Reading in the Age of AI
Key Points
- The rise of AI has sparked worries that knowledge creation is stagnating, but the real issue is that we lack clear methods for reading and learning in an information‑overloaded era.
- Reading—whether physical books, Kindle articles, or audio content—remains essential, yet our traditional habits were built for a selective information age and must be adapted for today’s flood of data.
- The speaker proposes three reading frameworks for the AI age: “awareness reading” for quick, surface‑level updates (e.g., news feeds, skim‑reading to pass a test).
- The second framework, “information retrieval” (also called domain completion), leverages tools like Google or AI assistants to locate and extract specific knowledge on demand, turning the reader into a more efficient, targeted learner.
Sections
- Reading in the Age of AI - The speaker examines concerns that AI‑generated summaries diminish traditional reading, argues that our reading practices are ill‑defined for today’s information flood, and urges a rethinking of how we consume and retain knowledge.
- The Cost of Deep Reading - The speaker explains how deep reading consumes significant brain energy, argues for selective use amid AI-driven retrieval tools, and calls for better filtering mechanisms.
- Different Reading Strategies for AI Books - The speaker explains three distinct approaches—retrieval, conneto (deep narrative), and awareness reading—using examples of AI‑focused titles to illustrate when each method is appropriate.
- AI Overload and Quality Filtering - The speaker likens the flood of AI‑generated content to a resume glut, arguing that without strong filtering mechanisms we risk drowning in low‑quality information and must curate trusted resources for both beginners and advanced users.
- Beyond Token Metaphors: Human Insight - The speaker contends that true innovation stems from deep, iterative reading and reflection—far richer than the token‑based AI model—and believes AI can serve as a catalyst to amplify this human‑driven creativity.
- AI-Enhanced Interdisciplinary Learning - The speaker urges passionate, multi‑domain learners to be discerning readers, use AI as a tool to link diverse fields—while avoiding inflated claims of AI‑generated breakthroughs—and follow curated learning pathways to turn their interests into meaningful contributions.
Full Transcript
# Reading in the Age of AI **Source:** [https://www.youtube.com/watch?v=1Q8CjX0SP0o](https://www.youtube.com/watch?v=1Q8CjX0SP0o) **Duration:** 00:19:03 ## Summary - The rise of AI has sparked worries that knowledge creation is stagnating, but the real issue is that we lack clear methods for reading and learning in an information‑overloaded era. - Reading—whether physical books, Kindle articles, or audio content—remains essential, yet our traditional habits were built for a selective information age and must be adapted for today’s flood of data. - The speaker proposes three reading frameworks for the AI age: “awareness reading” for quick, surface‑level updates (e.g., news feeds, skim‑reading to pass a test). - The second framework, “information retrieval” (also called domain completion), leverages tools like Google or AI assistants to locate and extract specific knowledge on demand, turning the reader into a more efficient, targeted learner. ## Sections - [00:00:00](https://www.youtube.com/watch?v=1Q8CjX0SP0o&t=0s) **Reading in the Age of AI** - The speaker examines concerns that AI‑generated summaries diminish traditional reading, argues that our reading practices are ill‑defined for today’s information flood, and urges a rethinking of how we consume and retain knowledge. - [00:03:28](https://www.youtube.com/watch?v=1Q8CjX0SP0o&t=208s) **The Cost of Deep Reading** - The speaker explains how deep reading consumes significant brain energy, argues for selective use amid AI-driven retrieval tools, and calls for better filtering mechanisms. - [00:07:10](https://www.youtube.com/watch?v=1Q8CjX0SP0o&t=430s) **Different Reading Strategies for AI Books** - The speaker explains three distinct approaches—retrieval, conneto (deep narrative), and awareness reading—using examples of AI‑focused titles to illustrate when each method is appropriate. - [00:10:18](https://www.youtube.com/watch?v=1Q8CjX0SP0o&t=618s) **AI Overload and Quality Filtering** - The speaker likens the flood of AI‑generated content to a resume glut, arguing that without strong filtering mechanisms we risk drowning in low‑quality information and must curate trusted resources for both beginners and advanced users. - [00:14:08](https://www.youtube.com/watch?v=1Q8CjX0SP0o&t=848s) **Beyond Token Metaphors: Human Insight** - The speaker contends that true innovation stems from deep, iterative reading and reflection—far richer than the token‑based AI model—and believes AI can serve as a catalyst to amplify this human‑driven creativity. - [00:17:23](https://www.youtube.com/watch?v=1Q8CjX0SP0o&t=1043s) **AI-Enhanced Interdisciplinary Learning** - The speaker urges passionate, multi‑domain learners to be discerning readers, use AI as a tool to link diverse fields—while avoiding inflated claims of AI‑generated breakthroughs—and follow curated learning pathways to turn their interests into meaningful contributions. ## Full Transcript
So, this one is by popular request. I've
gotten asked a ton. Nate, what do I
read? Look at the books off. What do I
read in the age of AI? How do I know I'm
reading the right things? It feels like
reading is becoming increasingly
irrelevant. The Wall Street Journal came
out this week with an assessment that
basically suggests we are rotting out
knowledge in the knowledge economy
because AI is repeating knowledge and
regurgitating knowledge and
recalibrating knowledge and there's no
actual new knowledge being produced
because people are just asking for
summaries. I don't think that's actually
true. I think that's panicinducing. But
I do think we need to talk about it and
I want to talk about it in the context
of how we read in the age of AI. And I
think this is a big big deal because
even if you don't read books, right?
