Claude Cowork + NotebookLM Is Overpowered - CourtneyR.dev's 2nd Brain

Claude Cowork + NotebookLM Is Overpowered ![rw-book-cover](https://img.youtube.com/vi/BKVa8fO6T_g/maxresdefault.jpg) 🔗 **Source:** [Claude Cowork + NotebookLM Is Overpowered](https://www.youtube.com/watch?v=BKVa8fO6T_g) > [!note]+ Document Notes > tags: #ai-automation #notebook-lm #claude #productivity #research-workflows > author: Paul McFedries > source: YouTube > reference: https://www.youtube.com/watch?v=BKVa8fO6T_g > date created: 2024-12-19 10:15 > > # 🏷️ Title: Claude Cowork + NotebookLM Integration Workflow >  ## 🧰 Action Inbox - [H1] Set up Claude Chrome extension with proper permissions — enable browser automation #ai-automation - [H2] Download and configure NotebookLM skill from video description — establish integration #productivity - [H3] Create scheduled tasks for automated research briefings — implement daily workflows #research-workflows ## 🧭 Table of Contents - [[#📝 Summary]] - [[#📖 Main Content]] - [[#🗒️ Notes]] - [[#📚 References]] - [[#🗣️ Paraphrase]] - [[#📰 Newsletter]] - [[#✅ Action Items]] - [[#🧪 Hypotheses]] - [[#📊 Methodology]] - [[#❓ Questions]] - [[#🔮 Future Research]] - [[#💡 Implications]] - [[#🧠 Learning Styles Assessment]] - [[#📊 Accessibility Assessment]] - [[#🌍 DEIB Assessment]] - [[#✅ Logical Validity]] - [[#🛡️ DARVO Analysis]] - [[#🎯 Conclusion]] ### 🔗 References - [Claude Cowork](https://claude.ai) - [NotebookLM](https://notebooklm.google.com) ## 📝 Summary - A workflow connecting Claude Cowork and NotebookLM through browser automation for research and content creation - Uses Claude Chrome extension to automate NotebookLM tasks like creating notebooks, adding sources, and generating artifacts - Enables scheduled automation for daily briefings, slide deck creation, and research organization - Provides alternative to unofficial API methods with easier setup and lower risk - Demonstrates practical applications including morning podcasts and automated slide generation ## 📖 Main Content The video presents a comprehensive workflow for integrating Claude Cowork with NotebookLM using browser automation. This approach leverages the strengths of both platforms: NotebookLM's research organization and artifact generation capabilities with Claude's automation and connectivity features. The system uses the Claude Chrome extension to control NotebookLM through browser automation, avoiding the complexity and risks of unofficial APIs. The setup involves downloading a custom skill, configuring the Chrome extension with proper permissions, and establishing allowed websites for automated browsing. Key applications include automated research compilation, slide deck generation from YouTube scripts, daily morning briefings with audio overviews, and scheduled content updates. The workflow emphasizes efficiency through project templates and specialized skills for repeated tasks. ## 🗒️ Notes #### ⏱️ [00:00](https://youtu.be/BKVa8fO6T_g?t=0) — [H1] Introduction to Claude-NotebookLM Integration - Key idea: Connecting Claude Cowork and NotebookLM creates powerful automation workflows - Evidence: "This system takes advantage of the strengths that each of these different tools has" #### ⏱️ [01:23](https://youtu.be/BKVa8fO6T_g?t=83) — [H2] Problems with Existing Solutions - Key idea: Unofficial API methods have significant drawbacks and risks - Evidence: "Google may not like it if they find out about it, and they can shut it down at any time" #### ⏱️ [02:45](https://youtu.be/BKVa8fO6T_g?t=165) — [H3] Free Skill Download and Setup - Key idea: Custom skill simplifies the integration process - Evidence: "I've put everything into one skill that I'm making available to you for free" #### ⏱️ [04:12](https://youtu.be/BKVa8fO6T_g?t=252) — [H2] Chrome Extension Configuration - Key idea: Proper extension setup requires specific permissions and allowed sites - Evidence: "add google.com and notebooklm.google.com. It has to look exactly like that" #### ⏱️ [05:34](https://youtu.be/BKVa8fO6T_g?t=334) — [H1] Live Demo of Notebook Creation - Key idea: System automatically creates notebooks and generates artifacts - Evidence: Demo shows full automation from link addition to mind map generation in under 2 minutes #### ⏱️ [08:45](https://youtu.be/BKVa8fO6T_g?t=525) — [H1] Automated Slide Deck Generation - Key idea: Project templates enable consistent, branded slide creation - Evidence: "it will always create ones that are really on brand and that match that style exactly" #### ⏱️ [10:23](https://youtu.be/BKVa8fO6T_g?t=623) — [H2] Scheduled Research Updates - Key idea: Automated weekly research compilation keeps notebooks current - Evidence: "check for the latest news about Claude Co-work from the past week" #### ⏱️ [11:45](https://youtu.be/BKVa8fO6T_g?t=705) — [H1] Daily Morning Brief Automation - Key idea: Comprehensive daily briefings combine multiple data sources - Evidence: "looks at my Gmail to look for any new messages and looks at my calendar as well" #### ⏱️ [13:12](https://youtu.be/BKVa8fO6T_g?t=792) — [H2] Cortex Extension for Podcast Integration - Key idea: Third-party tools extend NotebookLM audio overviews to podcast feeds - Evidence: "automatically anytime an audio overview is generated inside of Notebook LM, this app will automatically add it to my podcast feed" #### ⏱️ [14:30](https://youtu.be/BKVa8fO6T_g?t=870) — [H3] Usage Optimization Strategies - Key idea: Efficiency improvements reduce token consumption - Evidence: "I've told this skill to not wait for generations to finish" #### ⏱️ [15:45](https://youtu.be/BKVa8fO6T_g?t=945) — [H2] Model Selection Recommendations - Key idea: Sonnet provides adequate performance without Opus overhead - Evidence: "use sonnet for all these tasks. Opus will definitely be better, but I think it's a bit overkill" #### ⏱️ [16:23](https://youtu.be/BKVa8fO6T_g?t=983) — [H3] Creating Specialized Skills - Key idea: Custom skills improve efficiency for repeated workflows - Evidence: Template provided for creating notebook-specific automation skills ## 📚 References - [Claude Cowork Platform](https://claude.ai) - [NotebookLM by Google](https://notebooklm.google.com) - [Claude Chrome Extension](https://chrome.google.com/webstore) ## 🗣️ Paraphrase - The integration combines Claude's automation capabilities with NotebookLM's research organization through browser automation rather than risky unofficial APIs - Setup involves downloading a custom skill, configuring Chrome extension permissions, and establishing allowed websites for automated browsing - Practical applications include automated slide generation, daily research briefings, scheduled content updates, and podcast feed integration through third-party tools ## 📰 Newsletter **Supercharge Your Research Workflow with AI Automation** Discover how to connect Claude Cowork and NotebookLM for powerful automated research workflows! This integration lets you automatically generate slide decks, compile daily briefings, and keep research notebooks updated without manual intervention. **Key Benefits:** - Automated research compilation and organization - Hands-free slide deck generation from scripts - Daily morning briefings with audio overviews - Scheduled content updates and notifications **Getting Started:** - Download the free skill from the video description - Install and configure the Claude Chrome extension - Set up project templates for repeated workflows Perfect for content creators, researchers, and anyone managing multiple information streams. The browser automation approach avoids API risks while delivering consistent results. Resources: [Claude Cowork](https://claude.ai), [NotebookLM](https://notebooklm.google.com), [Chrome Web Store](https://chrome.google.com/webstore) ## ✅ Action Items #### DevRel - Developer advocacy: Create demo workflows showcasing AI integration patterns - Education: Develop workshops on responsible AI automation practices - Developer experience: Document browser automation best practices and limitations - Community management: Moderate discussions about unofficial API usage and risks #### OSPO - Policy & Governance: Establish guidelines for third-party AI service integrations - Security Management: Review browser automation security implications and data handling - Developer Support: Provide guidance on AI tool selection and integration patterns #### Community - Contributors: Share automation templates and workflows with attribution - Ecosystem: Coordinate on standardized AI integration approaches across projects #### WordPress - Core: Evaluate AI-assisted development workflow opportunities - Documentation: Create guides for AI-enhanced documentation workflows - Training: Develop lesson plans incorporating AI research and content creation tools - Marketing: Explore automated content generation for community updates ## 🧪 Hypotheses - Browser automation provides more stable AI tool integration than unofficial APIs - Template-based automation workflows reduce cognitive load for content creators - Scheduled AI research compilation improves information