• Everyday AI
  • Posts
  • Ep 627: NotebookLM: New features, what’s next and complete walkthrough

Ep 627: NotebookLM: New features, what’s next and complete walkthrough

Inside NotebookLM's recent updates, NVIDIA to invest $2 billion in xAI, OpenAI’s trillion dollar compute lineup, Google’s new model uses the computer and more

Author

Everyday AI
October 08, 2025

Sup y’all 👋

Random… does your company have an official AI policy?

Does your company have an official AI policy?

🗳️ VOTE to see live results 🗳️

Login or Subscribe to participate in polls.

✌️

Jordan

(Let’s connect on LinkedIn. I’d love to hear your company’s AI policy.)

P.S. — You REALLY gotta see what we cooked up at the end of today’s newsletter. Seriously. We should charge a buncha money for this resource but it’s free 99.

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: MORE new NotebookLM features? Yup. You voted for the coverage, so we dish the details. Give it a watch/read/listen.

🕵️‍♂️ Fresh Finds: OpenAI releases Sora 2 prompting guide, Google Deepmind’s agent that automagically fixes bugs, xAI releases impressive AI video update and more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: NVIDIA to invest $2 billion in xAI, OpenAI’s trillion dollar compute lineup, Google’s new model uses the computer and more. Read on for Byte Sized News.

💪 Leverage AI: Cool bro. New NotebookLM features. We get it. But here’s why you should pay attention and how it’ll help. Keep reading for that!

↩️ Don’t miss out: Did you miss our last newsletter? We talked about Inside ChatGPT's new Agent builder, Deloitte in hot water after AI missteps, AI bubble fears on circular funding trends, IBM and Anthropic team up and more. Check it here!

Ep 627: NotebookLM: New features, what’s next and complete walkthrough

Have you been sleeping on NotebookLM? 😴

If so, you're leaving hours of productivity (and probably a lot of money) at the door. 

But real talk -- the team is shipping fast. The NotebookLM you met last year from the viral Audio Overviews is not the NotebookLM of today. 

It's slowly turned into a robust, multimedia powerhouse. And the last feature update only adds to its utility. 

If you missed it. Don't worry. We're putting AI to work on Wednesday with the newest updates (and a complete walk through!) of NotebookLM. 

Also on the pod today:

• Shorter, snappier AI responses ⚡
• AI-generated mind maps explained 🧠
• Interactive quizzes and flashcards 📝

It’ll be worth your 39 minutes:

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – QA Tech is AI agents for QA, AI AppGen by Retool goes from idea to app, Promptius is AI agents built with natural language.

AI Coding — Google is releasing its Opal vibe coding platform in 15 more countries.

Model Alignment — Anthropic’s just-released Petri automates audits to uncover subtle AI misalignment

AI coding — Google’s DeepMind unveiled CodeMender, an AI agent that auto-detects and patches vulnerable code. Here’s why that matters.

AI Video — xAI released Imagine v0.9, their new AI video tool.

 

Agentic AI — MIT-IBM Watson AI Lab released Toucan: 1.5M real tool-use trajectories that boost agent performance.

Claude Code — Claude Code may be rolling out soon to mobile apps.

AI Robotics — Softbank is spending $5 billion plus on AI robotics.

AI and National Security — OpenAI shut down several ChatGPT accounts tied to Chinese and Russian-linked actors. Find out why.

AI Deepfakes — Robin Williams’ daughter blasted AI deepfakes of her late father as “gross,” urging people to stop sending them

 

1. Google debuts Gemini 2.5 Computer Use in public preview 🖥️

According to Google, the new Gemini 2.5 Computer Use model is now live in public preview via the Gemini API, promising faster and more accurate automated UI control for web and mobile than leading rivals.

Built on Gemini 2.5 Pro’s visual reasoning, it operates in a loop using screenshots and action history to click, type, scroll, and even request user confirmation for high-stakes steps, with safety guardrails like per-step reviews and system instructions. Early users and internal Google teams report major gains in speed, reliability, and recovery of failed workflows, while benchmarks from Browserbase and Google show strong browser control at lower latency.

2. IBM launches AI-first Project Bob IDE in private preview ⚒️

IBM unveiled vibe coding tool Project Bob at TechXchange to automate coding tasks across editing, testing, documentation, security and CI/CD, with Anthropic’s models integrated and governance tools like AgentOps extending to mainframes.

Reported early results show 6,000 testers with 45% average productivity gains, and Gartner expects 40% of enterprise software to use vibe coding by 2028. For day-to-day builders and growing teams, this signals faster prototyping with more built-in oversight, but it also means leaders must balance speed with code quality, security and clear accountability.

3.JPMorgan’s $2B AI Bet Pays For Itself 🏦

According to Bloomberg TV, Jamie Dimon said JPMorgan spends about $2 billion a year on AI and is seeing roughly the same amount in savings, calling it only the “tip of the iceberg.”

Dimon noted hundreds of AI use cases already in play and touted longer-term benefits, from helping cure cancer to shorter work weeks, while acknowledging some jobs will be displaced. He added consumers look stable but a recession could happen in 2026, and he criticized government shutdowns while pointing to strong merger activity.

4. OpenAI’s trillion-dollar compute spree raises eyebrows 😲

According to THE DECODER, OpenAI has inked roughly $1 trillion in compute contracts with Nvidia, AMD, Oracle, and CoreWeave, a timely move that dwarfs its current finances and could mean a loss of about $10 billion this year.