Maybe you're on a Kindle, maybe you are
reading articles, you are still
consuming Hunter, maybe you're in this
video. I think if you are listening to
this video, it's like an audible book.
It counts as reading. Maybe you're
you're feeling like that's being
generous, but it's true. It counts as
reading. I want to suggest that the way
we read is not well enough documented.
And that's part of why we're suffering
is we don't know how to be readers in an
age when we're drowning in information
because reading evolved for a selective
information age. We learn to read
because we needed to remember things and
remembering by writing down works pretty
well. And to retrieve it, you have to
read it, right? That was the fundamental
idea. That's why a lot of the initial
records we have for writing are commerce
records. You are trying to read and
understand. Court documents are also
very early examples of writing. You're
trying to read, understand, remember. We
have since evolved and we read for all
kinds of reasons. In the age of AI, I
want to give you three ways that I think
we read and learn. And I want to give
you frameworks for how you think about
this when you have like let's assume you
have an intelligent AI counterpart, a a
buddy, a chat GPT in your pocket that
you're reading with. How do you use
that? Well, so you are not just wasting
your time and information goes in one
ear and it goes out the other ear. So
here are the three frames that we're
going to walk through. Number one, I
think it's the lightest is what I would
call awareness reading. So that's the
idea that you just need to catch up on
the information of the day. This might
be the person who is scrolling X. This
might be the person who is trying to
understand what's in the news. This
might be the person who is just skimming
through a book in the bookstore or maybe
skimming through a book that they're
going to get a test on and they just
need to have enough awareness to pass
the test. Right? That's that lightest
layer. The second layer is what I call
information retrieval. You also hear it
called domain completion. It actually
maps pretty well to a lot of what we do
with Google. I think it's underthought
of that we Google as readers. In a
sense, Google is a form of reading
because we're trying to go out and
retrieve information that has been
written down somewhere. That is the act
of reading. And when we do a lot of our
sort of intermediate level reading, what
we're really doing is we're saying fact
A is in a book. Fact A is on a web page.
I need to go and get it and retrieve it
for a particular purpose. I'm going to
use it. I'm going to repurpose it to
borrow token architecture. I'm going to
stick the token in and I'm going to
produce a new token. Right? That's
retrieval reading. And then three, and
this is the one that people get like
lots of big feelings about. It's what I
there's not been a good name for it. I'm
calling it conneto reading. Other people
call it deep reading. I like to call it
conneto reading because my thinking is
that, you know, our brains are plastic.
