New AI Update Let's Me Chat With My Books (And I've Switched from Google to Perplexity Thanks To Henri Matisse)
What a time to be a curious bookworm 🐛📚
Last week, I received an email update from Readwise, an app that lets you organize your highlights from the books you read, about a new feature that essentially lets me chat with my books.
I’m sure some AI enthusiasts out there have figured this out already without this feature and I’m late to the party, but this is a fascinating update I figured was worth sharing.
Here’s the exact update from Readwise:
You can now chat with your entire library of highlights directly in Readwise!
To put it simply, chatting is like “super-powered search” across everything you've ever highlighted, whether that’s Reader, Kindle, or any other connected reading app.
Want to ask a detailed question about a book you finished three months ago? Looking for a specific tweet you saved but can’t remember the exact words? Hoping to see connections across different domains on a particular idea you’ve been capturing over the years?
Chat With Highlights makes all of this both possible and easy.
This is exciting for a few reasons:
It’s useful for anyone who needs to recall specific passages from their highlights quickly.
Beyond getting a second opinion from your favorite AI model, you can now get feedback from your books.
It’ll be much easier to find interesting connection points between different authors’ ideas, which can serve as a launchpad for you to create something unique. In other words, your highlights become a serendipity generator.
It’s the second and third points that have me the most excited.
Here’s an example from my library of the feature in action:
From my simple question, Readwise pulled highlights from recent books I’ve read over the past twelve months, specifically Clear Thinking by Shane Parrish, The Great Mental Models Volume 1: General Thinking Concepts by Shane Parrish, and Originals by Adam Grant, to give me a response on how I can think more contrarian.
Connection points between these three different books are made clear within seconds. Now it’s up to you, the reader, to turn that response into real-world action.
A Hypothetical AI Exercise
If I wanted to jump down the rabbit hole created by my question, I could first craft a journaling prompt using the Clear Thinking reference (break free from social defaults) as a starting point:
“What social defaults am I following that breaking free from would make me uncomfortable?”
After a journaling session, I could then put my responses through rigorous questioning, inspired by the reference from The Great Mental Models Volume 1: General Thinking Concepts (use Socratic questioning to clarify your thoughts). I can start a conversation with my current preferred AI model, ChatGPT. For example, I habitually say “yes” to meeting invitations when I should be saying, “No, let’s discuss this asynchronously.” I could discuss this further with ChatGPT and let it rigorously question me to help me better understand this habit.
Once the conversation is done, I should have an opinion . . . a perspective . . . and I can state it back to ChatGPT. In a new conversation, I can ask it to challenge my perspective, inspired by the reference from Originals by Adam Grant (embrace criticism and invite dissenting opinions). This dialogue—turned into a debate by my prompt—will help reinforce my point of view since ChatGPT will look for any potential holes in my thinking.
Now, I’ve created this hypothetical exercise off the cuff, so it may need further refining, but you get the general idea—this new feature is a great way to stimulate serendipitous moments with your books and AI models.
Your questions can now return a mix of insights in the form of highlights pulled from autobiographies and history books and tweets from brilliant thinkers. Or you can simply use this feature to help you recall information you need the moment you need it.
I don’t want to over-romanticize this feature, but there’s a lot to be excited about here.
On Switching from Google to Perplexity Thanks To Henri Matisse
I’ve been using Google practically since I started using a computer. That’s years of habitual reinforcement, meaning that whenever I’m curious about something, I instinctively turn to Google to look for the information I need.
Recently, however, I’ve been asking myself if this is an “old school” way of doing things. There has to be a better, quicker way with AI, right? I mean, ChatGPT released its new search functionality last year.
But even though I heard Sam Altman tout its capabilities in an interview with The Free Press, I’m still not a regular user of the search feature. I’m not entirely sure why. As silly as it sounds, it may be due to the slight friction of having to toggle on the search button instead of using Google, a dedicated search engine. But even that feels like a weak excuse. I think it’s simply been hard to break the habit—Google is still synonymous with search in my mind, though that’s beginning to fade, as you’ll see in the rest of this article.
Now I have to admit that Google has gotten better with its recent addition of AI-generated summaries, aka “AI Overviews,” which are those little summaries at the top of a search result, as shown below:

Still, with all of the new tools I’ve been playing with over the past few years and all of the new tools I’ve had to learn in my career, specifically with my recent move into the DevOps space, I think if there’s any time for me to break habits, it’s now.
