{
  "version": "1.2.0",
  "chapters": [
    {
      "title": "Cold open\r",
      "startTime": 0,
      "endTime": 148
    },
    {
      "title": "Holden is back!\r",
      "startTime": 148,
      "endTime": 178
    },
    {
      "title": "An AI Chernobyl we never notice\r",
      "startTime": 178,
      "endTime": 459
    },
    {
      "title": "Is rogue AI takeover easy or hard?\r",
      "startTime": 459,
      "endTime": 1081
    },
    {
      "title": "The AGI race isn't a coordination failure\r",
      "startTime": 1081,
      "endTime": 1710
    },
    {
      "title": "What Holden now does at Anthropic\r",
      "startTime": 1710,
      "endTime": 1838
    },
    {
      "title": "The case for working at Anthropic\r",
      "startTime": 1838,
      "endTime": 2490
    },
    {
      "title": "Is Anthropic doing enough?\r",
      "startTime": 2490,
      "endTime": 2670
    },
    {
      "title": "Can we trust Anthropic, or any AI company?\r",
      "startTime": 2670,
      "endTime": 3011
    },
    {
      "title": "How can Anthropic compete while paying the “safety tax”?\r",
      "startTime": 3011,
      "endTime": 3433
    },
    {
      "title": "What, if anything, could prompt Anthropic to halt development of AGI?\r",
      "startTime": 3433,
      "endTime": 3604
    },
    {
      "title": "Holden's retrospective on responsible scaling policies\r",
      "startTime": 3604,
      "endTime": 4545
    },
    {
      "title": "Overrated work\r",
      "startTime": 4545,
      "endTime": 4678
    },
    {
      "title": "Concrete shovel-ready projects Holden is excited about\r",
      "startTime": 4678,
      "endTime": 4932
    },
    {
      "title": "Great things to do in technical AI safety\r",
      "startTime": 4932,
      "endTime": 5393
    },
    {
      "title": "Great things to do on AI welfare and AI relationships\r",
      "startTime": 5393,
      "endTime": 5811
    },
    {
      "title": "Great things to do in biosecurity and pandemic preparedness\r",
      "startTime": 5811,
      "endTime": 5857
    },
    {
      "title": "How to choose where to work\r",
      "startTime": 5857,
      "endTime": 6218
    },
    {
      "title": "Overrated AI risk: Cyberattacks\r",
      "startTime": 6218,
      "endTime": 6808
    },
    {
      "title": "Overrated AI risk: Persuasion\r",
      "startTime": 6808,
      "endTime": 7051
    },
    {
      "title": "Why AI R&D is the main thing to worry about\r",
      "startTime": 7051,
      "endTime": 7770
    },
    {
      "title": "The case that AI-enabled R&D wouldn't speed things up much\r",
      "startTime": 7770,
      "endTime": 8006
    },
    {
      "title": "AI-enabled human power grabs\r",
      "startTime": 8006,
      "endTime": 8764
    },
    {
      "title": "Main benefits of getting AGI right\r",
      "startTime": 8764,
      "endTime": 9104
    },
    {
      "title": "The world is handling AGI about as badly as possible\r",
      "startTime": 9104,
      "endTime": 9258
    },
    {
      "title": "Learning from targeting companies for public criticism in farm animal welfare\r",
      "startTime": 9258,
      "endTime": 9823
    },
    {
      "title": "Will Anthropic actually make any difference?\r",
      "startTime": 9823,
      "endTime": 10703
    },
    {
      "title": "“Misaligned” vs “misaligned and power-seeking”\r",
      "startTime": 10703,
      "endTime": 11056
    },
    {
      "title": "Success without dignity: how we could win despite being stupid\r",
      "startTime": 11056,
      "endTime": 11520
    },
    {
      "title": "Holden sees less dignity but has more hope\r",
      "startTime": 11520,
      "endTime": 11983
    },
    {
      "title": "Should we expect misaligned power-seeking by default?\r",
      "startTime": 11983,
      "endTime": 12456
    },
    {
      "title": "Will reinforcement learning make everything worse?\r",
      "startTime": 12456,
      "endTime": 12774
    },
    {
      "title": "Should we push for marginal improvements or big paradigm shifts?\r",
      "startTime": 12774,
      "endTime": 12932
    },
    {
      "title": "Should safety-focused people cluster or spread out?\r",
      "startTime": 12932,
      "endTime": 13195
    },
    {
      "title": "Is Anthropic vocal enough about strong regulation?\r",
      "startTime": 13195,
      "endTime": 13410
    },
    {
      "title": "Is Holden biased because of his financial stake in Anthropic?\r",
      "startTime": 13410,
      "endTime": 13677
    },
    {
      "title": "Have we learned clever governance structures don't work?\r",
      "startTime": 13677,
      "endTime": 13820
    },
    {
      "title": "Is Holden scared of AI bioweapons?\r",
      "startTime": 13820,
      "endTime": 14038
    },
    {
      "title": "Holden thinks AI companions are bad news\r",
      "startTime": 14038,
      "endTime": 14453
    },
    {
      "title": "Are AI companies too hawkish on China?\r",
      "startTime": 14453,
      "endTime": 14706
    },
    {
      "title": "The frontier of infosec: confidentiality vs integrity\r",
      "startTime": 14706,
      "endTime": 14875
    },
    {
      "title": "How often does AI work backfire?\r",
      "startTime": 14875,
      "endTime": 15770
    },
    {
      "title": "Is AI clearly more impactful to work in than other causes?\r",
      "startTime": 15770,
      "endTime": 16162
    },
    {
      "title": "What's the role of earning to give?",
      "startTime": 16162,
      "endTime": 0
    }
  ]
}