Our brains are forming all the time,
evolving all the time, pruning pathways,
etc. We are deepening the conneto in our
brains when we read deeply. And this is
the kind of reading that teachers get
very excited about. This is the kind of
reading that everyone worries is
disappearing, but it's only one part of
this larger reading landscape. And by
all accounts, it's a very expensive form
of reading from an energy perspective. I
was doing some research on sort of
reading and the science of reading,
apparently, if you're doing that kind of
deep reading, you are burning a
tremendous amount of glucose in your
brain because you're trying to process
so much. And so as much as we may try
and lament and sort of put the
responsibility for reading more on
ourselves or on our peers, maybe part of
it is recognizing that this is a very
energetically expensive task and we need
to make the choice to use that reading
skill where it matters the most. And
that is the part that I would argue has
gotten harder in the age of AI. And
that's what I want to address. I think
that we need better tools for filtering
and simplifying reading tasks so that we
can truly build understanding at the
levels we need. We can build awareness
where we need it. We can retrieve
information where we need it. Uh which
also, by the way, I would argue LLMs are
retrieval reading. You're getting
answers there. There's a popular term
now called answer engine optimization
because people needed an answer for
search engine optimization in the age of
AI, so they made it up. But the idea is
the right. You're doing retrieval and
you're coming back with information. And
then there's conneto reading, right? The
the three I want to suggest that there
is a way for us to understand in advance
what kinds of books we need to dig into
in a conneto sense, in a deep reading
sense, what kinds we can just skim, what
kinds we can do some retrieval on. And
it has never been more possible to do
that kind of learning across multiple
book types, multiple retrieval types
using the assistance of AI. Yes, I
unlike the Wall Street Journal, I do not
think that AI is bad for reading. I
think people who passively use AI are
bad readers. And that's a that's a
different thing. So what do I mean here?
Let's go to a few examples. Let's for
example look at AI literacy
fundamentals. The author is Ben Jones
and the pitch is pretty simple, right?
If if you're overwhelmed by AI, this
will help you, right? You can join in
the conversation, etc. It is clearly a
beginner grade book. I would say just
looking at it and this is again I want
to talk this through so you get this
into your head. This is an example of a
book where there are lots of facts that
you want to retrieve and you want to
understand. This is a good example of a
book where you should have a copy and
then you should read with AI as an
information retrieval system. So you
should say this is my current level of
fluency in a prompt. This is the book
that I'm holding and looking at. By the
way, I don't advocate not getting the
book. I think you should get the book
and I want to read only the sections
that I'm missing on. Maybe I want to
read about how token architecture works.
Maybe I want to read about how next
token prediction works. Maybe I just
want to understand how AI conversations
happen in a chatbot. That's the level
I'm at. You can then go in and talk to
the AI and say, "What parts of the book
do I need to read? What elements of this
book speak to that knowledge gap, the
thing I'm trying to learn? And maybe
what other books do you recommend?" And
by the way, I tested this on AI to see
how much hallucination I would get as
far as madeup books. It is remarkably
accurate. I was working with Perplexity
for this. I find that Perplexity is a
great search engine here. I only had one
instance in about 30 books where
Perplexity made up the book entirely.
And I mean, is that great, right? You
have to check it. I caught it. But the
other 29 books were legit and they were
good books and there was a great find.
Saved me a ton of time kind of sort of
pulling the list together and making
sure that I had a fluent understanding
because as much as I read, there's books
coming out on AI all the time and I
wanted that broadstroke view. So I think
AI literacy fundamentals is a good
example of the kind of retrieval motion
I'm talking about where you need to
acquire new facts. Let's talk about
something where I think you have a
different kind of reading. Let's talk
about co-intelligence living and working
with AI. It's a bestseller by Ethan
Mllik. People have heard about it. As
someone who has read that book, that
book is literally over here on my shelf.