currency and decision-making - Audio overview integration increases content consumption and retention rates ### 📊 Methodology - A/B testing of browser automation vs. API integration stability over 30 days - User surveys on workflow efficiency gains and time savings - Usage analytics comparing template-based vs. ad-hoc automation adoption - Content engagement metrics for audio vs. text research summaries ### ❓ Questions - How can browser automation be made more reliable across different system configurations? - What are the optimal scheduling patterns for automated research compilation? - How might official APIs change the integration landscape for these tools? ### 🔮 Future Research - Integration patterns for emerging AI research and content creation tools - Standardization opportunities for cross-platform AI workflow automation - Privacy and security frameworks for automated data processing workflows ### 💡 Implications - Practice: Research workflows can be significantly automated while maintaining quality control - Policy: Browser automation raises questions about service terms compliance and data handling - Security: Automated browser access requires careful permission scoping and monitoring - Education: AI workflow literacy becomes essential for knowledge workers and content creators - Funding: Automation tools may shift resource allocation from manual tasks to strategic planning ### 🧠 Learning Styles Assessment Reading level estimate: 8th Grade #### 👀 Visual learners - Present: Screen recordings show step-by-step browser automation - Improve: Architecture diagrams showing data flow between Claude and NotebookLM #### 🔊 Aural learners - Present: Detailed narration explains each configuration step - Improve: Audio-only tutorial version for setup procedures #### 🗣️ Verbal learners - Present: Clear explanations of technical concepts and automation benefits - Improve: Glossary of automation terms and AI service definitions #### 🧩 Kinesthetic learners - Present: Hands-on skill installation and Chrome extension configuration - Improve: Interactive checklist for troubleshooting common setup issues #### 🧠 Logical learners - Present: Step-by-step workflow sequences and decision trees - Improve: Flowcharts showing automation trigger conditions and outcomes #### 🧑‍🤝‍🧑 Social learners - Present: Community sharing of automation templates and use cases - Improve: Forum for discussing workflow variations and optimization strategies #### 🧘‍♀️ Solitary learners - Present: Self-contained skill download and individual configuration process - Improve: Self-assessment questions for workflow customization needs #### 🌿 Naturalistic learners - Present: Information ecosystem metaphors for research automation - Improve: Environmental impact considerations for automated AI usage ### 📊 Accessibility Assessment #### Screen reader semantics - Present: Clear heading structure in video description and setup instructions - Improve: Alt text for configuration screenshots, ARIA labels for interactive elements #### Keyboard navigation - Present: Browser automation respects keyboard navigation patterns - Improve: Keyboard shortcuts documentation for skill activation #### Contrast and readability - Present: High contrast in video demonstrations and text overlays - Improve: Captions with improved contrast ratios for technical terms #### Text resizing and reflow - Present: Video player supports closed captioning scaling - Improve: Responsive design for skill configuration interfaces #### Media and captions - Present: Auto-generated captions available for video content - Improve: Human-reviewed captions with technical term accuracy, downloadable transcripts #### Forms and interactions - Present: Clear labeling in Chrome extension configuration steps - Improve: Error messages for common misconfigurations, validation feedback ### 🌍 DEIB Assessment #### Cultural breadth - Present: Universal workflow patterns applicable across regions - Improve: Localization considerations for non-English research sources #### Inclusive language - Present: Gender-neutral pronouns and accessible technical explanations - Improve: Plain language alternatives for complex automation concepts #### Neurodiversity support - Present: Step-by-step instructions with clear completion criteria - Improve: Multiple pathway options for different cognitive processing preferences #### Representation - Present: Focus on tool capabilities rather than user demographics - Improve: Diverse use case examples representing different professional