1. Combining NotebookLM and Anti-gravity for Complete AI Systems [0]

1.1. Understanding the Roles of NotebookLM and Anti-gravity [0]

Combining NotebookLM (the reader) with Anti-gravity (the builder) allows users to create complete, cost-effective AI systems instead of just using a chatbot.

  1. The Power of Combination [0]

    1. NotebookLM recently became available within Google's Anti-gravity, which fundamentally changes how these tools are used [0].
    2. Most users treat these tools as simple chatbots, but combining them allows any research to be turned into a complete AI system that operates 24/7 [1].
    3. This combination helps users save time, avoid costly mistakes, and create custom apps, websites, and dashboards without incurring expensive API costs [1].

  2. Defining the Tools [6]

    1. Anti-gravity is the software builder, capable of creating apps, websites, and AI systems, powered by the world's most powerful models [6].
      1. It acts as the builder in this relationship, creating software, systems, and connections, serving as the command center [19].
      2. Anti-gravity is to building software what Microsoft Word is to documents and Excel is to spreadsheets [12].
    2. NotebookLM is the research tool, capable of understanding and researching anything, turning content into audio, and acting as an unlimited knowledge base [7].
      1. It acts as the reader, understanding, reading, and holding massive context, often handling 40 to 50 different documents, resources, and videos [16].

  3. Solving the Token Cost Problem [23]

    1. Using massive files directly within Anti-gravity for personal operating systems or building projects costs tokens every time the AI reads the files, leading to poor performance and high usage costs [23].
    2. Connecting Anti-gravity and NotebookLM solves this cost problem while unlocking incredible capabilities [24].
    3. This connection allows for the programmatic creation of unlimited notebooks within Anti-gravity, enabling deep research agents to run outside of Anti-gravity, which saves money and increases performance [25].
    4. Users can upload documents to Anti-gravity and add them to their NotebookLM notebooks [32].
    5. The combined system allows users to ask questions about any notebook in one chat and create infographics, reports, slide decks, and audio overviews at scale [27].

1.2. The Basics of NotebookLM [33]

NotebookLM is Google's primary tool for understanding resources, allowing users to create notebooks from various sources and customize the AI's interaction style.

  1. Access and Core Functionality [39]

    1. NotebookLM is Google's main software for understanding any resource, accessible at notebooklm.google.com [39].
    2. Users can create notebooks and drop in any resources, including articles, YouTube links, Google Drive files, or copied text [43].
    3. For example, a user can copy a YouTube link about Anti-gravity skills, drop it into NotebookLM, and add it as a resource [46].

  2. Research and Interaction Capabilities [47]

    1. NotebookLM can perform fast research by compiling a list of 30 to 40 or more resources on a given topic, such as Anti-gravity skills [47].
    2. Users can then chat with the compiled knowledge base, asking for summaries or specific information [49].

  3. Output and Customization [53]

    1. NotebookLM can export knowledge into various formats, including audio overviews, video overviews, mind maps, reports, flashcards, quizzes, infographics, slide decks, and data tables [53].
    2. Users can customize the AI's conversational goal, style, or role using toggles and custom examples [54].
    3. Users can untoggle specific resources to focus the conversation on a select few items [56].
    4. The platform makes learning and creating resources significantly easier, and users can chat with their notebooks directly within Gemini [57].
    5. The platform's main function is to ingest various files (PDFs, documents, research) into one large database, answer questions, and turn that knowledge into presentations or other resources [59].

2. Level 1: Programmatic Notebook Creation and Setup [61]

2.1. Installing and Connecting NotebookLM to Anti-gravity via MCP [61]

The first step is connecting NotebookLM to Anti-gravity using the Model Context Protocol (MCP) to enable programmatic creation of infinite, project-specific notebooks.

  1. The Goal of Level 1 [61]

    1. The first level is creating NotebookLM notebooks from within Anti-gravity, which allows users to connect the knowledge to any project and spin up an infinite number of notebooks [61].
    2. Users can ask Anti-gravity for topics (e.g., 10 topics on creating blue bottles) and then have Anti-gravity create 10 expert notebooks in NotebookLM for querying and project integration [62].
    3. The initial requirement is installing and connecting NotebookLM to Anti-gravity using the Model Context Protocol (MCP) [64].

