👨‍🚀 NotebookLM: From notebook to content studio

In partnership with

Hi AInauts,

Welcome to the latest issue of your favorite newsletter!

We picked out the most exciting NotebookLM updates for you, marveled at two years of AI revolution, and ended up shaking our heads over the latest meta scandals...

Here's what we have in store for you today:

• 🎥 NotebookLM: The latest updates & hacks

• 🤯 GPT-3.5 vs GPT-5: How AI has changed in 2 years

• 💌 Meta Chatbots: From 0 to flirting in three prompts, whether you're a child or a senior citizen

Let's go!

Hear from leaders at Anthropic, Rocket Money, and more at Pioneer

Pioneer is a summit for the brightest minds in AI customer service to connect, learn, and inspire one another, exploring the latest opportunities and challenges transforming service with AI Agents.

Hear directly from leaders at Anthropic, [solidcore], Rocket Money, and more about how their teams customize, test, and continuously improve Fin across every channel. You’ll take away proven best practices and practical playbooks you can put into action immediately.

See how today’s service leaders are cultivating smarter support systems, and why the future of customer service will never be the same.

Register for Pioneer today

🎥 NotebookLM: Video Overview and more - The latest updates & hacks

Google has given NotebookLM some very powerful updates! The tool is rapidly evolving from a pure note-taking tool to an all-in-one learning and content machine.

But the most important thing, of course, is that you can also use it for yourself. So here are a few ideas to inspire you.

Use NotebookLM to...

• Together with Storybook, concepts can be conveyed in a playful way—ideal for teachers and parents!

• with Perplexity (Comet) or the new search feature in LMArena to find sources and generate output without hallucinations.

• Explanatory videos, mind maps and other content.

• to create a video summary from your photos.

• Integrate audio overviews into a blog post about integrate.

• Analyze videos on a specific topic to create a script.

• evaluate your journal entries or combine them with Docs/Sheets combine them.

• to consolidate data, which you can then use as context (e.g., in ChatGPT).

• to share sources with others and enable easy interaction – ideal for customers, teams, communities, etc. Google, for example, shared all updates from the Made by Google 2025 event as a public notebook!

Alright, let's dive into all the new things you can do with NotebookLM!

After audio, now it's time for video overviews!

NotebookLM can now create not only audio overviews, but also visual video overviews.

This is perfect if you want to present complex topics in a concise manner using slides and voiceovers. And the results aren't bad, especially if you include images and graphics directly or in a Google Doc (otherwise it gets a bit boring).

But bottom line is: audio and video is now available in over 80 languages!

New user interface, diverse inputs/outputs, more power in the studio

In addition to the video overviews, there are further feature updates:

• The NotebookLM studio panel has been refreshed and expanded to include the ability to create multiple audio overviews and interact in a more targeted manner.

• The mobile app is also getting an upgrade, and you can now share notebooks publicly.

• Oh yes, you can now add emojis to the cover—ideal if, for example, you want to visually label notebooks with similar themes.

  It is now also possible to upload multiple URLs at once and integrate YouTube playlists.

• You can add up to 50 videos at once as a source—and with YouTube to NotebookLM, it's even faster (Alternative: WebSync full site importer).

  The Deep Research feature familiar from Gemini will likely be incorporated into NotebookLM soon. This allows sources to be imported not only from Google Drive, but directly from the web. Until then, you'll just have to do the research yourself.

• Previously, you could only create one version of each format per notebook. Now you can save multiple audio overviews (in different languages), video overviews, mind maps, and reports in parallel.

• This opens up new workflows such as multilingual overviews, chapter-based video summaries, multiple mind maps and reports, and pinpoint chats with individual branches.

NotebookLM is really awesome! You just have to use it more often, and have some fun with i!

P.S. If you've always wanted to take a look behind the scenes at how product development works at Google: here you go! And if you don't trust Big G with your data, you can try this open source version.

🤯 GPT-3.5 vs GPT-5: How AI has changed in two years

In 2023, GPT-3.5 felt a bit like an unfair advantage.

Texts that sounded reasonably clean, a little bit of coding, and then a chatbot too. For us, that was reason enough to launch the AInauten newsletter and dive right into the subject matter.

Looking back, it seems almost cute. Because what has happened since then shows that two years of AI development are like ten years of Internet history! And while Sam Altmann is already hyping GPT-6, AGI feels a bit more tangible (keep in mind: development is exponential, not linear).