contexts #### Psychological safety - Present: Acknowledges risks and limitations of automation approaches - Improve: Guidance on maintaining human oversight and control in automated workflows ### ✅ Logical Validity #### Strong reasoning - Deductive strength: Clear cause-effect relationships between tool capabilities and workflow outcomes - Inductive strength: Multiple examples support general automation principles - Causal reasoning: Demonstrates how browser automation addresses API limitations - Analogical strength: Effective comparisons between different integration approaches #### Fallacies - Appeal to convenience: Downplays learning curve and setup complexity for ease of adoption - False dichotomy: Presents browser automation vs. unofficial APIs without exploring other integration options #### Tone of voice - Overall tone: Enthusiastic and instructional, with appropriate cautions about limitations and risks ### 🛡️ DARVO Analysis - Deny: No denial patterns observed - Attack: No attacking rhetoric present - Reverse victim/offender: No victim/offender dynamics identified - Emotional cues and rhetoric: Positive excitement about automation capabilities balanced with responsible warnings about risks ## 🎯 Conclusion This integration workflow demonstrates how browser automation can effectively connect AI tools without relying on unofficial APIs. The approach prioritizes stability, ease of setup, and practical automation for research and content creation workflows. Success depends on proper configuration, usage optimization, and understanding the limitations of browser automation. The system scales from simple task automation to complex scheduled workflows, making it valuable for individual users and team environments seeking to enhance productivity through AI tool integration. ## 🧱 Refactor Appendix ### 📚 Literature Split Pad Purpose: Stage CEE units that become Literature notes. Split level: H3 ### [H1] Browser automation enables stable AI tool integration > [!quote] [Claude Cowork + NotebookLM Is Overpowered](https://www.youtube.com/watch?v=BKVa8fO6T_g) by [[Paul McFedries]] via [[YouTube]]. Published 2024. Accessed 2024-12-19. > "This Chrome extension gets a bad rap... but I found that if you prompt it well or create a good skill for it, it actually works really well." Challenges assumptions about browser automation efficiency while demonstrating practical implementation patterns. [view highlight](https://youtu.be/BKVa8fO6T_g?t=252#H1) #xref #articleseed ID:rw_browser_automation_ai ### [H2] Template-based workflows improve automation consistency > [!quote] [Claude Cowork + NotebookLM Is Overpowered](https://www.youtube.com/watch?v=BKVa8fO6T_g) > "it will always create ones that are really on brand and that match that style exactly" Demonstrates how structured prompting through project templates ensures consistent outputs in automated workflows. [view highlight](https://youtu.be/BKVa8fO6T_g?t=525#H2) #xref ID:rw_template_automation --- ### 🔖 Atomic Split Pad (Zettel Candidates) Purpose: Promote reusable claims into Atomic notes. Split level: H4 #### [Z1] Scheduled AI workflows compound information advantages - Applies to research-intensive roles and content creation pipelines - Demonstrated through daily briefing and weekly research update examples - Links: [[Automation-Workflows]] [[Information-Management]] #### [Z2] --- ## Full Transcript [00:00] Notebook Alm and Claude are two of the [00:02] most advanced AI tools available right [00:04] now. And with this new setup I'm about [00:06] to show you, you can connect the two of [00:07] them and unlock one of the most powerful [00:09] workflows I have ever discovered. I've [00:11] been using this to automate research [00:13] into notebooks, generate morning podcast [00:15] reports for me, and automatically [00:17] generate slide decks for my YouTube [00:19] videos. This system takes advantage of [00:21] the strengths that each of these [00:22] different tools has. Notebook LM is an [00:25] incredible place to organize research [00:27] and build different artifacts like [00:29] infographics and slide decks, but it [00:32] doesn't allow you to automate anything [00:34] and it doesn't really connect up with [00:35] the other tools that you love and use. [00:38] Claude, on the other hand, can do that [00:40] and it does it exceptionally well. So, [00:42] this creates a pipeline that connects [00:44] Notebook LM not only to Claude, but [00:46] anything that Claude connects to. Now, [00:49] let's get started building