recording_1.json

[
  {
    "speaker": "Speaker 1",
    "text": "Back in May this year, my longtime friend Alex reached out and asked me if I wanted to collaborate with him on a book about DeepSeek.",
    "timestamp": "00:00-00:08"
  },
  {
    "speaker": "Speaker 1",
    "text": "I would love to tell you the story of how I thought long and hard before getting back to him.",
    "timestamp": "00:08-00:12"
  },
  {
    "speaker": "Speaker 1",
    "text": "I just said yes",
    "timestamp": "00:12-00:14"
  },
  {
    "speaker": "Speaker 1",
    "text": "Now, six months later, I'm excited to announce that Deep Seeking Practice is out and available for pre-order.",
    "timestamp": "00:14-00:19"
  },
  {
    "speaker": "Speaker 1",
    "text": "The book is published by Packt and is a collaboration between Andy Pang, Alex Strick van Linschoten, and me",
    "timestamp": "00:19-00:26"
  },
  {
    "speaker": "Speaker 1",
    "text": "As the title suggests, this book focuses on DeepSeek models, which famously took the world by storm earlier this year.",
    "timestamp": "00:26-00:32"
  },
  {
    "speaker": "Speaker 1",
    "text": "It's organized into three parts.",
    "timestamp": "00:32-00:35"
  },
  {
    "speaker": "Speaker 1",
    "text": "Understanding and Exploring DeepSeek examines their role in the wider LLM world.",
    "timestamp": "00:35-00:40"
  },
  {
    "speaker": "Speaker 1",
    "text": "Using DeepSeek is entirely dedicated to applying DeepSeek models to real-world problems.",
    "timestamp": "00:39-00:45"
  },
  {
    "speaker": "Speaker 1",
    "text": "Finally, distilling and deploying DeepSeek covers distillation and deployment.",
    "timestamp": "00:45-00:50"
  },
  {
    "speaker": "Speaker 1",
    "text": "Even though the book was a joint effort, my main focus was on part two and how to use DeepSeek models to tackle real-world problems.",
    "timestamp": "00:50-00:57"
  },
  {
    "speaker": "Speaker 1",
    "text": "In particular, two chapters.",
    "timestamp": "00:57-00:59"
  },
  {
    "speaker": "Speaker 1",
    "text": "Chapter 5, Building with DeepSeek, walks through creating an alternative to Garmin's daily summary notifications.",
    "timestamp": "00:59-01:05"
  },
  {
    "speaker": "Speaker 1",
    "text": "We go through building a prototype using DeepSeek's API, how to leverage local models, frameworks like XGrammar and vLLM",
    "timestamp": "01:05-01:13"
  },
  {
    "speaker": "Speaker 1",
    "text": "all the way to deploying your own model using AWS large model inference or LMI.",
    "timestamp": "01:13-01:19"
  },
  {
    "speaker": "Speaker 1",
    "text": "Chapter 6, Agents with DeepSeek, is all about agents.",
    "timestamp": "01:19-01:22"
  },
  {
    "speaker": "Speaker 1",
    "text": "We start with a short intro to agents, tools, and the inner workings of the MCP protocol.",
    "timestamp": "01:22-01:28"
  },
  {
    "speaker": "Speaker 1",
    "text": "After that, we built, from scratch, three different agents powered by DeepSeek models.",
    "timestamp": "01:27-01:32"
  },
  {
    "speaker": "Speaker 1",
    "text": "An evaluator optimizer that summarizes arXiv papers, an orchestrated worker that generates research reports, and a tool calling agent that can search the web and answer complex questions.",
    "timestamp": "01:32-01:42"
  },
  {
    "speaker": "Speaker 1",
    "text": "Overall, I'm really excited about this book.",
    "timestamp": "01:42-01:45"
  },
  {
    "speaker": "Speaker 1",
    "text": "Finally gets to see the light of day.",
    "timestamp": "01:45-01:47"
  },
  {
    "speaker": "Speaker 1",
    "text": "Contributing to a book is no small feat.",
    "timestamp": "01:47-01:49"
  },
  {
    "speaker": "Speaker 1",
    "text": "The process is rough.",
    "timestamp": "01:49-01:50"
  },
  {
    "speaker": "Speaker 1",
    "text": "Lots of discussion and lots of drafts thrown in the trash.",
    "timestamp": "01:50-01:53"
  },
  {
    "speaker": "Speaker 1",
    "text": "It was hard work, but it was also a lot of fun.",
    "timestamp": "01:53-01:55"
  },
  {
    "speaker": "Speaker 1",
    "text": "I hope you enjoy it.",
    "timestamp": "01:55-01:57"
  }
]