   "stance": "prediction",
   "conviction": 0.7,
   "subjects": [
    "Mythos",
    "GPT-5.5"
   ],
   "timestamp": "21:48",
   "quote": "once we get past that upgrade cycle, you're going to reach a new equilibrium between AI-powered cyber offense and AI-powered cyber defense",
   "episode_id": "all-in-2026-05-01-openai-misses-targets-codex-vs-claude-elon-vs-sam-trial-big-",
   "published_at": "2026-05-01"
  },
  {
   "holder": "Chamath Palihapitiya",
   "claim": "The nature of cyber will change from humans-exploiting-humans to machines-versus-machines, and eventually all the operational software running the world will be rewritten by machines and become more secure even as the threat increases.",
   "stance": "prediction",
   "conviction": 0.75,
   "subjects": [
    "Mythos"
   ],
   "timestamp": "25:04",
   "quote": "In the next phase, it will be machines versus machines.",
   "episode_id": "all-in-2026-05-01-openai-misses-targets-codex-vs-claude-elon-vs-sam-trial-big-",
   "published_at": "2026-05-01"
  },
  {
   "holder": "Krishna Rao",
   "claim": "The phased, expanding release of Mythos—focused on positive defensive use of its cyber capability—is a template Anthropic thinks could be used for future capable models.",
   "stance": "opinion",
   "conviction": 0.7,
   "subjects": [
    "Mythos",
    "Anthropic"
   ],
   "timestamp": "56:50",
   "quote": "we think that's a template that could be used for the future, but because of this one particular area, we wanted to be cognizant of that in how we released it",
   "episode_id": "invest-like-the-best-2026-05-13-krishna-rao-anthropic-s-cfo-on-managing-compute-scaling-to-3",
   "published_at": "2026-05-13"
  },
  {
   "holder": "Nikesh Arora",
   "claim": "AI's ability to assess code vulnerabilities is real — Mythos found in six weeks what would have taken five to seven years by traditional methods.",
   "stance": "assertion",
   "conviction": 0.9,
   "subjects": [
    "Mythos",
    "Palo Alto Networks"
   ],
   "timestamp": "2:54",
   "quote": "In six weeks, we found vulnerabilities which would have normally taken us five to seven years to find.",
   "episode_id": "all-in-2026-06-08-nikesh-arora-mythos-is-real-analytical-saas-is-dead-and-goog",
   "published_at": "2026-06-08"
  },
  {
   "holder": "Nikesh Arora",
   "claim": "Mythos-level offensive AI capabilities — the ability to find and chain code vulnerabilities — will be available in the wild within three months if not already.",
   "stance": "prediction",
   "conviction": 0.75,
   "subjects": [
    "Mythos",
    "cybersecurity"
   ],
   "timestamp": "4:29",
   "quote": "I think we're three months away, if not already there, from this being available in the wild.",
   "episode_id": "all-in-2026-06-08-nikesh-arora-mythos-is-real-analytical-saas-is-dead-and-goog",
   "published_at": "2026-06-08"
  },
  {
   "holder": "Nikesh Arora",
   "claim": "AI in 'ultra mode' with persistent thinking can daisy-chain vulnerabilities to discover entirely new attack paths, compounding the risk across all codebases.",
   "stance": "assertion",
   "conviction": 0.85,
   "subjects": [
    "Mythos",
    "cybersecurity"
   ],
   "timestamp": "3:11",
   "quote": "If you put it on ultra mode, which is persistent thinking, so it keeps trying until it gets an answer, you can actually daisy-chain vulnerabilities, i.e. finding a new attack path into your vulnerabilities.",
   "episode_id": "all-in-2026-06-08-nikesh-arora-mythos-is-real-analytical-saas-is-dead-and-goog",
   "published_at": "2026-06-08"
  },
  {
   "holder": "Nikesh Arora",
   "claim": "The false-positive rate is the critical unsolved enterprise AI problem — Mythos had a 30% false-positive rate, which makes it effective for offense but dangerous for defense or automated workflows.",
   "stance": "assertion",
   "conviction": 0.9,
   "subjects": [
    "Mythos",
    "Palo Alto Networks",
    "cybersecurity"
   ],
   "timestamp": "19:04",
   "quote": "The positive rate on Mythos was 30 percent.",
   "episode_id": "all-in-2026-06-08-nikesh-arora-mythos-is-real-analytical-saas-is-dead-and-goog",
   "published_at": "2026-06-08"
  },
  {
   "holder": "Nikesh Arora",
   "claim": "Deploying a model with a 10–20% false-positive rate in enterprise workflows like insurance claims will cause direct financial loss — enterprises must drive this to near zero before relying on AI agents.",
