Taste, Curation and AI Using NotebookLM
One blog that I have enjoyed reading recently is Ted Gioia's blog, called The Honest Broker. Ted Gioia is an amazingly well read man, with rich and diverse experiences, and an encyclopedic knowledge of music - if you aren't subscribed already, you may want to consider doing so.
In a recent post, titled "I Answer 18 Questions" he, well, answers 18 questions. Among those answers was this line:
Heidegger warns us against viewing the world primarily as raw material—and he’s right. The tech agenda must always be balanced with more humanizing forces. I read William Blake as saying the same thing. Shakespeare, too, for that matter.
This answer is in response to a question about technology and monopolies, and Ted begins his answer by rephrasing it to be "Why is tech evil?". Now, you may or may not agree about tech being evil, but I found the answer to be fascinating. I am very far from being an expert on Heidegger, William Blake and Shakespeare, but the comment did make me curious, and particularly so about Shakespeare.
I've read some of the plays (but not all, alas), and certainly have not studied them enough to be able to meaningfully interpret Ted's answer. And so I set out to do what you'd expect me to do: I was going to ask an LLM to tell me more about this.
Meet Steven Johnson
Steven is an American author. I have not read any books written by him yet, and the reason I know of him is because for the past year or so, I have been a huge fan of a product called NotebookLM.
Here's Steven himself about the product, and how it all started:
About five months ago, I shared the news that I was collaborating with Google on a new AI-based tool for thought, then code-named Project Tailwind. If you happened to have missed that post, or are a new subscriber to Adjacent Possible, the quick backstory is that I have had a career-long obsession with using software to augment the research and ideation phase of my work as a writer, some of which I have explored in the “creative workflows” series in this newsletter. About a year and a half ago, the folks at Google Labs reached out to see if I was interested in helping to build a new tool for thought designed around a language model, and that collaboration led to the announcement of Project Tailwind at Google’s I/O conference in May.
A number of exciting developments have happened since then, so I figured it was time for an update. There are two main headlines. First we have dropped the code-name of Project Tailwind, and are now calling it NotebookLM. In the world before computers came along, the primary tool we had for organizing and remixing our ideas and the ideas we encountered in other people’s work was a notebook. This project is what you get when you try to reimagine note-taking software from the ground up knowing that you have a language model at the core. Hence NotebookLM.
OK, But What IS NotebookLM
Think of it as your research assistant.
Here we go again, you might think. But hang on, this is different. When I say research assistant, what I mean is that NotebookLM is an assistant for the research you have done.
As opposed to what? As opposed to an assistant who goes out and does the research for you.
A Deep Research query and NotebookLM are both AI tools, sure, but they're very different.
A Deep Research Query goes out into the Internet world and does research on your behalf. It comes back a few minutes later with a goopy report that you must plow through, but the research is usually quite good, with an ever narrowing confidence interval. When you factor in the time and cost advantages, a Deep Research query is a no-brainer.
But NotebookLM doesn't go out and trawl the internet for you. In fact, let's stick to that analogy. Think of Deep Research as a tool that trawls the ocean and comes back with a excreta-ton of links and references.
NotebookLM? He's the guy on the dock, taciturn look on face, toothpick sticking out of the side of the mouth, assessing the fish and reporting on the quality of the catch... to you.
And the analogy really works, because our guy simply refuses to do any work until he has any fish to examine:
There's literally nothing to do in the product until you load a source, and it can take time to assemble a truly rich notebook on a particular topic. Source curation (or context engineering, as we would now call it) is not something that the average user is familiar with.
So if, for example, you want to learn about Keynesian macroeconomics, there are two ways you can do it:
Ask Deep Research to write you an explainer by citing the best sources it can find (and how to write out this prompt would be a nice blogpost in and of itself)
Upload to NotebookLM the best sources you have been able to find. These are the fish we were talking about. Once he has fish to examine, our guy will tell you all about them.
There are three things I need to make clear at this point:
Agency (and this is important)
Curation and Taste
Connections
Agency: You Are An Active User
NotebookLM only gets to work once you upload sources into it. Remember, our guy will chill until he has fish to examine. It is your job to get him the fish!
Where do you get the sources from? Which ones have you chosen, and why? Anybody can get a half-decent Deep Research report out. But you are required to have some knowledge of the subject before you can complement it with the use of AI.
Let's say you are an undergrad student, and you have to show that you have studied and understood the theory of comparative advantage. Maybe you "write" a report and submit it to the professor. But what if five of you have to submit a curated NotebookLM, and the professor judges you on the kind of conversations you (all of you, the professor included) are able to have with the curated sources?
Which kind of assignment requires more work and agency on part of the student?