The deals feature circular funding arrangements like Nvidia’s planned $100 billion investment that OpenAI would largely spend on Nvidia chips, while AMD is offering warrants to buy up to 10 percent of its shares at a token price, and partner stocks have already jumped. Analysts warn the company lacks cost discipline as it chases unprecedented scale, a strategy that makes sense only if revenue, currently about $12 billion annually, grows much faster.

5. xAI eyes $20 billion in fresh funding as Nvidia circles the deal 🤑

According to Bloomberg, Elon Musk’s xAI is reportedly pursuing a $20 billion financing package using a special-purpose vehicle that would buy Nvidia chips and rent them back to the startup, with Nvidia potentially investing up to $2 billion in the equity slice.

The structure, which blends $12.5 billion in debt with $7.5 billion in equity and is backed by GPUs rather than the company itself, spotlights the creativity and urgency behind AI firms racing to lock down high-end compute. This continues Nvidia’s aggressive ecosystem strategy after its commitments tied to OpenAI, Intel, and CoreWeave, raising fresh questions about “circular” financing while underscoring how central the chip maker has become to AI’s supply chain.

🦾How You Can Leverage:

That's because OpenAI’s been stealing the AI headlines all week. 

So what did Google unleash inside NotebookLM?

Now, you can customize AI responses and length, which fixes the nightmare scenario of asking NotebookLM a simple question and getting a novel’s worth of words back in response. 

That’s right — one of the common annoyances of NotebookLM has been squashed and now the AI cheat code feels truly personal. 

So on today's show, we're breaking down the new customization features that actually work, the learning guide that makes you retain information, and why September plus October updates combined make NotebookLM absolutely untouchable right now.

Let’s dive in. 

1 – Grounded AI means zero hallucinations 🔒

New to the NotebookLM hoopla? We gotchyu. 

It only works with your uploaded sources.

Period.

Upload 50 white papers and ask who won the World Series? It'll say it has no clue because that's not in your documents.

This seems limiting until you realize it's legit the most powerful business feature that exists.

You can't accidentally get Reddit threads mixed into your proprietary strategy analysis. You can't get outdated blog posts injected into your competitive intelligence. You can't get random internet wisdom contaminating your client research.

NotebookLM won't even start chatting until you upload sources.

Your team can dump the entire content library into one notebook and asks strategic questions knowing you'll get pure analysis of your actual materials with receipts for every claim.

Try This

Create a notebook and upload every doc from your most important current project. 

Client briefs, internal memos, competitive analyses, meeting notes. 

Ask something complex like "what are the three biggest risks across all these sources and which stakeholders mentioned each one?" Watch it synthesize with citations. 

Click those numbers and verify the exact source material. This is how AI should work for business applications where accuracy actually matters instead of vibes-based responses mixing your data with random internet content.

2 – Customization makes it your learning partner 🎯

October's NotebookLM updates finally delivered truly personalized AI.

Custom response style and length while chatting.

This is the breakthrough everyone missed. Your competitive intel notebook can be direct with zero fluff. Your learning notebook can be detailed and patient. Your client research can be thorough and professional.

And they don't bleed into each other.

We tested this live. Same question twice. Default settings gave us a massive wall of text. 

Custom settings saying "be direct, don't waste words, keep it short" cut the response by s75%. 

The learning guide conversational style is the other massive October update.

Instead of just answering questions, it asks YOU questions first. You say "recap this report" and it responds asking what your goal actually is. Business implications? Technical details? Strategic positioning?

This forces active thinking instead of passive scrolling.

Your competitor launches a new product. You dump all their materials into NotebookLM. Switch to learning guide mode. It makes you clarify what actually matters before information overload hits.

Then it quizzes you. Makes you explain concepts back. Forces real retention instead of skimming and forgetting everything ten minutes later.

Plus it auto-generates flashcards, quizzes, and mind maps from any document set.

Try This

Create three notebooks for three different work contexts. Client presentations, competitive intel, personal learning. 

Click configure notebook in each and write exactly how you need it to respond. Something like "professional and thorough" for clients versus "direct AF, zero fluff" for competitive research. 

Upload one doc to each and ask the same question across all three. The response differences show you why one-size-fits-all AI communication is fundamentally broken for how humans actually work throughout the day.

3 – September plus October equals untouchable 🚀

October's customization drop is huge.

September brought custom audio overview formats. The brief, the critique, the debate, the deep dive. All with customizable length and tone. Flashcards and quizzes with instant feedback. Dynamic report templates that adapt to your source material.

July's updates that we covered in episode 578 added video overviews that are legitimately mind blowing. Multi-version creation. Multitasking while listening to audio overviews.

All grounded in your actual documents with zero hallucinations.

Beutimous. 

Try This

Open NotebookLM and explore the full workflow with a real project. 

Upload multiple sources on the left using whatever format your materials are in. Configure the chat in the middle with custom style for that specific use case. 

Then use the studio on the right to generate an audio overview using the brief format, create a mind map, and generate flashcards. 

Experience how all three panes work together to turn any document set into an interactive learning environment that actually adapts to how you need to consume information instead of forcing you into one communication style forever.

 🎁 Bonus Content 🎁

Not gonna lie… it would be kinda foolish to not repost today’s LinkedIn livestream.

We’ve created about 30 ‘Bonus Guides’ over the past 3 years and this is by far the most useful.

Reply

or to participate.