I haven't picked up a book in months, maybe longer, and honestly I don't think it has hurt my work.
When I need to learn something, I ask AI. When I'm stuck on a problem, I describe it and get an answer. The old model was: read the book, absorb the framework, apply it later. The new model is: start the work, hit the wall, ask the question, get the answer, keep moving.
This feels more efficient, but something about it bothers me, and I can't fully articulate why.
What is the case against books?
Here's the logic that's been running through my head:
People learn in order to achieve things, and every goal breaks down into tasks. Tasks can be solved by learning on-the-fly—ask a question, get an answer, apply it, move on. Why front-load hundreds of hours reading a book when you could just solve problems as they emerge?
AI is getting remarkably good at answering questions. And questions are the driver of learning. The information is there when you need it. Why store it in your head?
For procedural knowledge—how to do X—this seems obviously right. I don't need to read a book about Python syntax when I can ask for the exact function I need in the moment.
Where do good questions come from?
But here's what I keep circling back to: where do good questions come from?
When I ask AI a useful question, I already know enough to recognize what I don't know. I have some mental model of the problem space. I can identify the gap.
Where did that scaffolding come from? Probably from years of reading things I didn't immediately need.
This is where the research gets interesting. Cognitive scientists distinguish between different types of knowledge: procedural (how to do things), conceptual (mental models and frameworks), and tacit (judgment and intuition). Q&A with AI excels at procedural knowledge. The question is whether it can build the other two.
What does the research say?
Maryanne Wolf, a cognitive neuroscientist at UCLA, has spent decades studying what she calls "deep reading." Her finding: deep reading activates different neural pathways than information retrieval—specifically areas associated with critical analysis, empathy, and integration of background knowledge. The difference shows up in measurable brain activity, so it is more than a matter of taste.
"If you think about the brain as a series of muscles, these young readers do not exercise the muscles that, through labor and effort over time, contribute to the growth of the circuit. Writing is a perfect example. If you farm it all out to ChatGPT, children and youth are not only vulnerable, but they are losing the potential for their own brains to become ever more elaborated, and thus more able to evaluate and to articulate their own best thoughts. Our society's rapacious appetite for efficiency and effort-saving is ultimately anathema to the developing reading brain."
When we skim and search, we're optimizing for speed. When we read deeply, we're building something else—what Wolf calls the "contemplative dimension" of cognition.
There's also Keith Stanovich's work on the "Matthew Effect" in reading: children who read more develop larger vocabularies, which enables more reading, which enables more learning. The compounding happens because sustained reading builds foundational schemas that make future learning possible.
The uncomfortable implication: maybe the reason I can learn effectively through Q&A now is because I spent years building mental models through reading. I'm drawing down a balance I'm no longer depositing into.
Should kids still learn from books?
When I first wrote this piece I said I didn't know. After looking harder at the research, I now think the answer for children is yes, and the evidence is stronger than I expected.
The clearest long term evidence comes from the 1970 British Cohort Study. Alice Sullivan and Matt Brown compared children from similar backgrounds who had the same test scores at ages 5 and 10. The ones who read books often at 10 and more than once a week at 16 made more progress in vocabulary, spelling and maths by age 16. Maths is the interesting part, because reading a novel does not teach algebra directly. The researchers estimated that reading for pleasure mattered more for a child's progress than having a parent with a degree.
A study of 1,890 pairs of identical twins in the UK points in the same direction. Because identical twins share their genes and their home, a gap between them has to come from somewhere else. The twin who read better at age 7 tended to score higher on later intelligence tests, including nonverbal reasoning tests that have nothing to do with words. A later reanalysis argued that some of this effect may come from other stable factors, so I treat it as supporting evidence rather than proof.
The newest piece is a 2023 study of more than 10,000 American children in Psychological Medicine. Children who started reading for pleasure between ages 2 and 9 scored higher on cognitive tests at ages 9 to 13, had fewer behavioral problems, and showed differences in brain structure in regions tied to cognition. Like the cohort study, it shows a link rather than a controlled experiment.
None of these studies is perfect on its own, but together they tell a consistent story. Following a long argument across hundreds of pages trains a child to hold ideas in mind, connect them, and notice when something does not follow. That is close to what I would call the logic brain, and I have not seen evidence that a child gets the same training from a stream of answers.
Project based learning can also work well. Studies show that students in well structured project programs can outperform traditional classrooms, and second graders in one program gained five to six extra months of learning in social studies. Those programs still had a curriculum, scaffolding and a teacher guiding the inquiry, so they are not the same as letting a child figure everything out with AI.
For adults, is Q&A a faster way to learn than reading?
For adults I am less sure, and my own experience says that Q&A is simply faster. I can go from a question to a usable answer in minutes, while a book makes me read two hundred pages to find the twenty I needed.
There is good evidence that questions are a powerful way to learn. Psychologists call it the testing effect: people remember material much better when they practice answering questions about it than when they reread it. Q&A with an AI can work like that, as long as I am the one doing the thinking.
The risk is that the AI ends up doing the thinking instead. In a 2025 set of experiments with more than 10,000 people, participants who researched a topic through ChatGPT summaries put in less effort and came away with shallower knowledge than those who used ordinary web search, even when the facts were identical. A field experiment with nearly 1,000 high school students found something similar. Students with unrestricted GPT-4 access did much better on practice problems, but when the tool was taken away they scored 17% worse than students who never had it. A version of the tutor that gave hints instead of answers avoided most of that damage.
So my working answer is that Q&A is the faster way to take in knowledge, but speed is not the same as understanding. It works best when I ask follow up questions, push back, and try to explain the idea in my own words. It works poorly when I accept the first summary and move on.
So are books still worth it?
For children, I now think books are clearly worth it. The evidence links regular reading to better vocabulary, better maths and better reasoning, and the benefit seems to come from the long, effortful practice of following an argument. I would not swap that out for an AI tutor, though I would happily add one on top.
For adults who already have strong mental models, I think Q&A is often the better tool for getting knowledge quickly. Books still earn their place when I am entering a completely new field, because that is when I don't yet know which questions to ask.
What I'm most curious about is whether anyone has built genuine expertise in a complex field purely through Q&A, without the reading underneath it. I suspect the answer will shape how the next generation learns, and I think it is a question worth asking.