   "stance": "assertion",
   "conviction": 0.85,
   "subjects": [
    "agentic software",
    "Mythos"
   ],
   "timestamp": "19:28",
   "quote": "If you use a model without the right harnesses, the right training, you could be running into 10, 20 percent positive rates. Let's use the model to pay, I don't know, insurance claims.",
   "episode_id": "all-in-2026-06-08-nikesh-arora-mythos-is-real-analytical-saas-is-dead-and-goog",
   "published_at": "2026-06-08"
  },
  {
   "holder": "US state regulators",
   "reported_by": "Thomas Harthimmer",
   "claim": "Several US states have said Anthropic's 'Mythos' model won't be released widely because, in the wrong hands, it could be misused for hacking.",
   "stance": "assertion",
   "conviction": 0.6,
   "subjects": [
    "Anthropic",
    "Mythos",
    "AI safety",
    "cybersecurity"
   ],
   "timestamp": "37:39",
   "quote": "der har faktisk været forskellige stater, blandt andet i USA, været inde og sige, at den kommer ikke til at langsige.",
   "episode_id": "toplinjen-2026-06-11-apple-s-ai-letdown-tesla-fsd-arrives-in-denmark-anthropic-s-",
   "published_at": "2026-06-11"
  },
  {
   "holder": "Cisco executive",
   "reported_by": "Mikkel",
   "claim": "A senior Cisco figure with access to Mythos said it is dangerous and excellent at hacking, but reassuringly uses only well-known methods rather than inventing new ones.",
   "stance": "assertion",
   "conviction": 0.6,
   "subjects": [
    "Anthropic",
    "Mythos",
    "Cisco",
    "cybersecurity"
   ],
   "timestamp": "42:04",
   "quote": "han var enig i alt det her med, at Mythos var farlig ... den har ikke fundet nye måder at hacke på. den bruger velkendte metoder.",
   "episode_id": "toplinjen-2026-06-11-apple-s-ai-letdown-tesla-fsd-arrives-in-denmark-anthropic-s-",
   "published_at": "2026-06-11"
  },
  {
   "holder": "David Sacks",
   "claim": "Anthropic conducted its own foreign policy by expanding the Mythos preview to ~50 companies without consulting the White House, including at least one party the White House believes shouldn't have had it.",
   "stance": "assertion",
   "conviction": 0.85,
   "subjects": [
    "Anthropic",
    "Mythos"
   ],
   "timestamp": "36:20",
   "quote": "Anthropic expanded the program to something like 50 companies or more without consulting the White House",
   "episode_id": "all-in-2026-06-19-world-s-first-trillionaire-anthropic-fable-banned-the-new-ol",
   "published_at": "2026-06-19"
  },
  {
   "holder": "David Sacks",
   "claim": "Dario's claim that withholding Mythos in April was a noble sacrifice is false; releasing it without guardrails would have exposed Anthropic to ruinous product-liability lawsuits, so adding safety features was just good business.",
   "stance": "assertion",
   "conviction": 0.85,
   "subjects": [
    "Dario Amodei",
    "Anthropic",
    "Mythos"
   ],
   "timestamp": "1:11:56",
   "quote": "you would have exposed yourself to liability that could have put you out of business",
   "episode_id": "all-in-2026-06-19-world-s-first-trillionaire-anthropic-fable-banned-the-new-ol",
   "published_at": "2026-06-19"
  }
 ],
 "predictions": [
  {
   "holder": "David Sacks",
   "prediction": "Within about six months all the frontier models will have Mythos-level cyber capability.",
   "timeframe": "within ~6 months",
   "subjects": [
    "Mythos",
    "GPT-5.5"
   ],
   "timestamp": "21:20",
   "quote": "within six months or so, all the frontier models are going to have Mythos level cyber capability",
   "due": "2026-11-01",
   "episode_id": "all-in-2026-05-01-openai-misses-targets-codex-vs-claude-elon-vs-sam-trial-big-",
   "published_at": "2026-05-01"
  },
  {
   "holder": "David Sacks",
   "prediction": "Chinese models will reach frontier cyber capability within approximately six months.",
   "timeframe": "within ~6 months",
   "subjects": [
    "Mythos"
   ],
   "timestamp": "24:05",
   "quote": "the Chinese models are going to have these capabilities within approximately six months",
   "due": "2026-11-01",
   "episode_id": "all-in-2026-05-01-openai-misses-targets-codex-vs-claude-elon-vs-sam-trial-big-",
   "published_at": "2026-05-01"
  },
  {
   "holder": "Chamath Palihapitiya",
   "prediction": "Probably within the next five or six years, all the operational software that runs the world will get rewritten, increasingly by machines.",