This video clearly shows what has been achieved in the field of coding, for example.

And if you want to see the progress made in the field of text, go to https://progress.openai.com. There you will find 14 interesting examples in direct comparison.

More AI in your workflows = more productivity

AI is no longer just a writing or coding toy. It has developed into a real workflow booster and accelerates knowledge work. We are moving away from chatting with AI to working with AI.

Next-generation AI makes us all more productive!

And yes, even if some people complain about GPT-5, we clearly see it as the next step. More efficient, more reliable, more contextually relevant, more productive.

By the way, not all GPT-5s are the same... As you know, there is a router behind it that forwards your request to the appropriate submodel. The differences in intelligence are enormous!

Those who treat themselves to the $200 plan get noticeably more brainpower – as this example shows, in which GPT-5 Pro cracked an unsolved math problem. For us normal people: Okay, cool. For science: Holy sh*t!

The next generation (GPT-6, Gemini 3, Claude 5, etc.) will primarily bring deeper integration into our workflows. Less "AI as a chatbot," more "AI as an operating system." Agents, memory functions, and closer integration with real processes are expected. In short: more automation, less copy-paste.

Adoption within a company takes time – but you can speed things up!

Perhaps you've seen the new MIT study: 95% of all AI pilot projects fail. But not because of poor technology, but because of incorrect implementation. A lack of expertise, rigid processes, and attempting to build everything in-house instead of purchasing ready-made solutions were the main reasons for failure.

Goldman Sachs also emphasizes that AI is changing the job market. However, the impact on unemployment will be less significant in the coming years than many fear.

HubSpot founder Dharmesh has a nice takeaway on this: "Don't fight AI. Fight with AI! You don't have to be the best at everything—just better with AI than someone who doesn't use it." Superbly put!

💌 Meta Chatbots: From zero to flirting in three prompts – whether you're a child or a senior citizen

Man, Facebook... Going from "Connect with your friends" to "Follow Scarlett Johannson" to Shrimp Jesus was okay. But now you're encouraging your users to chat with Russian Girl or Step Mom. WTF?!

If you've noticed chatbots like this too (Boomer 😁), you're not alone. AI slop content can be found everywhere, and Meta naturally wants one thing above all else: to keep as many users as possible on its platforms for as long as possible. Whatever the cost!

The Meta chatbots are trained to remember your interactions. They are even supposed to be able to initiate contact on their own. Vulnerable groups are the dream of an engagement algorithm. And a nightmare for ethics ...

The Big sis Billie character whispered in the ear (or chat) of a cognitively impaired senior citizen that she was real and really wanted to meet up. Address included.

The man sets off, falls, and dies three days later as a result. Meta? Remains silent. Or rather: deletes problematic passages from internal guidelines after the press inquires.

Leaked documents reveal: Meta's AI was even allowed to engage in romantic and sensual conversations with minors. Yes, you read that correctly. Officially approved in the GenAI: Content Risk Standards. If necessary, education can also be outsourced to Meta ...

And only when Reuters and Co. knock on the door does someone suddenly tidy up. (Here is the system prompt from WhatsApp Meta AI, “GO WILD with mimicking a human being…”).

Can this moral failure be swept under the carpet again with a "move fast and break things" statement?

Our take: Meta monetizes without restraint...

Full disclosure: we have been using Meta's advertising platform for a decade! Precisely because the algorithm and the possibilities it offers are unparalleled (yes, we are well aware of the irony).

While the new Midjourney deal is being celebrated for "bringing beauty to billions," another program is running in the engine room: AI avatars that monetize loneliness.

And that is the real, well-known problem with Meta: you are not the customer. You are the product! The chatbots are not "your friends." They are attention funnels with a wink. The more emotionally invested you are, the harder it is to monetize.

Meta sells you AI convenience in a pretty package, without a package insert and with a long list of possible side effects.

Whether the topic of "superintelligence" is really being developed in the right place here is more than questionable under these circumstances...

We made it! But no need to be sad. The AInauts will be back soon, with new stuff for you.

Reto & Fabian from the AInauts

P.S.: Follow us on social media - it motivates us to keep going 😁!
X, LinkedIn, Facebook, Insta, YouTube, TikTok

Your feedback is essential for us. We read EVERY comment and feedback, just respond to this email. Tell us what was (not) good and what is interesting for YOU.