I’m ready to switch my way of doing things, starting with how I hunt for answers on the internet.
Researching Perplexity
Perplexity is a tool I’ve been hearing about for months on the podcasts I listen to, and it’s come up frequently in my conversations.
It recently came up in a conversation with a friend who said one of his new customers found his store through Perplexity, which was a first for him (and surprising) given the thousands he spends on Google Ads each month. Not to digress, but if these AI assistants and models become the dominant way people get their answers, shifting consumers away from traditional ways of searching for products and local businesses (e.g., using search engines and clicking on blue links), that’s going to be a major disruptor for companies and the ad agencies in charge of their marketing budgets, not to mention the way those agencies currently handle the dark magic that is SEO. At present, there’s no way to advertise on any of these AI assistants, so does that mean those businesses should be preparing to move to Meta or other social platforms to advertise? Are we going back to billboards? All of this is a topic for another article.
Back to my switch.
I’ve always thought of Perplexity as a tool mainly used by people who need deeper levels of research. Journalists, students, scientists—those were the kinds of users that came to mind for me. And as much as I’m willing to try new things, the amount of tools and AI models and assistants right now is both exciting and disorienting. Should I be using ChatGPT, Claude, Gemini, Copilot, DeepSeek, or Qwen as my daily driver for work? Should I be using this app or that? Is the premium version worth it?
Like all of us, I’m a bit conflicted when signing up for anything new—part of me is excited at the opportunity of optimizing my current processes, and another part of me is thinking, “Yet another subscription that’s ‘less than a cup of coffee a day.’ How does this thing work again? Is this really the best one? You know what, let’s hold off for now.”
Despite my inner conflict, I decided to continue with my research, and I came across an article from a professor talking about his experience switching from Google to Perplexity.
Moderately convinced and realizing how easily reversible this decision was, I downloaded Perplexity and tried it out.
Making The Switch
My initial impressions were that Perplexity’s interface is as simple as it gets (like many of the current AI assistants out right now).
I started asking questions and getting up-to-date answers from the web. Small numbers sit beside Perplexity’s responses indicating where the information was pulled from so that I can verify that the AI didn’t just hallucinate.
Also worth noting, with the Pro version, you can choose your preferred AI model to assist with your searches. As of writing this article, you can use DeepSeek’s R1, Anthropic’s Claude 3.5 Sonnet, Perplexity’s Sonar, OpenAI’s GPT-4o, and xAI’s Grok-2. At the current rate of innovation, I’m fully aware this list will be out of date in a few weeks, so refer to Perplexity's website for more information.
If you care to know my preference, I’ve been testing DeepSeek’s R1, and I’m enjoying watching its chain-of-thought reasoning when answering my queries. As Ethan Mollick said in a recent tweet, it’s hard to not anthropomorphize R1 when you can watch how it’s thinking through answering your question in real-time, which feels very human.
I love how R1 starts with, “Okay, let’s tackle the query about chain of thought reasoning,” almost as if it’s grounding itself before beginning its work. Again, very human.
Back to my switch, here’s an example of a typical Perplexity search:

Clean, simple, and elaborate.
I committed to using the tool every day for a week and did my best to resist returning to Google—unless absolutely necessary. You know, break in case of emergency.
During my first week, it was business as usual: jumping on meetings, having discussions, hearing unfamiliar terms, searching with Perplexity—all the standard way I’d use Google.
One day, on my lunch break, I was going through my personal inbox and reading random newsletters when I stumbled upon a promotional email from The Skateroom.
They announced they just dropped a Henri Matisse collection.
As a former skateboarder in my youth and now a regular museum-goer, I like looking at their new designs whenever they release them. This email, however, sparked my curiosity to learn more about Henri Matisse. I knew of him as an artist and I’d seen his work before, but I had never really known the story behind the cutouts. (Perhaps due to not paying attention in that one art history class I had in college.)
Nevertheless, I went to Perplexity, starting my search with a simple question, “Who is Henri Matisse?”
Immediately, I found myself having a conversation. Perplexity, like Google, also gives you related searches, which sparked ideas about new bits of information I wanted to know. The experience was novel. It was fun. I felt pulled to explore deeper, every response feeling like a puzzle piece being handed to me while the puzzle itself got bigger.
Eventually, the conversation led me to learn about how, after intestinal cancer surgery in 1941, Matisse was confined to a wheelchair, so he started “painting with scissors” and developed his famous cutout technique out of necessity since he wasn’t able to paint. His initial pieces of art were met with skepticism, but eventually, those very same critics saw the new style as innovative.
