{
  "video": {
    "id": "y2dyD6slKxE",
    "title": "Why AI Won't Replace Stock Pickers - Jack Woodruff, Candlestick & Vinoo Ganesh, Kepler",
    "duration": 1126,
    "upload_date": null,
    "channel": "AI Engineer",
    "source": "AI Engineer",
    "event_format": ""
  },
  "analysis": {
    "video_id": "y2dyD6slKxE",
    "title": "Why AI Won't Replace Stock Pickers - Jack Woodruff, Candlestick & Vinoo Ganesh, Kepler",
    "one_liner": "Former Citadel/Surveyor PM Jack Woodruff argues that AI turns today's analyst work into table stakes but won't replace stock pickers: alpha still comes from finding a divergent view, so the winners will use AI to decide which names deserve their limited time.",
    "summary": "In a fireside chat with Kepler's Vinoo Ganesh, Jack Woodruff (founder and former CIO of Candlestick Capital Management, previously a partner and PM at Surveyor Capital, a division of Citadel) says long-term stock picking won't change much. What will change is how information is put together and what counts as table stakes. He treats investment management as a time-management exercise. AI can shrink 6-8 hours of prep to 5 minutes, but competitors will spend the saved time on proprietary research, so the bar simply rises. He insists on being able to audit every number and expects the analyst pyramid to become flatter, more vertical teams. He thinks the 'no-analyst hedge fund' can only replicate last year's funds.",
    "key_points": [
      "Alpha still comes from the 'mosaic': three or four data points from different sources that line up behind a divergent view. That doesn't change with AI; how people put the information together does.",
      "Humans are still the bottleneck, because the information has to get into your head. Coverage has diminishing returns: the more stocks you cover, the less time each one gets.",
      "AI might cut 6-8 hours of prep to 5 minutes. In a competitive industry the freed time goes into proprietary research, channel checks and industry backdrop work, so everyone's workload goes up rather than down.",
      "Junior analysts will still read 10-Ks and transcripts, perhaps as summaries. Big-data wrangling can be outsourced. AI's real value is steering analysts toward the right names to spend time on.",
      "Being able to audit numbers matters most. As a former CPA, Woodruff would never sign off on a model's number without seeing where it came from. A single data-entry mistake could cost millions. People who use Claude to build models still have to audit every number, which raises the question of how much time is really saved.",
      "Coverage maths: cover 100 stocks with a 20-name book and you pick 20%; at 200 names it's 10%; at 400 it's 5%. More ideas should raise the hit rate, but only if the research on all 400 is as solid as it was on 100.",
      "Org structure shifts from a pyramid (many analysts → senior stock pickers → PMs → CIO) to something more vertical. He'd keep the 'makers' and need fewer 'preparers', and might hire investigative journalists, deep-project researchers and data engineers instead.",
      "The 'no-analyst hedge fund' run on AI agents will replicate a fund that existed a year ago but won't match what human-run funds are doing next year. If fundamental investing ever removes humans entirely, it becomes a contest of who has the most compute and information. Quant funds took humans out of the loop decades ago."
    ],
    "takeaways": [
      "Use AI to set priorities ('here's where you should spend your time'), not to produce buy/sell calls. Its output is a pointer to a special project, not the trade.",
      "Build finance AI tools so every number traces back to its source. Without that, users re-audit everything and lose the time they saved.",
      "Assume anything AI automates becomes table stakes. Put the freed time into proprietary research, channel checks and industry builds that others don't have.",
      "When selling AI software to hedge funds, expect data security and isolation questions to go to the CTO, and expect nervousness about internal data leaving the firm.",
      "Rethink hiring around fewer 'preparers' and more people who can feed proprietary information into the machine: investigative journalists, deep-project researchers, data engineers."
    ],
    "topics": [
      "finance",
      "hedge funds",
      "human-in-the-loop",
      "auditability",
      "ai adoption",
      "org design",
      "data security"
    ],
    "tools": [
      "Kepler",
      "Candlestick Capital Management",
      "Surveyor Capital",
      "Citadel",
      "Claude"
    ],
    "quotes": [
      {
        "text": "investment management is a time management exercise.",
        "at": "02:41",
        "url": "https://www.youtube.com/watch?v=y2dyD6slKxE&t=161s"
      },
      {
        "text": "investment management isn't about doing earnings previews. It's not about building DCFs. It's not about that. It's making money.",
        "at": "10:41",
        "url": "https://www.youtube.com/watch?v=y2dyD6slKxE&t=641s"
      },
      {
        "text": "What I do believe is that they will be able to replicate a fund that existed a year ago.",
        "at": "17:57",
        "url": "https://www.youtube.com/watch?v=y2dyD6slKxE&t=1077s"
      },
      {
        "text": "I don't think you're ever going to really be able to take humans out of the loop.",
        "at": "18:12",
        "url": "https://www.youtube.com/watch?v=y2dyD6slKxE&t=1092s"
      }
    ],
    "words": 4005
  },
  "summary_url": "/#y2dyD6slKxE",
  "transcript": {
    "html": "/transcripts/y2dyD6slKxE.html",
    "txt": "/transcripts/y2dyD6slKxE.txt",
    "vtt": "/transcripts/y2dyD6slKxE.vtt"
  }
}