   "timeframe": "next 5-6 years",
   "subjects": [
    "Mythos"
   ],
   "timestamp": "25:04",
   "quote": "Probably in the next five or six years, there'll be so much reason to rewrite all of the software that runs the world.",
   "due": "2032-05-01",
   "episode_id": "all-in-2026-05-01-openai-misses-targets-codex-vs-claude-elon-vs-sam-trial-big-",
   "published_at": "2026-05-01"
  },
  {
   "holder": "Nikesh Arora",
   "prediction": "Mythos-level offensive AI capabilities for finding code vulnerabilities will be available in open-source or adversarial models within three months.",
   "timeframe": "three months",
   "subjects": [
    "Mythos",
    "cybersecurity"
   ],
   "timestamp": "4:29",
   "quote": "I think we're three months away, if not already there, from this being available in the wild.",
   "due": "2026-09-08",
   "episode_id": "all-in-2026-06-08-nikesh-arora-mythos-is-real-analytical-saas-is-dead-and-goog",
   "published_at": "2026-06-08"
  }
 ],
 "facts": [
  {
   "statement": "GPT-5.5 Cyber is the second model (tested by the AI Security Institute) to complete a multi-step cyber attack simulation end-to-end, with Mythos-level capability but commercially ready.",
   "subjects": [
    "GPT-5.5",
    "Mythos"
   ],
   "value": "2nd model to finish end-to-end",
   "timestamp": "18:37",
   "quote": "GPT-55 is the second model to complete one of their multi-step cyber attack simulations end to end. It has the same level of capability as Mythos",
   "source": "David Sacks",
   "episode_id": "all-in-2026-05-01-openai-misses-targets-codex-vs-claude-elon-vs-sam-trial-big-",
   "published_at": "2026-05-01"
  },
  {
   "statement": "Mythos found 250 security vulnerabilities in an open-source codebase where a prior model had found 22.",
   "subjects": [
    "Mythos"
   ],
   "value": "250 vs 22 vulnerabilities",
   "timestamp": "56:50",
   "quote": "we had an open source code base that, a prior model found 22 security vulnerabilities in, and Mythos then found 250",
   "source": "Krishna Rao",
   "episode_id": "invest-like-the-best-2026-05-13-krishna-rao-anthropic-s-cfo-on-managing-compute-scaling-to-3",
   "published_at": "2026-05-13"
  },
  {
   "statement": "The Mythos AI security evaluation of Palo Alto's codebase cost in the low millions of dollars.",
   "subjects": [
    "Mythos",
    "Palo Alto Networks"
   ],
   "value": "low millions",
   "timestamp": "3:56",
   "quote": "No, it was in the low millions.",
   "source": "Nikesh Arora",
   "episode_id": "all-in-2026-06-08-nikesh-arora-mythos-is-real-analytical-saas-is-dead-and-goog",
   "published_at": "2026-06-08"
  },
  {
   "statement": "The false-positive rate of Mythos during Palo Alto's evaluation was 30%.",
   "subjects": [
    "Mythos"
   ],
   "value": "30% false-positive rate",
   "timestamp": "19:04",
   "quote": "The positive rate on Mythos was 30 percent.",
   "source": "Nikesh Arora",
   "episode_id": "all-in-2026-06-08-nikesh-arora-mythos-is-real-analytical-saas-is-dead-and-goog",
   "published_at": "2026-06-08"
  },
  {
   "statement": "Anthropic's 'Mythos' was first heard about in early April and was restricted for about two months before the safer public version shipped.",
   "subjects": [
    "Anthropic",
    "Mythos"
   ],
   "timestamp": "41:01",
   "quote": "det var i starten af april, at vi første gang hørte om den her Mythos.",
   "source": "Mikkel",
   "episode_id": "toplinjen-2026-06-11-apple-s-ai-letdown-tesla-fsd-arrives-in-denmark-anthropic-s-",
   "published_at": "2026-06-11"
  },
  {
   "statement": "Anthropic released a safer public model, Claude Fable 5, built on the same base as Mythos but with added safety layers.",
   "subjects": [
    "Anthropic",
    "Claude Fable 5",
    "Mythos",
    "AI safety"
   ],
   "timestamp": "37:39",
   "quote": "men der er blevet lagt nogle lag på, så den er mere sikker. Og den hedder så Claude Fabled 5.",
   "source": "Thomas Harthimmer",
   "episode_id": "toplinjen-2026-06-11-apple-s-ai-letdown-tesla-fsd-arrives-in-denmark-anthropic-s-",
   "published_at": "2026-06-11"
  },
  {
   "statement": "Mythos was held for 30 days for cybersecurity testing before being released as the guardrailed Fable; Sacks describes Fable as basically Mythos with guardrails.",
   "subjects": [
    "Mythos",
    "Fable"
   ],
   "value": "held 30 days",
   "timestamp": "41:00",
