New NotebookLM Video Update is INSANE!
Julian Goldie breaks down the major NotebookLM updates from late 2025 through April 2026, including the new cinematic video overviews powered by three AI models. He then explains how he pairs NotebookLM with the open-source Hermes agent and Obsidian to build a persistent, personalized content pipeline for his AI Profit Boardroom community.
Summary
The video opens with Julian Goldie positioning NotebookLM as having evolved from a simple note-taking and podcast tool into a full research-to-content studio. He walks through eight major updates spanning roughly seven months: the 1-million-token context window expansion in October 2025, the Deep Research feature launch in November 2025 (enabling autonomous source-finding rather than relying on manually uploaded files), support for new file types like Google Sheets and CSVs, the switch to Gemini 3 as the core reasoning model in December 2025, and the headline update of March 2026 — cinematic video overviews.
The cinematic video overview feature is described as a fundamentally different product from earlier narrated slideshows. It uses three AI models in sequence: Gemini 3 as the creative director determining narrative arc and pacing, Imagen (referred to in the transcript as 'Nano Banana Pro') for visual asset generation, and Veo 3 (referred to as 'VO3') for final animated output. This feature is currently exclusive to Google AI Ultra, the paid premium tier. Additionally, in April 2026, Google shipped three updates in 24 hours: auto-labeling of sources, bulk sharing, and improved flashcard memory, along with a new three-column interface that allows one-click generation of multiple content formats simultaneously from the Studio panel.
Julian then identifies the core problem this creates: NotebookLM generates vast amounts of content but has no persistent memory, no organization layer, and no sense of the user's voice or brand. Every notebook starts fresh. To solve this, he introduces Hermes Agent, a free open-source autonomous AI agent launched in February 2026 by a lab called News Research, which reached 135,000 GitHub stars in under three months. Hermes is distinguished by cross-session persistent memory, the ability to create reusable 'skill documents,' and native support for MCP (Model Context Protocol), allowing it to connect to NotebookLM without custom code.
Julian describes his specific workflow for the AI Profit Boardroom: he feeds member questions into NotebookLM's Deep Research, generates podcasts, infographics, and mind maps from the Studio, and then Hermes picks up all outputs, organizes them into a searchable media library, and applies persistent knowledge of his tone, audience, and prior content. Obsidian serves as the memory vault, automatically storing every generated asset and feeding that context back into future content creation, visible as a knowledge graph. The video concludes with a pitch for the AI Profit Boardroom, which offers pre-configured versions of this entire stack including Hermes, the Obsidian vault, prompts, a 30-day roadmap, and live coaching calls.
Key Insights
- Julian explains that NotebookLM's cinematic video overviews use three AI models in sequence — Gemini 3 as creative director, a visual asset generator, and VO3 for animated output — meaning the final video is produced through a multi-model pipeline triggered from a single click on uploaded documents.
- Julian argues that the November 2025 Deep Research update fundamentally changed NotebookLM's nature, shifting it from a passive summarizer that only works with manually uploaded sources to an active research agent that autonomously browses hundreds of websites and compiles citation-backed reports.
- Julian claims that Hermes Agent's defining differentiator over other AI tools is cross-session persistent memory — it builds a cumulative model of the user's projects, preferences, and voice over time, and also writes reusable 'skill documents' each time it solves a complex problem so performance improves with use.
- Julian asserts that the memory layer — specifically using Obsidian as a vault that Hermes reads before generating anything new — is what prevents AI-generated content from sounding generic, because it feeds the system an evolving model of the user's topics, voice, and audience rather than starting fresh each time.
- Julian contends that the reason most people fail to implement AI workflows is not tool complexity — Hermes installs in one command, NotebookLM is free, and Obsidian takes five minutes to configure — but rather the absence of a pre-built system with workflows, configuration files, prompts, and structure ready on day one.
Topics
Transcript
[0:00] New Notebook LM video update is insane. Have you seen what Notebook LM just did? Because if you blinked, you miss it. We went from podcast overviews to full cinematic AI videos. From a note-taking app to a complete research to content studio. Most people are still using it like it's 2024. The digital avatar of Julian Goldie. I help people actually learn and use AI tools in their work. And today I'm breaking down every major Notebook LM update from the last several months, why it matters, and exactly how I've wired it into Hermes agent to build a content pipeline for the AI Profit Boardroom that would have taken a full [0:30] team to run just a…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Julian Goldie SEO
NEW Nvidia Autonomous AI is WILD!! 🤯
Nvidia announced Nemo Clo, a new autonomous AI agent system that operates independently without continuous prompting. Powered by Nemotron 3 Ultra (a 550 billion parameter model), the system is five times faster and cheaper than previous versions, with OpenShell providing secure sandboxed execution.
Laguna XS 2.1: New FREE + Opensource Local AI!
Julian reviews Laguna XS 2.1, a new free open-source local AI coding model from Poolside that performs comparably to Qwen 3.6 and outperforms Claude Haiku on benchmarks. He demonstrates its practical capabilities by building landing pages and functional apps, highlighting its speed, offline functionality, and multiple deployment options through local setup, Claude Code, or OpenRouter's free API.
How to Run Hermes FREE Forever!
The video demonstrates how to run the Hermes AI agent for free using Gemma 4, a local open-source model from Google, with significant speed improvements through MLX optimization. The setup works on Apple Silicon Macs or via free APIs on Open Router, enabling autonomous agents to work offline and privately without subscription costs.
This NEW Chinese AI is INSANE! (FREE + Open Source!)
Long Cap 2.0 is a new open-source Chinese AI model from a food delivery app company that offers 1 million tokens of free context memory, beats GPT-4.5 on SWE bench pro benchmarks, and uses efficient parameter activation to reduce computational overhead while maintaining high performance.
Claude Code is now FREE: Here’s how…
Google's new Gemma 4 model running on Ollama is 90% faster on Apple Silicon, enabling free Claude Code usage locally without token costs. The setup requires three simple steps: downloading Ollama, Gemma 4, and installing into Claude Code, with alternatives available via OpenRouter API for non-Mac users.