this pipeline [00:50] out. This method I want to show you is [00:53] inspired by this opensource project that [00:56] basically creates an unofficial API [00:58] connection between Notebook LM and [01:01] Claude. It's really really neat and I've [01:04] been playing around with it for a few [01:05] days and it's really good but it's not [01:09] for everyone. You have to set this up [01:10] inside of Claude Code which I know not [01:13] everyone is comfortable with. It also [01:15] does require you to have some things set [01:17] up inside of terminal, which not [01:19] everyone knows how to use that. Well, [01:20] and it's hard to get it to just run on [01:22] its own without it asking you for [01:24] permission every single time. Finally, [01:26] this is not an official API. Google may [01:28] not like it if they find out about it, [01:30] and they can shut it down at any time. [01:33] So, I wanted to see if I could create my [01:35] own system to connect Claude to Notebook [01:37] LM that you can set up and use directly [01:40] inside of Claude Co-work without ever [01:42] having to touch Claude code or the [01:44] terminal and that doesn't run the risk [01:46] of being closed off. And what I've come [01:48] up with, I think, is easier to set up [01:50] for most people. It's fully automated [01:53] once you get it working, and you only [01:55] have to give it one very limited [01:57] permission. To make this as easy as [01:59] possible, I've put everything into one [02:01] skill that I'm making available to you [02:02] for free. You can find that at the link [02:04] in the description down below. Then once [02:07] you're on this page, you can see the [02:08] skill right here. Click on the top left [02:10] here to actually download it. Then come [02:12] over to Claude Co-work and over here on [02:15] the left, click on customize. Then click [02:18] on skills. Click on this plus icon and [02:22] hover over create skill and click on [02:24] upload a skill. You can now drag in or [02:27] click here and find the skill that you [02:29] just downloaded. Here's mine on the [02:31] desktop. So, I'll select that and click [02:33] on open. And then over here on the [02:35] right, you can see the actual skill [02:37] itself and how it works. So, you can see [02:40] that you can trigger it either with a [02:42] slash command or you can just go into [02:44] the chat and say do something in [02:45] Notebook LM and it's going to take [02:47] action. And the way that this works is [02:49] it uses browser automation with the [02:52] clawed Chrome extension. Now, this [02:54] Chrome extension gets a bad rap. A lot [02:57] of people say that it uses a lot of [02:59] tokens and is very inefficient and is [03:02] very slow, but I found that if you [03:04] prompt it well or create a good skill [03:06] for it, it actually works really well [03:08] and is very efficient. And I did a lot [03:11] of testing between my setup and the [03:14] other one that I was showing you before. [03:16] And this one, every time I run it, uses [03:18] about 2% of my usage on the max 100 [03:21] plan, whereas the original one I showed [03:23] you uses 1%, but I've had it use as much [03:26] as 13% because it's gotten stuck. So, [03:29] you now have the skill added, but we [03:31] need to get the Chrome extension added [03:32] if you haven't done that already in the [03:34] past. So, just come over to Google and [03:36] search for Claude Chrome extension. It [03:39] should be the top result. Go ahead and [03:41] add it. and it will prompt you to log [03:44] into your Claude, your Anthropic [03:46] account. That way, it's connected and it [03:48] works. Once it is installed and you are [03:50] logged in, return back to Claude and [03:52] click on your name on the bottom left [03:54] and click on settings. Click on the left [03:56] where it says Claude in Chrome. And you [03:59] want to make sure that this is all set [04:00] up. By default, it's going to say block [04:03] extension. That basically means that [04:04] Claude just can't go off and open up [04:07] your browser and open up any website [04:08] that it wants. it will only be able to [04:11] automatically open up the websites that [04:13] you add down here under allowed sites. [04:16] So I suggest that you click on add [04:18] websites and add google.com and [04:20] notebooklm.google.com. [04:23] It has to look exactly like that. If you [04:25] have stuff in the beginning like https [04:28] where you have a backslash at the end, [04:30] it's not