  2. Understanding and Using MCP [71]

    1. MCP stands for Model Context Protocol, which allows Anti-gravity (or any app) to connect to any other app [72].
    2. The MCP is analogous to a universal remote control, enabling the intelligence within Anti-gravity to connect to services like Notion to create articles or perform other actions [74].

  3. Installation Process [77]

    1. Since NotebookLM is not in the default MCP store, users must use a specific prompt from a GitHub repository to install the server [77].
    2. The user pastes the prompt into the Anti-gravity agent, selects a model (like Claude Sonnet or Gemini), and allows the system to run the installation [86].
      1. NotebookLM is powered by the powerful Gemini 3 model [90].
    3. Anti-gravity will automatically install necessary components and check the MCP configuration file [94].
    4. Users can manage the frequency of permission requests in the Anti-gravity settings, choosing options like "always proceed" or "ask questions" [96].
    5. After successful installation, the user must connect the NotebookLM MCP by authorizing it via browser credentials [106].
      1. Anti-gravity opens a new browser window (indicated by a blue hue) and prompts the user to sign into NotebookLM to grab the necessary tokens [112].
    6. Once signed in, the NotebookLM MCP server appears in the "manage MCP servers" list, and Anti-gravity adds it to the raw config where all connection data resides [118].

2.2. Programmatically Creating Deep Research Notebooks [124]

By leveraging the MCP connection, users can instruct Anti-gravity to generate multiple deep research notebooks simultaneously based on specific topics.

  1. Generating Topics via Anti-gravity [124]

    1. The user starts a new chat and describes a business bottleneck or project goal to Anti-gravity [125].
    2. The user asks Anti-gravity to identify five key areas of knowledge needed for success (e.g., growing on YouTube), requesting a mix of specific and hyper-specialized topics [129].
    3. Anti-gravity responds with topics like content strategy, audience psychology, packaging, retention, and the YouTube algorithm [136].
    4. In the "old world," the user would have manually searched for and copied resources for each topic [140].

  2. Automating Notebook Creation and Deep Research [141]

    1. The user instructs Anti-gravity to use the NotebookLM MCP to create five different notebooks, one for each topic, using the deep research function [143].
    2. Anti-gravity triggers the MCP tool, programmatically spinning up the five notebooks simultaneously [149].
    3. This process demonstrates how Anti-gravity can build systems like Lego blocks, especially when provided with contextual information [156].

  3. Reviewing the Automated Output [158]

    1. Anti-gravity confirms that the deep research has been initiated for all five topics [160].
    2. Each research task takes about three to five minutes and gathers around 40 high-quality sources from across the web [162].
    3. The user could have specified the research sources (e.g., only from certain creators) or influenced the research using a "business context" file containing information about their business and customers [164].
    4. The user confirms that five new notebooks have been created on the NotebookLM homepage, each with a prompt and deep research underway [177].

3. Level 2: Codifying Behavior with Anti-gravity Skills [186]

3.1. Creating a NotebookLM Skill for Controlled Behavior [186]

Creating a skill in Anti-gravity allows users to codify specific instructions for how the AI should interact with NotebookLM, ensuring consistent, tailored results.

  1. Defining an Anti-gravity Skill [185]

    1. A skill is a repeatable, specific feature used consistently within Anti-gravity that describes how the AI should act [188].
    2. When the AI accesses the MCP, the skill ensures it follows a specific set of principles [192].

  2. The Skill Creation Process [193]

    1. The user grabs a skill template resource and instructs Anti-gravity to create a skill based on the provided information for accessing the MCP with NotebookLM [187].
    2. The user copies the NotebookLM capabilities from the MCP server management section and pastes them into the chat [199].

