NemoClaw Knowledge Wiki

New Gemini Tutorial

4 min read

New Gemini Tutorial

https://www.youtube.com/watch?v=cwai8SjMZWM Ali H Salem Here is a comprehensive guide and summary of the Gemini tutorial video, structured by feature.

The Complete Guide to Google Gemini (2025 Edition)

This tutorial covers the end-to-end functionality of the Gemini application, highlighting its native integration into the Google ecosystem and its advanced reasoning capabilities.

1. Foundations & Interface

  • Workspace Integration: The strongest differentiator for Gemini is its native integration with Google Workspace. By typing @, you can directly call tools like Gmail, Google Calendar, Docs, and Drive.
    • Example: @Google Calendar schedule a meeting at 16:00 tomorrow for 'Deep Work'.
  • Models:
    • Fast Model: For quick, lower-complexity tasks.
    • Gemini 3 (Reasoning): The recommended model. It uses a “thinking level” variable to adjust effort based on prompt complexity.
  • Context Window: Supports up to 1 million tokens, allowing you to upload entire books, massive codebases, or multiple large reports for analysis.

2. Deep Research

Gemini’s Deep Research goes beyond standard web search by allowing you to cross-reference specific user data (Gmail, Drive, Uploaded Files) with the web.

  • Gap Analysis: You can upload a specific report (e.g., a PDF) and ask Gemini to perform a “Gap Analysis” against external web benchmarks.
  • Process: It generates a research plan executes the research compiles a structured report.
  • Output Options: Once research is complete, you can use NotebookLM features to convert the report into a podcast (Audio Overview), study guide, or FAQ.

3. Gemini Gems

Gems are custom, reusable “mini-experts” or personas designed for specific recurring tasks.

  • How to Build:
    1. Name: e.g., “Executive Gem”.
    2. Instructions: Define the persona and output style (e.g., “You are an executive communications specialist. Turn input text into concise executive briefs.”).
    3. Knowledge: Upload reference files if needed.
  • Benefit: Saves you from re-prompting the persona every time you need a specific type of output.

4. Video & Image Generation

  • Video (Veo 3.1): Generates high-quality video with sound.
    • Capabilities: Handles multiple sequences and cinematic instructions well (e.g., specific camera angles, lighting).
  • Images (Nano Banana Pro):
    • Creation: Generates high-fidelity images.
    • Editing & Continuity: The standout feature is context continuity. You can ask it to edit an image (e.g., “Switch the menu theme to forest green”) and it will modify the style while retaining the exact layout and text content of the previous image.

5. Canvas (Built-in Workspace)

“Canvas” is a workspace mode that outputs structured, editable documents rather than a chat stream.

  • Use Case: creating dashboards, reports, or web pages.
  • Iterative Building: You can highlight specific sections (e.g., a chart) and ask Gemini to expand on them or move them into a separate tab/dropdown.
  • Result: A visually rich, HTML-style dashboard that is interactive and presents data much better than a wall of text.

6. Guided Learning

Ideal for digesting complex corporate documents or studying new topics.

  • Scaffolded Learning: Instead of dumping information, Gemini introduces a topic simply and asks where you want to go next.
  • Workflow: Upload a complex PDF Activate “Guided Learning” tool Ask “Where do I start?“.
  • Interaction: Gemini provides a high-level summary and distinct paths/buttons to explore specific sub-topics deeper.

7. Visual Layouts

Transforms text-heavy explanations into magazine-style layouts with interactive widgets and images.

  • How to use: Enable “Visual Layout” in tools and ask for an explanation (e.g., “Explain how LLMs work”).
  • Output: Instead of text paragraphs, you get a graphical breakdown with steps (e.g., Training Tokenization Prediction) and visual aids.

8. Scheduled Actions

Automates recurring prompts to run at specific times.

  • Setup: Simply ask in chat: “Create a weekly scheduled action. Every Monday at 8:00 AM, send me a summary of the latest AI updates.”
  • Management: You can toggle, edit, or delete these automations in Settings > Scheduled actions.

9. Prompt Engineering (Gemini 3 Pro)

Because Gemini 3 is a reasoning model, prompt engineering differs slightly from older models:

  1. Be Concise: It performs best with direct, clear instructions rather than verbose, over-engineered prompts.
  2. Specify Output Verbosity: By default, it is concise. If you want a chatty or elaborate response, you must explicitly ask for it.
  3. Context Placement: Place your specific question or instruction at the end of the prompt, after the data/context, to prevent the model from getting lost in large datasets.

10. Key Settings

  • Instructions for Gemini: Set a global baseline instruction (System Prompt) that applies to every chat (e.g., “Always be concise, direct, and MECE”).
  • Apps: Manage connections to GitHub, YouTube Music, and Google Workspace.
  • Public Links: Manage or delete links to chats you have previously shared publicly.

Graph View