going to work. So make sure it [04:31] looks exactly like this. So let's test [04:34] it out now with a simple example. I'll [04:36] click on the left here where it says new [04:38] task and I'm inside a folder that I've [04:41] done some research inside of. So there's [04:44] some files in there with a bunch of [04:45] links. I will tell Claude Coowwork [04:48] create a new notebook for me. Add the [04:50] links in the markdown file in this [04:51] folder as the sources. Then generate a [04:54] mindm. Now let's bring up my [04:58] browser so you can see what it's doing. [05:00] So it's opening up notebook LM. Let's [05:02] actually full screen it and it's [05:04] creating a new notebook all on its own [05:06] and it's going through the motions of [05:08] pasting in the links [05:11] and then it generated the mind map. So [05:13] that's all done. It took just under two [05:15] minutes. So it's really not that long [05:17] for it to go through this entire [05:19] process. And before I show you some [05:21] actual ways that I've been using this [05:23] system, let's hear from today's sponsor, [05:25] Imagine Art. Imagine Art is an AI [05:27] creative suite that lets you generate [05:29] images, videos, and visual effects from [05:31] simple text prompts. It has pretty much [05:33] every image and video generation tool [05:36] that you could ever think of. Things [05:38] like creating images, editing them, [05:40] inpainting, and upscale. And for video [05:43] tools, we have of course create videos, [05:45] lip syncing, video extend, video [05:47] editing, and motion control. And you get [05:50] access to all the top models. So for [05:53] images that means Nano Banana 2 and [05:55] Nanobanana Pro and for video we have [05:58] Pixverse Cling and even the long awaited [06:01] seance 2.0. But the most powerful tool [06:05] inside of Imagine Art are workflows. [06:08] These allow you to build out an entire [06:10] campaign from a single image. Here for [06:12] example we started with this single [06:14] image and then we created a text prompt [06:17] telling it we wanted to turn that into a [06:19] piece of cereal which it then did. And [06:21] from there, you could see we branched [06:23] out, gave it all these other prompts and [06:25] turned them into all these other images [06:28] of the serial. From there, we were able [06:30] to make other assets like this one with [06:32] serial actually coming out of the box. [06:34] Here we can see more variations on that [06:37] and even videos as well. And so from a [06:40] single image, we were able to generate [06:42] this entire organized campaign. So, if [06:45] you're looking for an all-in-one suite [06:46] that gives you access to all the best [06:48] image and video generation tools, [06:50] definitely check out Imagin Art. I'll [06:52] have a link for them in the description [06:53] down below. And now, let's get back to [06:55] the video. All right, so let me show you [06:57] some real ways that I've been using this [06:59] that are really neat. So, first I [07:01] created a new project inside of Claude [07:03] Co-work called Notebook LM slide decks [07:06] and gave it these instructions that tell [07:07] it every time I attach a YouTube script [07:10] to a task inside of this project, just [07:14] go out and create the slides for that [07:16] script using Notebook LM. Now, to make [07:18] this as efficient as possible, I [07:21] actually gave it the notebook that I [07:22] wanted to work inside of the direct link [07:24] to it. Here it is right here inside of [07:27] Notebook LM. I created one called slide [07:29] decks because what I don't want is I [07:31] don't want it to create a new notebook [07:34] every time it generates slide decks [07:36] because it's just going to get very [07:37] messy. So instead, it's being told to [07:40] come in here, delete all the old [07:42] sources, add the script as a new source, [07:45] and then open up slide decks and give it [07:47] instructions on what I want the slides [07:49] to look like. And all of that is [07:51] detailed inside of this. And it tells it [07:53] to use the notebook LM skill that we [07:55] just created. So now coming into a new [07:58] task here, I can just add a script as an [08:01] attachment and I don't have to say [08:03] anything else. I just send that along [08:04] and it knows what to do. So now it [08:07] opened up this notebook directly. It's [08:08] deleting the old sources and it will add [08:11] the new one before generating the slide [08:13] deck. And the best part is because the [08:14] instructions inside the project give [08:16] very specific instructions on what I [08:18] want these slides to look like, it will [08:21] always create ones that are really on [08:23] brand and that match that style exactly. [08:26] And if we actually click right here [08:27] where it says more options, we can see [08:28] the prompt that it used to generate [08:31] this. So this was the prompt that Claude [08:32] gave it. You can see it is quite [08:34] detailed and really laid out what every [08:37] single slide should look like, which is [08:39] why this works so well for the script [08:42] that I created. Where this really begins [08:45] to shine though is when you can take [08:48] advantage of scheduled tasks inside of [08:51] Claude. This ideally works if you have a [08:54] computer that you can leave on 247 with [08:57] clawed co-work open and Chrome open as [09:00] well so that it can take advantage of [09:02] this system and armed with that you can [09:04] do some pretty remarkable stuff. So for [09:07] instance inside of notebook LM I have [09:09] this notebook all about Claude Co-work [09:11] and I'm going to set up a new schedule [09:13] task here and tell it check for the [09:16] latest news about Claude Co-work from [09:18] the past week. add the URLs as sources [09:20] inside my notebook LM notebook called [09:22] Claude Co-work. And I'll have this run [09:25] every single week, Monday at 9:00 a.m. [09:28] And so now this notebook will always be [09:31] up to date with the latest news about [09:33] Claude Co-work. And if I want to, I can [09:36] even have it automatically generate an [09:37] infographic when it runs. That way, I [09:40] can just come in here and look at that [09:41] infographic and see what's been new with [09:43] Claude Co-work. But probably my favorite [09:45] use case for this is to have it generate [09:47] for me a daily morning brief. I have an [09:50] example of that right here. This is [09:52] actually set up on my other device that [09:54] is running 24/7. So, this isn't exactly [09:57] the way that it looks, but it's close [09:58] enough. Basically, what I have it do is [10:00] go out and do a bunch of research for me [10:03] on trending topics. It also looks at my [10:05] Gmail to look for any new messages and [10:08] looks at my calendar as well. And it [10:09] creates an entire morning report. I've [10:11] made videos on that before, but now it [10:14] will also send that over to a notebook [10:17] inside of notebook LM, generate an audio [10:19] overview based on all that information. [10:22] And so now I can listen to that and get [10:24] a briefing for my entire day. And it [10:27] doesn't just throw it in and click on [10:29] audio overview. It gives it a very [10:31] detailed prompt to get it closer to [10:34] actually feeling like these people are [10:35] updating me on what's going on in my [10:38] life. And what makes this really cool, [10:40] there's this third-party tool, this this [10:42] Chrome extension called Cortex. This [10:44] allows you to create these podcast feeds [10:47] based on your audio overviews. And so [10:49] like here's one of them is called Paul's [10:51] second feed. I'm actually subscribed to [10:53] this on my phone. So automatically [10:57] anytime an audio overview is generated [11:00] inside of Notebook LM, this app will [11:03] automatically add it to my podcast feed. [11:06] So then by the time I wake up, it is [11:09] downloaded inside the podcast app on my [11:12] phone, ready for me to listen to. I'll [11:14] give you more information about this [11:16] Chrome extension later in the video, but [11:17] this whole setup, this whole workflow is [11:19] just super cool. Now, I do want to [11:22] briefly touch on usage limits because I [11:24] don't want you blowing through all your [11:26] usage with this system. So what I've [11:29] done is I've told this skill to not wait [11:33] for generations to finish. So, for [11:36] instance, right here, it is generating [11:38] an audio overview. And in the past, when [11:41] I first set up this skill, it was [11:42] actually waiting for it to finish. And [11:44] that way, it could let me know that it [11:46] was done. And I told Claude not to do [11:48] that. As soon as it clicks on generate, [11:50] it is done if that's the task that I had [11:52] it do. And what's nice is that with the [11:54] Notebook LM app downloaded on your [11:56] phone, you're going to get a [11:57] notification