I think that benefits more from conneto
reading. I think that's an example of a
book you should read end to end because
the narrative builds over time. You may
not have a paper book. Maybe it's on the
Kindle, but you should read it and you
should let it sink in. The good news is
this particular book is written very
clearly. It's easy to understand and
digest. As much as you're doing conneto
reading, it's not complex and hard to
read. Now, an example of awareness
reading. I actually think a fantastic
example of awareness reading is the
Washington is the I talked about at the
beginning of this video. The one that
talks about knowledge rot. I think you
just need to be aware of it. It's a
great example of something that frankly
perplexity could summarize for you and
just let you know what the state of the
conversation is. It is not something you
need to spend time digging into because
frankly the level of effort and
investment by the authors in a newspaper
article is something that varies a lot
and you have to use your judgment about
whether there's something new and
noteworthy there from an AI perspective
that is going to get you value for the
time you're putting in to read that
article end to end. And I can sort of
hear and feel the journalists that might
be listening to this sort of having
heart palpitations because like that
feels like it disintermediates the the
journalist and and the print experience
a little bit. I love good journalism. I
think there's a place for good
journalism and I've been pretty honest
about the fact that the number of good
articles that deeply understand
artificial intelligence, I can count
them on like one hand. And I think
that's a larger issue that I would love
to see addressed and I'd love to see a
larger journalistic conversation about
better quality newspaper articles, but I
got to be honest about where we are. I
think a lot of the articles that I see
are not well enough researched to
deserve deep reading at this time. And
so I think they go in the awareness
bucket. So when you think about it that
way, my thesis is that this dramatically
simplifies the whole problem. I think if
we think of knowledge rot as a problem
and we obsess over the the the gap the
the so-called rot in the space, we're
focusing on the wrong thing. I think we
should instead focus on reading as a
skill, recognize that reading is a skill
issue that needs to be updated for the
age of AI and that the primary way it
gets updated is by figuring out how to
work with this new co-intelligence to
borrow Ethan's phrase to read well to
read in a way that we remember to save
room for deep reading where it matters.
And I think if we think about it as wow
AI is enabling us to produce so many
more tokens than we had before.
Therefore our filtering problem is
really the issue that simplifies a lot
of this discourse because then it's
really not our fault. Like if you think
about it this is analogous to the job
market situation where as most people
who are in the job market know we are in
a world where recruiters are snowed
under with resumes that have been
published prepared by AI. the it's a
terrible experience for the applicants
because the applicants can't get any
attention because there's so many
perfect resumes now. It's a terrible
experience for the recruiters and for HR
because they can't figure out how to
sift for quality anymore and the whole
system has broken down. In that world,
AI producing tokens has led to a crisis
of sifting and businesses are reading
candidates quote unquote by bringing
them on site by doing anything they can
to get an actual candidate experience.
Some of them are trying AI interviews
etc. In the same way, we need to get
aggressive about sifting for quality
information. And so, part of what I'm
doing by sort of building this book list
is I want to have a source for quality
AI information. And some of it will be
retrieval and some of it will be
introductory and for beginners and some
of it's going to be for like advanced
users as well. Like one of it, one of
the advanced ones that I absolutely
love, it's from Stripe Press is the art
of doing science and engineering. It is
not directly about AI. It is a fantastic
book. It's by Richard W. Hamming and if
you are anywhere in the technical space,
I would recommend it. It helps you
understand how innovation happens in
technical spaces and I think it's a
highly relevant foundational text. So
there's there's going to be sort of
books like that as well. Please, please,
please do not listen to the people who
tell you that the problem is information
rot per se. Because if you believe that,
you are going to get into a position
where you think there's nothing you can
do about it. The problem is not
information rot per se. The problem is
that we need new skills in a world where
we have artificial intelligence right
next to us all the time. I think one of
the key skills is learning when to read
for retrieval, when to read for
awareness, when to read deeply for
conneto reading. The other key skill is
going to be learning when and how to put
out what we actually think into the
world. I'll give you an example. This is
a work of passion for me. I care about
reading. Are you surprised? Look at
these books. I care passionately about
reading and supporting readers,
supporting next generation readers. I
see the anecdotes coming out of the
education system where people are
struggling to get students engaged on
reading where people in high school and
above are saying that students are
phoning it in and just using AI to sort
of put the essays in and the educational
system that was predicated on the idea
that people needed to write essays that
people needed to demonstrate their
understanding by producing words is just
not working anymore. I think that the
way back the way back is reigniting
passion for deeply understanding a
subject. Reigniting curiosity,
understanding that there is too much
information to process and that wrote
assignments were never that effective
anyway. And that we need to challenge
people to grapple deeply with things to
get them to read. And if you want to
learn about AI, your best asset,
ironically, despite this being about how
you use AI to read better, your best
asset is still your curiosity. Your best
asset is still your passion for
artificial intelligence and the subject
matter. That matters. That matters a
lot. And I want to remind you of that
because that is empowering. If you are
curious, if you are passionate about
something, not only will you read
better, not only will you learn better,
you're going to find your voice. You're
going to find something you want to say
into the world. Maybe it's about AI.
