Google finally gives NotebookLM users a way to breathe

Google hasn’t even finished rebranding NotebookLM, and it’s already fixing one of the biggest pain points for heavy users: a cluttered dashboard.

Starting this week, Gemini Notebook Collections are rolling out to everyone. The new tab lets you group related notebooks while keeping everything visible under My Notebooks. No forced reorganization. No rigid folder trees. Just a little structure for the chaos.

The timing makes sense. Google is tying the former NotebookLM more tightly into Gemini and Google Search, turning a standalone research tool into something that follows you across the company’s AI products. More places for notebooks means more notebooks. And more notebooks means you need a way to sort them.

Collections act like labels, not folders

The core idea is refreshingly simple. A notebook can belong to multiple Collections at once. That research project on climate policy could sit under both “Work” and “Environment” without duplicating a single file or forcing you to pick one permanent home.

Think of Collections like photo albums or playlists. They’re flexible by design. Anything that doesn’t need sorting stays ungrouped under My Notebooks, so your existing library doesn’t get turned upside down.

Google’s own announcement on X put it plainly: “zero rigid folder structures with maximum flexibility.” You can add notebooks to as many collections as you want, or skip the whole system entirely if it doesn’t fit your workflow.

Why this matters for heavy researchers

If you’ve ever stared at a wall of 50 notebooks with no way to tell them apart, you know the pain. Collections give crowded dashboards some breathing room without introducing a complicated filing system you’ll have to maintain forever.

The setup is deliberately basic. That’s a feature, not a bug. Google isn’t asking you to reorganize everything — just to group what needs grouping.

Organization becomes critical as Notebooks spread

This update isn’t happening in a vacuum. Notebooks can now appear inside the Gemini app, where you can work with uploaded material alongside your previous chats and custom instructions. Changes sync with the separate Gemini Notebook service, so the same project continues seamlessly across both products.

Google is also pulling Gemini Notebook into Search results. With notebooks gaining more places to appear and more ways to accumulate information, leaving them in one increasingly crowded library would become frustrating fast.

Collections add an organizational layer before that expansion makes the dashboard impossible to navigate. Smart timing, honestly.

What Collections still don’t give you

Let’s be clear about the limits. Google hasn’t announced nested Collections, automatic sorting, or anything resembling a full file-management system. You can group notebooks, but you can’t build a detailed hierarchy or hand the cleanup over to Gemini.

That leaves Collections with a narrow job: divide a growing library into useful groups without duplicating projects or removing anything from the main dashboard.

For anyone already drowning in notebooks, that may be enough. But if you were hoping for the control of a proper folder system with subfolders and drag-and-drop organization, you’ll find this pretty basic.

How to start using Collections today

Ready to tidy up? Here’s what you need to know:

• Open Gemini Notebook and look for the new Collections tab at the top of your dashboard.
• Create a collection and give it a name — anything from “Q3 Research” to “Recipes I’ll never cook.”
• Add notebooks by selecting them from your existing library. A notebook can live in multiple collections simultaneously.
• Leave ungrouped notebooks alone if they don’t need sorting. They’ll stay visible under My Notebooks.

The rollout is gradual, so if you don’t see the tab yet, check back in a few days. Google typically pushes these updates to all users within a week or two.

For more on how to get the most out of Google’s AI tools, check out our guide on using Gemini for research projects and tips for organizing your digital workspace.

Why everyone’s missing the real story about AI progress

Whenever a new AI model drops, the hype machine goes into overdrive. We hear about how much smarter it is, how quickly it answers, how photorealistic its outputs have become. But here’s the uncomfortable truth: none of that matters if the AI can’t reliably plug into the apps and services you use daily.

That’s why an upcoming update to the Model Context Protocol (MCP) caught my eye. It’s not a flashy new chatbot or a breakthrough model. Most people will never even notice it’s happening. Yet it could quietly make the entire AI ecosystem far more robust.

If you’ve never heard of MCP, think of it as a universal translator for AI. It lets assistants like ChatGPT, Claude, or Gemini safely talk to Gmail, Slack, calendars, databases, and countless other tools. Instead of every company inventing its own proprietary bridge, MCP provides a shared rulebook.

