00:00:00Claude code might be the most powerful research agent
00:00:03on the planet, but you need to add this one tool
00:00:07to unlock it.
00:00:08Now, most people's version of Claude code research
00:00:11is just telling it to use the web search tool
00:00:13and praying that whatever it comes back with is good enough.
00:00:17But we can do better because what if I told you
00:00:19with just five minutes of setup,
00:00:21we could create workflows inside of Claude code
00:00:24that are able to scrape any section of YouTube,
00:00:26pull their captions, push that information
00:00:28into a free, robust, pre-built rag system
00:00:32that is able to do all the heavy lifting
00:00:35and analysis for us, and then take that analysis
00:00:38and give us deliverables like slide decks,
00:00:40infographics, podcasts, you name it,
00:00:43all while costing us virtually zero tokens.
00:00:46Now, if that sounds way too good to be true,
00:00:48you would normally be right, but not in this case.
00:00:51So let me introduce you to the most underrated AI tool
00:00:55in the game today, Notebook LM.
00:00:58So in today's video, I'm gonna show you
00:01:00how to combine the powers of Claude code and Notebook LM
00:01:03to replace a research stack for free
00:01:06that would otherwise cost you hundreds of dollars a month
00:01:10to build and maintain.
00:01:11I'm really excited to show you guys this one.
00:01:14So let's hop into it.
00:01:15So let's kick this video off with a demo
00:01:16so you can see how we can use Claude code
00:01:19to harness all the features of Notebook LM
00:01:22without us ever leaving the terminal.
00:01:24Now this prompt is gonna have Claude code
00:01:26do a number of things.
00:01:27First, we are gonna use our custom YouTube search skill
00:01:30to find the latest trending videos on Claude code skills.
00:01:33And don't worry, I will show you all these skills
00:01:35and how to get them in a second.
00:01:37After we find the video URLs,
00:01:39I want Claude code to send those URLs over to Notebook LM
00:01:43using the Notebook LM skill.
00:01:44I then want Notebook LM to do analysis on those videos
00:01:49to figure out what are the top load skills.
00:01:51And I wanna get that analysis sent to us.
00:01:53Furthermore, I want a deliverable.
00:01:54I don't just want the text analysis.
00:01:56I want an infographic in a handwritten blueprint style
00:02:00depicting that analysis on the top skills.
00:02:03So with one prompt, we are going to scrape YouTube.
00:02:06We're going to source all of our data.
00:02:08We're going to essentially put it into a rag system
00:02:11because that's what Notebook LM is.
00:02:13We're gonna have Notebook LM do all the analysis
00:02:15and the deliverables for us offsite,
00:02:18which means we don't pay for that in tokens.
00:02:20And we get all that for free.
00:02:22So let's see how it works.
00:02:23So here's what we got.
00:02:24Claude code uploaded 20 YouTube sources
00:02:26into Notebook LM for analysis.
00:02:29Notebook LM then came back
00:02:30with the top five Claude code skills that you see here,
00:02:34as well as emerging trends for how they're used.
00:02:37It then created that infographic deliverable for us,
00:02:39which automatically populated inside of our project folder.
00:02:42So here's a look at that infographic.
00:02:44Again, this is nano banana pro under the hood.
00:02:47It's actually being called and the text here
00:02:49and all the visuals, A, fit the style that we called for,
00:02:52which was like a handwritten blueprint type style.
00:02:55And secondly, and more importantly,
00:02:57all this content is based on the videos
00:02:59and the analysis of those videos.
00:03:01It isn't just making stuff up.
00:03:02And we can also see here inside of Claude code,
00:03:04the videos it grabbed, the title, the creator, the views,
00:03:06the duration, and the date.
00:03:08And all this is reflected inside of Notebook LM itself.
00:03:10I can see all the sources that upload.
00:03:12I can see all of the analysis.
00:03:14I can see the blueprint guide that we asked for.
00:03:18And while this demo might seem
00:03:19like a rather simplistic application of this,
00:03:21I cannot stress enough the value add
00:03:24of combining these two tools,
00:03:26because this goes way beyond just automating
00:03:28the source process for Notebook LM.
00:03:30Everything we did here,
00:03:31we could have done manually inside of Notebook LM, right?
00:03:33I could have gone through YouTube manually.
00:03:35I could have found all the videos I want.
00:03:37I could have copy pasted them in.
00:03:38I could have gotten the analysis
00:03:39and I could have gotten a deliverable.
00:03:41The fact that we can automate that is nice,
00:03:43but it's more than that.
00:03:44It's the fact that I can bring all that analysis
00:03:47into my Claude code ecosystem effortlessly
00:03:50and the use cases of that workflow are pretty much infinite.