  3. Benefits of Codifying Behavior [205]

    1. Creating a skill provides control over how the AI behaves, allowing the user to be specific about actions [204].
    2. For example, the user can instruct the skill to always reference an "About Jack" file before creating notebooks to ensure the content is specific to the user's niche (e.g., helping businesses with AI systems) [206].
    3. This codification prevents the user from having to repeat the context in every conversation [209].

3.2. Automating Resource Import and Future Actions [212]

The skill can be updated to automate manual steps, such as importing research resources after they are generated, making the process fully hands-off.

  1. Automating the Import Step [212]

    1. Currently, Anti-gravity does not automatically import the research resources into NotebookLM [212].
    2. The user can manually click the import button or instruct Anti-gravity to automate this step [213].
    3. The user updates the skill to include an instruction: whenever notebooks are created, Anti-gravity should manually open them in the browser after a few minutes (e.g., five minutes) and import all resources [216].

  2. The Value of the Skill [219]

    1. The skill acts as an instruction manual, codifying how the user wants Anti-gravity and NotebookLM to behave [219].
    2. This system allows for the creation of infinite, tailored notebooks without manual intervention [223].
    3. The next step is leveraging these skills and notebooks within projects or chat interfaces inside Anti-gravity [224].

4. Level 3: Building Applications from Notebook Knowledge [227]

4.1. Leveraging Multiple Agents and Contextual Information [227]

By using the Agent Manager and providing personal context, users can query the deep research notebooks to generate highly specific, actionable strategies.

  1. Using the Agent Manager [227]

    1. Anti-gravity's Agent Manager allows multiple agents to work on the same project simultaneously [227].
    2. Users can spin up a new conversation, select a notebook to chat with, and have an additional agent working on the task [229].
    3. The user instructs one agent to use the MCP tool to grab a list of all accessible notebooks and create a document called "Notebook LM brain resources" on the left-hand side [231].
    4. The user confirms that the automated process successfully imported 48–49 resources into each of the five notebooks [243].

  2. Querying Notebooks with Specific Context [253]

    1. The user asks the agent to query the NotebookLM resources for relevant advice on starting a YouTube channel about English Springer Spaniels, including tips for making first videos and a capacity limit of three videos per week [254].
    2. The user creates a file called jack.md to store personal context, including the channel topic, video capacity, communication weaknesses, and the goal (funding a Springer Spaniel Rescue Center charity) [259].

  3. Updating the Skill to Reference Context [270]

    1. The user instructs a second agent to update the NotebookLM skill, requiring it to check the information within the jack.md file whenever referencing NotebookLM to ensure the response is contextual [273].
    2. After the skill is updated, the agent uses the MCP to query the notebooks and returns highly specific advice, including niche-down strategies, weekly content mix, hook mastery, and an example schedule [295].
    3. The user then instructs Anti-gravity to create an action plan called "YouTube strategy" based on this information [299].
    4. This demonstrates how the AI can query the resources and build out a plan through self-contained, iterative cycles [302].

4.2. Building Interactive Dashboards from Knowledge [305]

The knowledge contained within the notebooks can be leveraged to programmatically build interactive, visually engaging applications and dashboards using Anti-gravity.

  1. Requesting an Interactive Dashboard [305]

    1. Level three focuses on using the knowledge from NotebookLM to create and build something interesting, such as an application with Anti-gravity [307].
    2. The user asks the model to build an interactive, one-page HTML dashboard based on all the information gathered from the notebooks, emphasizing visual beauty and helpfulness [309].
    3. The core capability being highlighted is leveraging the notebook knowledge to build any part of an application [316].

  2. Querying Multiple Notebooks for Data [321]

    1. Anti-gravity queries different notebooks with specific questions to gather the necessary data for the dashboard [321].
    2. Examples of queries include:
      1. Extracting specific mathematical formulas, metrics, and benchmarks for YouTube growth (e.g., outline ratio, CTR targets) [323].
      2. Listing power words and psychological triggers for titles, and identifying visual elements that work best for thumbnails [325].
      3. Identifying specific engagement techniques [328].

