What’s Going On With NotebookLM?

NotebookLM

NurPhoto through Getty Photos

It’s unusual to assume that Google NotebookLM is just three years outdated.

However it did initially come out in Could of 2023, and extra formally, close to the tip of that 12 months. Since then, it’s been wowing the crowds, astounding all bystanders, together with my associates, after I generated one of many platform’s signature “stay podcasts” round random supplies from our shared previous.

NotebookLM’s rise has been meteoric. I don’t assume it’s hyperbole to say that. So what are folks doing with this expertise three years out?

A Assortment of Use Circumstances

Once I researched the highest reported person functions for Pocket book, I got here up with a “prime 5.” I’ll simply briefly go over every one:

Analysis and summarizing – getting clear summaries, key factors, and explanations for no matter you’re engaged on (I’ll come again to this later.)

Writing and content material creation – turning sources into weblog posts, experiences, or outlines grounded in your personal materials.

Studying and learning – asking questions on your notes, creating research guides, and understanding powerful subjects sooner, with Pocket book as your indispensable “research buddy” (along with any human ones who could also be round.)

Assembly and doc evaluation – letting Pocket book check out assembly docs to extract motion gadgets, insights, and selections.

Thought era and synthesis – this one type of speaks for itself. Like vibecoding, it entails “letting the machine be artistic,” which has garnered its share of controversy.

The Abstract Product

Going again to that first level, as I’d promised, I got here throughout this text on Medium, the place Clare Spencer talks about placing one’s personal articles into NotebookLM to primarily use AI to critique using AI, provided that the primary topic is Spencer’s reporting on GenAI instruments.

The upshot, it appears, is that Pocket book took the subject, GenAI in journalism, and produced a set of infographics that, whereas slightly focused to the topic at hand, contained extreme typos that compromised the standard of the outcomes. You’ll be able to read all about it here, and let me know your ideas within the remark part.

The eventual abstract, fed by Spencer’s observations, included this:

“Profitable implementations rely closely on sturdy human oversight and structured coaching, designed to show journalists the technical limitations of LLMs and to prioritize energetic fact-checking of all AI output. Finally, the business is grappling with the right way to strategically apply this expertise whereas implementing stringent editorial requirements to safeguard accuracy and public belief.”

In different phrases, the platform isn’t fairly prepared for prime time.

Take Your Time, Pocket book!

Right here’s an fascinating testimony from a Redditor who discovered how immediate engineering will help with the standard of what Pocket book produces.

Able_Orchid_3818 writes this:

“Hey everybody, I’ve been experimenting closely with NotebookLM and located a workflow that drastically improves the standard of the outputs. In case you simply dump your recordsdata and ask for a abstract, you’re dropping a large quantity of useful info. Right here is my step-by-step methodology to get deep, complete, and extremely structured information out of NotebookLM.”

What’s on this multi-step technique? First, there’s the “index” methodology, described thusly:

Once you add your sources, don’t begin asking questions straight away. As an alternative, give NotebookLM a complete immediate asking it to index your sources into most important subjects, outputting solely the subject titles.

The remainder of it goes like this: you feed the index again into Pocket book, and ask it to clarify, not summarize. Then there’s the “One-by-One” Deep Dive, an elective addition, and the final half, which Able_Orchid identifies utilizing the phrase “endurance.”

“Go into the Customized settings and add a immediate like this: ‘Take your time researching. Dive deep, don’t rush, and be affected person in your evaluation and studying.,’” the poster writes. “It would sound bizarre to inform an AI to ‘take its time,’ however giving it this instruction grants the mannequin the conceptual leeway to generate for much longer, extremely detailed, and meticulously analyzed responses. Do that workflow subsequent time you’ve got a messy batch of notes or audio recordsdata.”

It’s an fascinating take: the concept by telling the mannequin to decelerate, you command a deeper, finer grained end result. It appears unusual to those that don’t perceive the ability and complexity of those methods. To me, it makes good sense.

That’s a bit about how individuals are utilizing NotebookLM in 2026, a 12 months that’s unrolling quickly. Keep tuned for extra.