00:03:55And the second reason why this combination of tools
00:03:56is so powerful has everything to do
00:03:58with the straight up power of Notebook LM.
00:04:01If you try to recreate what Notebook LM does,
00:04:04which is like a scraping system into a rag system,
00:04:07into an analysis system, into a deliverable system, right?
00:04:11With the infographics, the slide deck, all of that,
00:04:13it would be a gigantic pain in the butt to do.
00:04:15As someone who has tried it,
00:04:16at least on the research side with things like N8N,
00:04:18it's not a simple process.
00:04:20Furthermore, it costs money and this whole thing is free,
00:04:23which is in large part why I'm so excited
00:04:24to share it with you.
00:04:25And another reason you should be pumped
00:04:27is because how simple it is to set this whole thing up,
00:04:30which is what we're going to cover now.
00:04:32So when it comes to setting this up,
00:04:33you're probably thinking, hey, Chase,
00:04:34how do we actually connect Notebook LM to Claude code
00:04:38considering the fact that Notebook LM
00:04:40does not have a public API?
00:04:41Well, luckily for us, there's people much smarter
00:04:43than you or I who have already solved this problem.
00:04:46In this case, it is Tang Ling
00:04:48and we will be standing on his shoulders today
00:04:50as we use the Notebook LM-PI GitHub repo
00:04:54to essentially act as an unofficial Python API
00:04:57for Notebook LM.
00:04:58But before we set up Notebook LM,
00:05:00we need to solve the first part of our pipeline,
00:05:03which is actually the YouTube search
00:05:04and the sourcing of the data we want Notebook LM
00:05:07to do analysis on.
00:05:09And for that, I have a custom YouTube search skill
00:05:12for Claude code that does all of this for you.
00:05:15It uses a Python script that relies on the YT-DLP dependency
00:05:20to scrape YouTube's metadata for us.
00:05:22So when I ask it for, hey, Claude code skills,
00:05:24it's just like us going on YouTube
00:05:27and searching in Claude code skills.
00:05:28It grabs title, views, author, all that important stuff.
00:05:32And this skill you see right here inside of Claude code
00:05:35teaches Claude code how to best use this script.
00:05:38Now there's two ways to get this skill
00:05:41and this script up and running.
00:05:42The first is rather simple.
00:05:44You just go inside of Claude code
00:05:45and explain that you want it to build this custom skill
00:05:48for you, that you want to use the YT-DLP dependency
00:05:51to create a custom YouTube scraper.
00:05:54Or if you want this entire YouTube search skill
00:05:57set up MD file, and you can just download it
00:05:59and hand it to Claude code, you can do that too.
00:06:01And you can get it by heading to my free school community,
00:06:03which you can find a description for.
00:06:04Now, speaking of my school communities
00:06:06inside of Chase AI Plus,
00:06:07you can also find my Claude code masterclass,
00:06:11which I just released a few days ago.
00:06:13So if you're someone who is kind of in the beginning
00:06:14of their AI journey is trying to figure out
00:06:16how can I best leverage Claude code,
00:06:18even if I'm not technical,
00:06:19but I really want to master what is definitely
00:06:22the most powerful AI tool in the game.
00:06:24Well, that's the place for you.
00:06:25And if you're interested in that,
00:06:26just check the link in the pin comment.
00:06:28So once you download the YouTube search skill set up file,
00:06:31just give it the Claude code and tell it to go to work,
00:06:33or you can just manually prompt the Claude code
00:06:35to build it for you.
00:06:36Now, let's head back here
00:06:37and set up the notebook LM connection.
00:06:39So I'll put a link to this down in the description as well.
00:06:42And the install is pretty simple.
00:06:44So to install this,
00:06:45we are just going to copy these commands,
00:06:47paste it inside of our terminal,
00:06:49which means, hey, if you're looking at Claude code,
00:06:51you're wrong.
00:06:51You need to open a second terminal that looks like this
00:06:53and paste in those commands.
00:06:55After you run these initial installation commands,
00:06:57what you want to do is scroll down here
00:06:59and we're gonna have to run one more command in the CLI.
00:07:01And that's the notebook LM login command.
00:07:04Same thing as before,
00:07:04head into another terminal, paste this in there.
00:07:07That's going to open a new window in Chrome.
00:07:10All you have to do is log in.
00:07:11You just have to do it once and you're all set.
00:07:13Now, after you authenticate with the notebook LM login,
00:07:16there's one more thing you have to do
00:07:17and we need to do the skill.
00:07:20So to do the skill inside of Claude code, right?
00:07:22You can either run this command in the terminal
00:07:25or just tell Claude code to do it.
00:07:27So understand also what we've done.
00:07:29We've done two things on the notebook LM side.