The quote below from this link provided by Perplexity summarizes the story behind his cutout technique fairly well:
Matisse's journey into the world of cut-outs was fraught with challenges. In his final years, beset by health issues that limited his ability to paint, Matisse turned to cutting painted paper as his primary mode of creation — a move that was initially met with skepticism by critics and the public alike. Many saw this new direction as a departure from "serious" art, questioning the legitimacy and depth of the cut-outs. However, Matisse persisted, driven by an unwavering belief in his vision. This perseverance was eventually rewarded, as the cut-outs came to be recognised for their innovative approach to colour, form, and composition. They were celebrated not only as the culmination of Matisse's artistic journey but as a major contribution to the narrative of modern art, influencing countless artists and movements that followed.
I also learned that Matisse called the last fourteen years of his life “Une seconde vie”—a second life. He described this phase as a liberating time, stating, “I have needed all that time to reach the stage where I can say what I want to say. Only what I created after the illness constitutes my real self: free, liberated…”
What a powerful reframing of his illness. The Stoics would be proud. I’ll be using this example whenever someone tells me they feel too old to start a creative project.
Now, some of you reading this part of the article may be thinking that I could’ve easily gotten this information off of Google, so Perplexity didn’t do anything too novel here that merits a switch or all of this writing. (Some of you may also be thinking this cutout story is as common knowledge as Van Gogh cutting his ear, but hey, we learn something new every day.)
You’d be right to a certain degree. I could’ve easily gotten the information from Google. I also could’ve gotten it from ChatGPT. However, I appreciated the conversational aspect of my search, and I enjoy the way Perplexity currently presents its information.
At present, though this will probably change, Perplexity’s responses feel “fuller” when it comes to search versus ChatGPT. They feel more elaborate, and I like the “related” feature that shows additional prompts to keep the conversation going.
Compare and contrast the three searches below:
Searching with Google, I’m given a tiny blurb on the bottom right about who Matisse is; however, the directions are pretty clear: click blue links to learn more.
Searching with ChatGPT, I’m provided with a decent overview, a video to watch, and an invitation to chat about what I’ve learned.
Searching with Perplexity, I’m given an elaborate summary, like ChatGPT can have a chat about my search, and I’m prompted with ideas to make the conversation more interesting.
It feels like walking up to a museum worker and asking, “Who is Henri Matisse?,” and the Google worker points me to a room with a plaque on the wall, while the ChatGPT and Perplexity workers immediately engage me in a stimulating dialogue. I’m sure the Gemini worker would offer me a pleasant experience as well, but I only have so many hours to explore and learn about Matisse.
This new search experience is night and day. Whether you use Perplexity or ChatGPT or some other AI model that has access to current information from the web, at least for me, all of this showcases a fundamental shift in how research is done online.
I don’t want to have to click a bunch of links; I want to first have a conversation about my search. I want the “too long; didn’t read” version. And I want to be given ideas for us to have a deeper discussion because I don’t know what I don’t know.
Though I’ve spent a large portion of this article talking about Perplexity, I think the relevant part of the article’s title can be changed to (And I've Switched from Google to Conversational Search Thanks To Matisse). The tool isn’t the main point—my eyes have opened to how much better searching can be when it’s a conversation and not a bunch of links shoved in my face.
I suspect that as the average searcher (e.g., your non-tech-savvy family member) gets comfortable shifting away from previous methods and moves toward new platforms, they will also have a similar realization of how much better search can be when it’s a dialogue.
Overall, I genuinely enjoyed jumping down the Matisse rabbit hole guided by AI, even though I still need to double-check if the museum workers just hallucinated. (I guess the workers in this museum all took acid?)
Before I wrap, I want to leave this quote from the professor I mentioned earlier that sold me on switching to Perplexity:
Over the next few days, as you use Perplexity.ai, you’ll notice something. You’re not just getting search results; you’re getting insights. You’re not just finding information; you’re learning. It’s a subtle but profound difference that transforms the way you interact with the internet.
So for the foreseeable future, until ChatGPT gets a bit better with search, I’ll stick with Perplexity. But hey, I’m flexible, and I can always be swayed toward something new.
Oh, and thanks, Matisse!
PS. Just as I finished putting the final touches on this article, OpenAI published an article introducing “deep research,” which is an agent that can conduct multi-step research for you. It appears doing your own research is now becoming an antiquated method. Who would’ve thought?