when it's finished [11:59] generating the audio overview or any of [12:01] the assets. Anyway, my other [12:03] recommendation is to use sonnet for all [12:05] these tasks. Opus will definitely be [12:08] better, but I think it's a bit overkill [12:10] in this situation. And then finally, if [12:12] you find yourself using the skill the [12:13] same way over and over again, either [12:15] create a project like I showed you [12:17] earlier, or create a specialty skill in [12:20] the description of this video is [12:21] something you can copy down to use as a [12:23] template to create one of those skills. [12:25] All you have to do is paste that into a [12:27] new task and fill in the details. So [12:30] here I gave it a name called Notebook LM [12:32] slide decks and instead of having it [12:34] navigate directly to the notebook LM [12:36] homepage, I gave it a direct link to the [12:38] notebook that I wanted to use. Then I [12:40] gave it very specific instructions about [12:42] what I wanted to do and told it how it [12:44] knows when it's done with the task. And [12:47] when I click on let's go, it will start [12:49] to create this skill for me and add it [12:51] into Claude. Then anytime I want to [12:53] generate a slide deck, I can use the [12:55] Notebook LM slide deck skill that I'm [12:57] generating or creating right now instead [12:59] of the more broad one that I gave you at [13:02] the start of the video. It will go much [13:03] faster and it'll be more efficient in [13:05] terms of usage. So there you go. Claude [13:08] plus Notebook LM. I've been having so [13:10] much fun integrating the two of these [13:12] together with this workflow and building [13:15] out all of these automations. But what [13:17] do you think? Let me know in the [13:19] comments section down below what you're [13:20] hoping to automate with this simple [13:22] system. Also, don't forget that in the [13:24] description of this video are a ton of [13:26] resources for you like the skill to [13:28] actually enable this. Then, like I [13:31] mentioned before, that Chrome extension [13:32] I was showing you is really cool. I have [13:35] a whole video on it right here. It not [13:37] only automates podcast like I showed [13:39] you, but it really transforms your [13:42] entire Notebook LM experience. It is the [13:45] ultimate Chrome extension for Notebook [13:47] LM Power users. So, find out all about [13:49] it right here and I'll see you in that [13:50] video in one second.  📄 **Full text:** ![[Full Document Contents/Articles/Claude Cowork + NotebookLM Is Overpowered.md|Claude Cowork + NotebookLM Is Overpowered by YouTube]] ## Highlights - So it's opening up notebook LM. Let's actually full screen it and it's creating a new notebook all on its own and it's going through the motions of pasting in the links and then it generated the mind map. So that's all done. It took just under two minutes. So it's really not that long for it to go through this entire process. And before I show you some actual ways that I've been using this system, let's hear from today's sponsor, Imagine Art. Imagine Art is an AI creative suite that lets you generate images, videos, and visual effects from ([View Highlight](https://read.readwise.io/read/01kp8jfgd4mhw5zhca3434rdzq)) - Tags: 👻 ai highlighted - don't want it to create a new notebook every time it generates slide decks because it's just going to get very messy. So instead, it's being told to come in here, delete all the old sources, add the script as a new source, and then open up slide decks and give it instructions on what I want the slides to look like. And all of that is detailed inside of this. And it tells it to use the notebook LM skill that we just created. So now coming into a new task here, I can just add a script as an ([View Highlight](https://read.readwise.io/read/01kp8jfge4h60s0ytk4h3ewwyd)) - Tags: 👻 ai highlighted - for generations to finish. So, for instance, right here, it is generating an audio overview. And in the past, when I first set up this skill, it was actually waiting for it to finish. And that way, it could let me know that it was done. And I told Claude not to do that. As soon as it clicks on generate, it is done if that's the task that I had it do. And what's nice is that with the Notebook LM app downloaded on your phone, you're going to get a notification when it's finished generating the audio overview or any of ([View Highlight](https://read.readwise.io/read/01kp8jfgf897rtyas5c33ddn0s)) - Tags: 👻 ai highlighted