00:07:30You have the skill
00:07:32and then you have like the actual API of the program.
00:07:35Remember skills are just prompts.
00:07:37It's just text telling Claude code
00:07:39how to do something in a specific manner.
00:07:42So all this information you see here about,
00:07:44hey, here's how we generate content.
00:07:46Here's how we create the notebook.
00:07:47This skill is teaching Claude code how to do that.
00:07:50So once it's installed, you just need to tell Claude code,
00:07:52hey, I want you to use notebook LM to create flashcards
00:07:56or an infographic or slide deck.
00:07:58It's that easy.
00:07:58Everything's just in plain language.
00:08:00And if you're wondering what exactly you can do
00:08:02with this program, well, it's all here inside of the GitHub.
00:08:04Anything you can do in notebook LM manually,
00:08:06you can do with the API and some.
00:08:09As you can see here beyond the web UI,
00:08:11we can also do batch downloads,
00:08:13export the quiz and flashcards, et cetera, et cetera.
00:08:16So we actually get more functionality using this program
00:08:19than you would do just loading up notebook LM yourself.
00:08:22So let's actually go through it one step at a time
00:08:24so you understand how it's working.
00:08:25So the first thing is that YouTube search skill.
00:08:28So like any skill, we can either use it as a slash command
00:08:30or you can just use plain language.
00:08:32But if I do YT dash search, you see we have query
00:08:36and then the count.
00:08:37So what are we looking for?
00:08:37Hey, we're looking for plod code skills.
00:08:41And so while in the demo, we did everything at once,
00:08:43I think it's useful to break it up sometimes
00:08:45so you can first get eyes on what your sources
00:08:48are actually going to be.
00:08:50So here's the results.
00:08:51It comes back with, right at any time,
00:08:53we can also check the YouTube links ourselves.
00:08:55And what's nice with this skill
00:08:56is it will also give you some insight
00:08:58about what's actually going on with what it brought back.
00:09:01So if you're happy with the sources,
00:09:02now we can push it into notebook LM.
00:09:04So again, you can just use plain language.
00:09:05Create a new notebook in notebook LM titled chase demo
00:09:08with these sources we just pulled.
00:09:10And we can see it created the notebook
00:09:12and now it's going to be populating it with its sources.
00:09:14So after a couple of minutes, all 20 sources are loaded
00:09:17and you're limited to 50 sources with notebook LM.
00:09:19And at this point you can have notebook do whatever you want.
00:09:21So we can say, based on those videos,
00:09:23what does notebook LM believe
00:09:24is the number one plod code skill?
00:09:26Now, again, the cool thing with this
00:09:28is all this analysis is offloaded.
00:09:31Plod code isn't doing this analysis.
00:09:33Plod code isn't spending tokens.
00:09:35It's only spending a small amount of tokens
00:09:36to send that request to notebook LM and bring it back.
00:09:39But all the thinking is done by Google
00:09:42and they're paying for it.
00:09:43So plod code grabbed notebook LM's analysis.
00:09:47And we can see that reflected here
00:09:49inside of notebook LM itself.
00:09:50So you can always double check and click inside of notebook LM
00:09:52if you wanted to see like what captions it's referencing to.
00:09:55And this cadence also applies for all the deliverables.
00:09:58So right, if you want the audio overview,
00:09:59the mind map, the flashcards, the infographics,
00:10:01anything you see over here on the right,
00:10:03just prompt plod code and it will do it for you.
00:10:06It's that easy.
00:10:08So how you end up leveraging this research workflow
00:10:10is ultimately up to you,
00:10:11but I really can't stress enough how wild this thing is.
00:10:15It seems pretty simple on the surface,
00:10:17but I'm telling you,
00:10:18if you've tried to deal with anything like this,
00:10:20especially with the YouTube video stuff
00:10:22and actually trying to create some like corpus of knowledge
00:10:25from these videos in a way that plod code
00:10:27or some other agentic code can interact with it,
00:10:30it's pretty difficult, right?
00:10:31And it's very time consuming
00:10:32and it can be rather brittle.
00:10:34Yet all of this is abstracted away for free with notebook LM.
00:10:39So I think this is an awesome tool.
00:10:42I hope you can get some use out of it.
00:10:44As always, like I said before,
00:10:45all the resources can be found in my school communities.
00:10:48If you need the MD file for the skill,
00:10:52for the YouTube search skill,
00:10:53make sure you see that in the free one.
00:10:54And again, if you're a little more serious about this stuff
00:10:56and you're like,
00:10:57I really just want to have like a clod code masterclass
00:10:59that gets me from like zero to AI dev,
00:11:01make sure to check out Chase AI+.
00:11:03So let me know what you thought of this in the comments
00:11:05and as always, I'll see you around.