(Yes, of course this system can and will be gamed. Remember, better than the status quo is an improvement, and not to be sneezed at.)
Curation and Taste
Even if you were to try and game that hypothetical assignment, gaming it would require you to have a conversation with at least one LLM about the list of sources. Hunting down the sources that the LLM has suggested and adding them to the LLM is still likely to leave even the most disinterested student with some exposure to the topic.
On the other hand, a passionate, curious student may also have a conversation with an LLM about her curated list. But the difference here is the fact that the student is looking to make her already curated list better. There may be additions, there may be deletions, there may be an update to editions in the case of textbooks, etc. But this is a student with some expertise who is honing said expertise, as opposed to a student who is having the LLM substitute for her expertise.
And once that curated list is inside NotebookLM, the kind of questions one is able to ask (and one does ask) will very quickly show up in the differences between what one is able to elicit from the tool.
As a learner, NotebookLM forces you to have agency, in terms of curating the list of references for a given topic. It also forces you to have taste, in terms of the kind of questions you ask the list that you yourself have curated.
NotebookLM gives you helpful starting points. A briefing doc, a mind map, an FAQ, and a podcast even(!) are created by the tool once you upload the references. But how good these are, and the conversations they enable are both a function of the quality of the source documents.
Connections
Things get really interesting when you are able to draw a mind map across multiple documents. When you and the tool, working together, are able to get meaningful connections across different topics that are related (directly or otherwise) to each other... that is magic waiting to happen.
Wait, So What About Ted?
Right, so as I was saying:
In a recent post, titled "I Answer 18 Questions" he (Ted) answers 18 questions. Among those answers was this line:
Heidegger warns us against viewing the world primarily as raw material—and he’s right. The tech agenda must always be balanced with more humanizing forces. I read William Blake as saying the same thing. Shakespeare, too, for that matter.
Now, rather than ask o3 or Gemini 2.5 Pro, I have the ability to use a Featured Notebook.
One of the secrets to getting the most out of NotebookLM is assembling high-quality sources to help you explore your interests. Today, we’re rolling out a new feature making that easier than ever. We’re working with respected authors, researchers, publications and nonprofits around the world to create featured notebooks.
The notebooks cover everything from in-depth scientific explorations to practical travel guides to advice from experts.
Steven Johnson expands upon this in his own tweet on the subject:
So you can think of the Featured Notebooks we are launching today as a preview on two levels. For newcomers to NotebookLM, the notebooks are a preview of how useful the product can be when you've assembled a collection of sources for whatever project you're working on. But it's also a preview of a potential future where there are thousands of expert-curated notebooks on all sorts of topics that you can add to your own collection, to have the knowledge you need on tap. Our launch lineup is: Longevity advice from legendary scientist
@EricTopol, bestselling author of “Super Agers” Expert analysis and predictions for the year 2025 as shared in The World Ahead annual report by
@TheEconomist An advice notebook based on bestselling author Arthur C. Brooks' "How to Build A Life" columns in
@TheAtlantic A science fan’s guide to visiting Yellowstone National Park, complete with geological explanations and biodiversity insights An overview of long-term trends in human wellbeing published by the University of Oxford-affiliated project,
@OurWorldInData Science-backed parenting advice based on psychology professor Jacqueline Nesi’s popular Substack newsletter, Techno Sapiens
The Complete Works of William Shakespeare, for students and scholars to explore
A notebook tracking the Q1 earnings reports from the top 50 public companies worldwide, for financial analysts and market watchers alike
And it was the second last entry in this excerpt that caught my eye. Can I talk to an LLM that is "source-grounded" on only Shakespeare's works?
The chastening thing I learnt by reading the answer was that I do not know enough about either Shakespeare or this theme to be able to have a meaningful, in-depth conversation about the topic - but note that this is a feature, not a bug. That's the point of a great AI tool - it should help you understand where you need to get stronger, and how you might get there. I'm looking forward to chatting with the featured notebooks from The Economist, and the one that focuses on an overview of long-term trends in human wellbeing (courtesy our friends at Our World in Data).
Some Learning Possibilities
Imagine non-zero sum games being played in class, with a notebook for all the sources collated by all of the students. So imagine the entire class taking notes in a common GDoc, and using that GDoc as a way to curate sources for a class notebook.
Imagine universities sharing featured notebooks based around a variety of topics, both for their own students, and in some cases, for the world at large.
Imagine academic papers having a notebook rather than a bibliography.
Imagine there being notebooks for conference proceedings.
Syllabi for examinations, undergrad theses or Master's theses or even PhD theses with a notebook as an accompaniment.
WATTBA!