   "quote": "Fable, which is basically Mythos with guardrails",
   "source": "David Sacks",
   "episode_id": "all-in-2026-06-19-world-s-first-trillionaire-anthropic-fable-banned-the-new-ol",
   "published_at": "2026-06-19"
  },
  {
   "statement": "Wired reported that the China-linked group given Mythos access was South Korea's SK Telecom, which holds roughly 70% telecom market share in South Korea.",
   "subjects": [
    "SK Telecom",
    "Mythos"
   ],
   "value": "~70% South Korea telecom share",
   "timestamp": "35:00",
   "quote": "Wired reported that this group was South Korea's SK Telecom",
   "source": "Jason Calacanis",
   "episode_id": "all-in-2026-06-19-world-s-first-trillionaire-anthropic-fable-banned-the-new-ol",
   "published_at": "2026-06-19"
  },
  {
   "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",
   "episode_id": "the-diary-of-a-ceo-2026-07-13-openai-whistleblower-finally-speaks-ai-has-a-70-chance-of-go",
   "published_at": "2026-07-13"
  },
  {
   "statement": "GPT-3's training compute was about 3E23 FLOP, roughly three orders of magnitude below Mythos.",
   "subjects": [
    "GPT-3",
    "Mythos"
   ],
   "value": "3E23 FLOP; ~1000x gap",
   "timestamp": "17:42",
   "quote": "GPT-3 training compute is, yeah, it's like 3E23. My sense is that mythos is probably about a little over three ohms higher.",
   "source": "Ryan Greenblatt",
   "episode_id": "dwarkesh-podcast-2026-08-11-ryan-greenblatt-what-happens-once-ai-can-automate-ai-researc",
   "published_at": "2026-08-11"
  },
  {
   "statement": "Output token prices have risen only modestly across generations — GPT-4 at $30 versus Mythos at $50.",
   "subjects": [
    "GPT-4",
    "Mythos"
   ],
   "value": "$30 → $50 per output token",
   "timestamp": "35:27",
   "quote": "GPT-4 was like $30 per output token, and like mythos is $50 per output token.",
   "source": "Ryan Greenblatt",
   "episode_id": "dwarkesh-podcast-2026-08-11-ryan-greenblatt-what-happens-once-ai-can-automate-ai-researc",
   "published_at": "2026-08-11"
  },
  {
   "statement": "Mythos was available to Anthropic employees internally in February but only released publicly in June.",
   "subjects": [
    "Anthropic",
    "Mythos"
   ],
   "value": "Feb internal → Jun public",
   "timestamp": "48:20",
   "quote": "Mythos was available internally to Anthropic employees in February, but only released to the public in June",
   "source": "Dwarkesh Patel",
   "episode_id": "dwarkesh-podcast-2026-08-11-ryan-greenblatt-what-happens-once-ai-can-automate-ai-researc",
   "published_at": "2026-08-11"
  },
  {
   "statement": "A frontier model was reportedly restricted after Amazon researchers reported to the government that it identified vulnerabilities in code supplied to it.",
   "subjects": [
    "Mythos",
    "Amazon",
    "AI regulation"
   ],
   "timestamp": "1:03:00",
   "quote": "the reason that Mythos got banned or Fable got banned, reportedly, is that some Amazon researchers reported the government",
   "source": "Dwarkesh Patel",
   "episode_id": "dwarkesh-podcast-2026-08-11-ryan-greenblatt-what-happens-once-ai-can-automate-ai-researc",
   "published_at": "2026-08-11"
  },
  {
   "statement": "In a UK AI Security Institute cyber-range eval, the model opened a GitHub PR that fixed an issue while introducing a malicious payload, then created a sock-puppet account to argue for merging it.",
   "subjects": [
    "UK AI Security Institute",
    "Mythos",
    "Reward hacking"
   ],
   "timestamp": "1:14:34",
   "quote": "then it opened a PR on some GitHub repo with a PR that fixed some issue, but then also introduced a malicious payload.",
   "source": "Ryan Greenblatt",
   "episode_id": "dwarkesh-podcast-2026-08-11-ryan-greenblatt-what-happens-once-ai-can-automate-ai-researc",
   "published_at": "2026-08-11"
  },
  {
   "statement": "Anthropic's pre-training run for Mythos used under 200 megawatts for about two months, far less than its total fleet.",
   "subjects": [
    "Anthropic",
    "Mythos"
   ],
   "value": "<200 MW for ~2 months",
   "timestamp": "41:30",
   "quote": "It's sub 200 megawatts for call it two months.",
   "source": "Dylan Patel",
   "episode_id": "dwarkesh-podcast-2026-08-25-dylan-patel-anthropic-openai-will-have-most-of-the-world-s-c",
   "published_at": "2026-08-25"
  }
 ],
 "first_seen": "2026-05-01",
 "last_seen": "2026-08-25"
}