The problem isn’t the AI — it’s the plumbing

It’s tempting to assume AI only improves when companies release beefier models. In practice, many of today’s growing pains have nothing to do with intelligence. They’re infrastructure issues.

Imagine calling a friend every few minutes and having to reintroduce yourself each time. That’s roughly how many AI services operate today. Servers burn extra cycles tracking who’s talking to them, especially when millions of people hit the same service simultaneously. The next MCP version changes that dynamic. Instead of forcing one server to manage every conversation, the protocol makes requests easier to shuffle between different servers.

Sounds like a minor technical tweak, right? But it strips away a surprising amount of complexity for companies running AI at scale. That translates into faster responses, lower costs, and fewer bottlenecks — even if you never see the change.

What the MCP update actually does

The upcoming revision targets a core inefficiency in how AI systems authenticate and route requests. Current implementations often require a single server to hold context for an entire session. That works fine in a demo, but it crumbles under real-world load.

The new approach decentralizes that responsibility. Requests become more portable, so they can be handled by whichever server is best positioned at the moment. It’s a bit like upgrading from a single checkout line to a system that dynamically opens new lanes as queues grow.

For developers, this means less time wrestling with session management and more time building features. For users, it means AI tools that feel more responsive and reliable — even during peak usage spikes.

Why this matters beyond the tech bubble

This update won’t suddenly make your favorite chatbot feel like a genius overnight. What it does is lay the groundwork for AI to move beyond isolated chat windows and into your everyday workflow. That’s the direction the industry is heading: AI that drafts emails, updates your CRM, books meetings, and queries internal databases — all without you juggling a dozen logins.

That vision only works if the underlying connections are solid. MCP is the glue, and this update makes the glue stickier.

Sometimes boring is exactly what AI needs

I’ll admit, a protocol update isn’t headline-grabbing material. But it’s precisely these unglamorous fixes that separate a promising demo from a dependable product.

Consider how far we’ve come. Early AI integrations were brittle — one change in an API could break everything. MCP’s standardized approach changes that calculus. It creates a stable foundation that both startups and tech giants can build on.

That’s why this quiet update deserves attention. It’s not about teaching AI a new trick; it’s about fixing the pipes so everything else flows. And while that may not sound thrilling today, it’s exactly the kind of improvement that makes tomorrow’s AI feel effortless and genuinely useful.

For more on how AI is evolving beyond chatbots, check out our guide to AI-powered productivity tools or how to connect AI assistants to your daily apps.

Apple’s trade secrets lawsuit against OpenAI reads less like a legal filing and more like a spy thriller script. The 41-page complaint, filed Friday, doesn’t just accuse OpenAI of poaching talent. It paints a picture of a coordinated operation to siphon confidential Apple information — complete with smug text messages, stolen hardware parts, and an internal cheat sheet for dodging security.

Some of the allegations are so brazen they almost sound made up. One message allegedly sent by a former Apple engineer to a colleague still inside the company reads: “LOL, I found out I can access the [network storage], so funny.” The reply? “I’m ready.”

Here are the most striking claims in the Apple OpenAI lawsuit.

“Rotten to the core”: Apple takes a swipe at OpenAI’s hardware ambitions

Apple doesn’t mince words when describing the foundation of OpenAI’s rumored hardware business. You know, the one that might eventually challenge the iPhone.

“OpenAI’s nascent hardware business now rests on the shakiest of foundations, rotten to its core by its illegal reliance on misappropriated trade secrets,” the complaint states.

Leave it to Apple to work a rotting fruit metaphor into a legal document. The message is clear: whatever OpenAI is building, Apple claims it was built on stolen blueprints.

“This is the tip of the iceberg”

Apple isn’t just documenting what it knows. It’s signaling that this is only the beginning. The company argues that once the discovery process gets rolling — when emails, texts, and internal documents are handed over — the real scale of the alleged misconduct will come into focus.