Maybe it's about science and engineering
like Richard Hamming did. Maybe it's
about something else entirely. But that
voice, that desire to speak back, in my
experience, only comes when you are
reading enough to prime the pumps. When
you are thinking and interacting with
the world and taking in information and
meditating on it deeply and letting it
reprocess inside you. I don't think the
token architecture metaphor works very
well for people. I know we use it a lot
because that's the predominant tech hype
cycle of our age. What we do with
information is deeper than that. The way
we reprocess, the way we mix it, it's
not something that is easily
translatable in machine terms. And that
magical compost pile is where the best
authors have always gotten their stuff.
It is where people who come up with
ideas to build a business get their
stuff. Except for them, they're reading
customers. They are reading the market.
They are reading pain points and they're
thinking about it and obsessing over it
and and studying it. We need that kind
of passion in order to put something
back in, in order to have the skill to
fight the knowledge rot, to have the
skill to say something new. I don't
think that that's hopeless. In fact, I
have never been more bullish on it
because I think that properly used AI
can help us to say new things. I'm going
to give you an example here. I am as a
project as a as a brain stretching
project working on learning Swedish. I'm
not good at it. Please, no one in my
comments come for me and say, "Tell me
about your Swedish skills. They're
terrible. I'm learning." But I have a
background not just in English, but also
in Indonesian. There is nothing in the
literature on learning English or
learning Swedish if you know Indonesian
already. I could not find anything. And
so I worked with GPT5 Pro and I put
together a way of thinking about Swedish
grammar that makes sense to someone who
knows Indonesian and a way sort of to
walk into learning Swedish. As far as I
know, that hasn't been done before. I'm
sure that could be done by someone with
a linguistics degree who knows Swedish
and Indonesian. I don't pretend that
that's something that's like completely
impossible for people, but it hasn't
been done. And GP25 Pro enabled me, a
humble person who is not a linguist, to
just go and do it. That is an example of
passion working with an AI to put
something out that I can understand and
actually like gain a skill set on. It is
using AI to be active. And so I want to
challenge you. Don't be afraid to work
with AI to build new and active things.
It is worth trying. It is worth taking a
prompt and saying, "Hey, can you please
help me with this?" And I will say the
advanced reasoning models do a lot
better. In particular, if you are if you
were sort of inspired by this and you're
like, what what is the takeaway here? I
don't do Swedish, but is there something
I can learn? Think about it as GPT5 and
other advanced reasoning models are very
very good at stitching between domains.
And so if you have existing knowledge
domains, you will have cracks in those
knowledge domains. I described one
between Indonesian and Swedish
linguistics. No one had bridged that
before. It knows all the domains well
enough that it can establish a bridge or
at least the start of a bridge, a
scaffold of a bridge between those
domains that ends up being a new
connector in the human knowledge set.
That's really exciting. And so what you
should do is if you're passionate about
something and you cover multiple
domains, be a good reader, work with AI,
see if you can't connect those domains
together in a way that's interesting and
novel. This is not the same thing as the
people working with AI and claiming they
are coming up with novel physics when
they are not physicists. That is a very
different thing. I would say that is
trying to say that the reasoning models
can advance the edges of human domain
experience beyond what PhD physicists
can do. That is at best unproven at
best. And there are articles sort of
from physicists explaining why that's
challenging. You have similar things
with mathematics. I think physics just
gets more attention because people watch
Star Trek a lot. So, where does that
leave us? Please be a careful reader.
Please filter what you want to read.
Don't blame yourself if you feel
overwhelmed. Figure out what your
pathway is for reading and jump into it.
I'm going to sort of put something up on
the Substack with like learning pathways
for different goals in AI. But really,
it's not about AI. It is about AI, but
like use it to learn anything in the age
when you have a lot of information to
choose from. And then when you're
passionate about something, think about
ways that you can actually put your
voice back out there. And maybe it's AI
assisted like I talked about with GPT5
Pro and Indonesian and Swedish. And
maybe it's not. Maybe you have the voice
inside and you already know. An example
of that one, as far as I know, I came up
with a sort of retrieval awareness uh
deep reading thing. I don't know if
other people have done it. It certainly
didn't come from AI. It came from me
arguing with AI and saying, "You're
wrong. I think this is the way we should
categorize it." Out of my own head
because I had passion for it. Because I
care about reading. care about reading.
Learn. Put something out there.