  3. Reviewing the Dashboard Output [331]

    1. Anti-gravity finishes designing the interactive dashboard, which is titled the "Spring Spaniel Command Center" [331].
    2. The dashboard includes:
      1. Primary goals (three videos a week, watch time) and micro-niche focus [336].
      2. A growth bar where users can input video views and subscribers to detect "viral outliers" [342].
      3. A health check section for inputting metrics like click-through rate (CTR) and average view duration (AVD) to calculate "viral profit" [349].
      4. A checklist based on the "one man rule" that can be connected to a database like Superbase [355].
    3. The entire application is designed around the knowledge base contained in the notebooks [359].
    4. This capability can be used for business compliance (e.g., ISO 9000), where Anti-gravity queries the knowledge base to ensure website content is accurate [361].

5. Levels 4 and 5: Programmatic Content and Local File Uploads [367]

5.1. Level 4: Programmatic Content Generation [367]

Anti-gravity can programmatically generate various content formats—such as audio overviews and presentations—from the NotebookLM resources without manual interaction.

  1. Automating Content Creation [367]

    1. Level four focuses on programmatically creating slide decks, presentations, and audio overviews from within Anti-gravity [367].
    2. The user instructs Anti-gravity to use the NotebookLM MCP to generate an audio overview on the YouTube database specifically around "hooks," and also to create a presentation [370].

  2. The Efficiency of Automation [375]

    1. The MCP is configured to allow Anti-gravity to trigger the creation of audio overviews, mind maps, and reports, which are native NotebookLM functions [376].
    2. This is valuable because users can programmatically spin up resources on various topics (e.g., tax laws for moving abroad) and have the audio files ready to listen to without manually clicking buttons [378].
    3. The audio overview generation happens automatically after the request, without the user touching the keyboard [382].

  3. Scalability [386]

    1. Anti-gravity generates both the requested audio overview and the slide deck [387].
    2. This process can be done for all created notebooks, and users can create an infinite number of these notebooks [390].

5.2. Level 5: Uploading Local Files and Troubleshooting [394]

Level five enables users to upload local files, including complex formats like PDFs, from their computer into NotebookLM via Anti-gravity, ensuring all knowledge is centralized.

  1. Uploading Text Files [394]

    1. Level five is the ability to take files on mass from a local computer and create a notebook from them, which is useful for client meeting notes or business SOPs [394].
    2. The user instructs Anti-gravity to create a new NotebookLM notebook and add two existing local files (jack.md and notebook_brain resources) as resources [404].
    3. Anti-gravity automatically identifies the files, adds them, and confirms successful creation, demonstrating that any local text can be ingested [408].
    4. The user verifies that the new notebook contains the text from both files [422].

  2. Handling Complex File Types (PDFs) [428]

    1. While Retrieval Augmented Generation (RAG) is easy with text, PDFs can be challenging to process [430].
    2. The user downloads a PDF file and creates two new folders in Anti-gravity: notebook to add (for files waiting to be processed) and notebook added (for completed files) [439].
    3. The user drags and drops the PDF into the notebook to add folder [443].
    4. The user instructs Anti-gravity to add every file within the notebook to add folder into the notebook and then move the file to the notebook added folder when complete [451].
    5. If Anti-gravity struggles with the PDF, it can create a script to translate the PDF into text, which can be turned into a skill for future automation [465].
    6. This system allows for the automated ingestion of documents that change regularly or documents created by other tools and systems [472].

  3. Successful PDF Conversion and Upload [476]

    1. Since the PDF could not be loaded directly, Anti-gravity converted it to text and then uploaded it, using a Python virtual environment and other processes [476].
    2. The user confirms that the "Anti-gravity School Creator PDF" has successfully arrived in NotebookLM [479].
    3. If users encounter problems with unofficial MCPs or specific document types, Anti-gravity can help troubleshoot and build a system to ensure any file can be transferred [482].
    4. The final step is ensuring users leverage Anti-gravity's skills to maximize the platform's capabilities and develop systems faster [486]. </Document A> </input_data>