NotebookLM Just Got Automations: Build POWERFUL AI Automations!

Summary (Tech concepts & how-to)

The video demonstrates how to combine NotebookLM (for research) with Codex (for automation/execution) to create an “end-to-end” AI workflow. The system continuously updates knowledge in NotebookLM and then turns that knowledge into actionable outputs (e.g., emails and content).

---

1) Why combine Codex + NotebookLM?

• Codex is positioned as strong for development and everyday work, but weaker for research.
• NotebookLM is positioned as the better research tool, including:
  • Deep research
  • Finding high-quality resources
  • Grounding answers in provided sources
• The overall goal:
  • Use NotebookLM for knowledge/research
  • Use Codex as the automation layer to keep everything updated and produce deliverables

---

2) Setup requirement: make Codex and NotebookLM communicate

Codex gets a new Chrome extension capability that allows it to control/connect with what’s open in the browser.

Steps shown in the demo

1. Enable Codex plugin support:
   • Install/enable “Chrome control” plugin
   • Confirm the browser extension shows as connected
2. Connect Codex to a specific NotebookLM notebook:
   • Copy the notebook link
   • In Codex, instruct it to connect to that notebook
   • Look for a confirmation message indicating the notebook is open in Chrome and ready for adding sources

---

3) Automation example: daily source ingestion into NotebookLM

Use case: Automatically build a recurring “database” of a topic/channel’s latest information.

Example: tracking a YouTube channel

• Every day at 9:00 a.m.
• Codex automatically:
  • Opens the browser
  • Navigates to the YouTube channel
  • Finds the latest video
  • Retrieves the link
  • Adds that link as a new source in the connected NotebookLM notebook

The demo shows the automation running and NotebookLM receiving the newest upload as a source without manual steps.

Where to find automation details in Codex

• Automation list/status
• Scheduling details (daily repeat, next run time)
• Model/reasoning/debug metadata

---

4) Limitation of NotebookLM (and workaround)

NotebookLM currently has limited “output formats” available from its UI (examples mentioned):

• audio
• slide deck
• video
• mind map
• report

The presenter’s argument: sharing raw/comprehensive notebooks or long reports isn’t always useful for teams/clients—hence the value of using Codex to turn research into practical outputs.

---

5) Next automation step: turn NotebookLM insights into an email (via Gmail)

Use case: Convert “new research” into a ready-to-send business message.

Setup

• Enable additional Codex plugins, specifically Gmail
• The video also mentions other possible integrations such as:
  • Google Drive
  • Teams
  • Outlook

Workflow shown

1. In NotebookLM, create a draft/command
2. Use Codex + Gmail plugin to generate an email draft based on:
   • the latest learning/takeaways from the newest NotebookLM sources (e.g., newest video)
3. Codex requests the recipient/business email details
4. Codex produces a Gmail draft, including:
   • a subject like “latest learnings…”
   • content summarizing key themes
   • source inclusion for traceability

The presenter confirms the draft appears in Gmail and the email content is ready.

---

6) Suggested “pipeline” (ideas for custom workflows)

The video suggests chaining automations into a pipeline:

1. NotebookLM maintenance automation
   • Continually ingest sources such as websites, channels, topics, competitors, stocks
2. Work-product automation
   • Email team/clients with key insights
3. Optional expansion:
   • Turn gathered knowledge into content pipelines (e.g., tweets)
   • Automate posting/creation with Codex after NotebookLM research

---

Main speakers/sources

• Primary speaker: The YouTube creator “Universe of AI” (referenced repeatedly in the demo and associated with their channel/newsletter)
• Tools/secondary references:
  • Codex
  • NotebookLM
  • Chrome (extension control)
  • Gmail plugin (also mentions Google Drive/Teams/Outlook as possible integrations)