“Discovery will expose that the misappropriation has been occurring on a scale many times greater than the several instances described below,” the complaint reads.

Translation: Apple thinks it’s caught a few fish, but it believes there’s a whole school down there.

The “LOL” message that started it all

One of the most damning pieces of evidence involves Chang Liu, a former senior systems electrical engineer at Apple who later joined OpenAI. According to the complaint, Liu exploited an authentication bug to access Apple’s systems — from the Apple-issued work computer of Yu-Ting “Alyssa” Peng, a colleague who allegedly acted as a conduit between the two companies.

Liu’s message to Peng — “LOL, I found out I can access the [network storage], so funny” — is included in the filing almost verbatim. Peng allegedly replied, “I’m ready.” She later left to join OpenAI herself but isn’t named as a defendant.

Then there’s the follow-up. Within hours of leaving Apple, Liu allegedly texted: “I still have another computer.” That message was discovered on Peng’s work laptop. Apple claims Liu planned to use that second machine to keep accessing confidential data after his departure.

“Didn’t even know we could take those from the office”

Some of the wildest allegations involve OpenAI’s hiring practices. Apple claims that OpenAI chief hardware officer Tang Yew Tan — who spent 24 years at Apple, most recently as VP of product design for iPhone and Apple Watch — directed job candidates still working at Apple to bring “actual parts” from Apple to their interviews.

The purpose? “Show and tell sessions.”

One candidate was reportedly surprised by the request, saying he didn’t even realize Apple parts could be taken out of the office. Apple also alleges candidates were told to bring “CAD/design artifacts” and “prototypes” to interviews.

That’s not a job interview. That’s a shopping list.

Avoiding the “dreaded walkout”

Apple’s complaint claims OpenAI went as far as coaching departing employees on how to evade Apple’s security procedures. The alleged method? An internal Apple document bearing a “Need to know” designation was circulated to new hires.

The document reportedly contained details on how to avoid the “dreaded walkout” — Apple’s practice of immediately removing employees from the premises when they give notice. By dodging that, employees could stay for the typical two-week notice period, giving them more time to access confidential information.

And if Apple asked departing employees to sign anything at their exit interview? OpenAI allegedly advised them not to sign — and to “let OpenAI know ‘asap.’”

Over 400 former Apple employees now work at OpenAI

The complaint also reveals a striking number: more than 400 former Apple employees now work at OpenAI. Apple uses that figure to underscore the scale of the problem.

“It is not surprising that certain OpenAI personnel have knowledge of Apple’s confidential and proprietary information, which they are obligated to keep confidential. But OpenAI has resorted to exploiting this confidential information,” the complaint states.

That’s a lot of people with a lot of institutional knowledge walking out the door.

The io connection: metal-finishing secrets and a $6.5 billion deal

Then there’s io, the hardware firm founded by former Apple employees — including Jony Ive — that OpenAI acquired last year in a deal valued at $6.5 billion. io is now a defendant in the lawsuit.

Apple alleges io used its proprietary industrial design techniques by misleading an Apple partner into believing it had permission to carry out a “confidential metal-finishing technique.” The complaint also claims OpenAI approached a supplier using confidential information about power and battery components — even using “internal terminology” to ask questions that “only Apple-insiders would know to ask.”

If you’re wondering whether this kind of behavior is normalized at OpenAI, Apple has an answer for that too. The company describes the alleged misconduct as “normalized and exemplified by leadership.”

“Apple is left with no choice”

Apple says it tried to resolve this quietly. The company claims it reached out to OpenAI back in February to raise concerns. OpenAI never responded.

So here we are.

OpenAI’s only public response so far came via a statement on X on Friday: “We have no interest in other companies’ trade secrets. We remain focused on building innovative technology that empowers people everywhere.”

Whether that holds up in court remains to be seen. But if even half of these allegations are proven, the Apple OpenAI lawsuit could become one of the most consequential trade secrets cases in tech history. For more on how companies protect their intellectual property, check out our breakdown of trade secret litigation trends and the growing tension between Big Tech hiring practices and non-compete agreements.