alt Hacker NewsNobody cracks open a programming book anymore
Comments
The decline of programming books removed a constraint on programming language complexity. At one point, the basic set of books for Java was six volumes. That's when language books broke down from sheer complexity. The combination of Google search and Stack Overflow allowed programming to become more complex than anyone could keep in their head. C++ bloated to the point that people who used to be C++ language lawyers couldn't keep up.
This follows a general trend. Areas which used to be bounded by the limits of the human mind stopped being bounded that way some time back. This first appeared in corporate structure. Through the 1970s or so, there was an upper limit on corporate complexity. Beyond some point, connectivity problems started to choke the organization. There were classic ways around this, mainly dividing companies into sub-companies with their own profit lines. "The Concept of the Corporation" by Peter Drucker describes how General Motors did that. GM was at the time a group of loosely connected car companies under one corporate roof.
A few companies figured out scaling early. Sears was famous for having developed the "Schedule System", which reduced fulfillment overhead from O(N * M) to O(N log M). This allowed Sears to run a giant ordering plant out of Chicago to serve the whole country. But many companies didn't scale well, and choked as they grew. Westinghouse is a classic example.
As computers came in, the scaling problems receded. Airlines got their reservation systems under control, and seat utilization went up. Logistics went from warehouses to fulfillment centers, with much shorter holding times in inventory. Chains no longer were limited in size - WalMart, McDonalds, and the big banks could expand to planetary scale. The giant corporate paper-pushing plants disappeared.
So did forced organizational simplicity. Companies had, at some level, to be simple. Otherwise they became unmanageable. As computerization proceeded, that constraint was relaxed.
Finance achieved previously unimaginable levels of complexity. Until the 1980s, most financial products were rather simple. Now, there's no limit, and the tail wags the dog. Futures markets are far bigger than the volume in the underlying commodity, and zero-sum activity dominates.
AI will accelerate this. There will be businesses no human can comprehend or manage. This may not be productive but will be profitable for someone.
I tried to introduce a partner to programming with an introductory Python book one year ago. It was brand new on the shelf in the impulse purchase area at Micro Center. It looked nice on the outside and decently vetted at a glance of the intro and a page or two, I trusted Micro Center (undeserved in retrospect), and I was in a bit of a rush. I gave it to my partner to try out on their own and they started having trouble pretty quick, and it wasn't really their fault - it was using a lot of technical terms and concepts with no explanation that you wouldn't expect someone to know who hasn't taken a few Computer Science classes.
And the best part.. it was Python 2.7. Micro Center sold me a brand new, glossy covered "Learn Python" book based on 2.7 in the year Anno Domini 2025. Its instructions didn't even properly tell you to install that version, so if you even make it that far you're going to be lost why the syntax is wrong for every example.
Moral is, books are just as easy to strike out on as a bad online resource. Honestly, I feel like Googling "x language tutorial" is probably going to get you the best results much more easily than picking something off the book shelf - if I can't vet a book reliably, and I already know the damn language inside and out, what hope does a newcomer have?
There is a good ending at least. Among a bunch of random stuff I got from an infrared spectroscopy shop that was closing down and practically giving away all their cool equipment, I found a copy of K&R C. I'd never read it myself but I'd heard so much about it online over the years that I figured it was as worth a try. So I got the victim of the Python book set up with WSL and gcc, and they had a much better time that time around.
Not true for everyone. I learned Rust from The Rust Programming Language ("The Rust Book") and "Rust for Rustaceans." Sure, coming from C/C++, I could have learned the syntax online but learning best idioms and styles required the time and commitment to read a book cover-to-cover. In fact, I've probably read each page in "Rust for Rustaceans" at least twice to ensure that I understood some of the more subtle points. I could have developed a half-baked notion of how the borrow-checker worked by fooling around and reading blurbs on Stack Exchange. But Rust for Rustaceans made clear the more subtle points that might have taken years of tinkering to understand. Thank goodness people still write excellent books on computer programming.
Beyond the slowing you to type, the key part of the good books was the considered and mindful order of presentation. This is what had me spending money when I could get the reference manual for free - a guide, a book that taught me unfamiliar concepts in top down fashion, and took some degree of responsibility to be both accessible and comprehensive.
I love the tutoring of LLM, but to this day as a complement to a guided book. I don't find such guided books in computer science much anymore sadly, but for now I still do it in other venues - French, Biology Astrophysics and such. I grab a book, and then use LLM to supplement my reading as my mind always has a myriad questions :).
Not entirely sure why computer science is so radically different - maybe because things change and get obsolete too fast? At any rate, cuddling with a book is still my favourite way to learn a new topic, much as I spend 12 hrs a day eagerly typing and staring at the screen as well :).
> Stack Overflow is receiving about 3,800 questions a month
The crazy thing is that SO is dying so quickly that it's already under half that amount.
https://data.stackexchange.com/stackoverflow/query/1926661#g...
> The kid who is right now learning to code by chatting with an agent is not a worse programmer than I was at 12, hunched over Learning Perl, retyping examples that would not run because I missed a semicolon.
To be honest, I'm 17 y.o., I'm coding by chatting with an agent, but it seems like we can't tell the distinction too absolutely.
At the first time writing a React app, I forgot to name a file with a .tsx extension and I used .ts instead, then spotting ugly error lines across my JSX syntax, confusing and sharing with my friend, and laughing this little funny thing all the day.
I once spent the whole afternoon choosing a js linter, reading their docs and perceiving different tastes. In my early twelve-ties (uh this sounds funny too) I'm always arrested by configuring Windows PEs, installing different Linux distributions on my PC, etc. Today I still read tech books, alongside videos, articles and also chatbots. Chatbot is a new tool, but there's no doubt it cannot replace other media types and what they bring to us/me.
What may I express is that a natural interest in programming or computer things cannot really be overwhelmed by LLM things. I don't know how to use vim skillfully since I majorly used Windows at my early age and I'm not familiar with vim's logic, but this practically doesn't stop anything. I still found Linux's fantasy, at last. And same for LLMs.
It's a shame because to guide a coding agent, you need to have the right grammar and vocabulary to describe what you want and how you want it to be built. Junior devs should read not because they need to know how to write the code, but they need to know the vocabulary and the grammar to guide the agents.
> You already know why, more or less. ChatGPT has over 900 million monthly active users. GitHub Copilot has 4.7 million paying subscribers as of January 2026, up roughly 75% in a year. You can’t imagine writing software without Claude Code anymore.
I just bought $600 worth of programming books and I’m pretty stoked to read them. Mostly a lot of titles considered “the classics” but my brain works best with hard print materials.
My two favorite non-fiction sections of the bookstore are dead and dying. The computer section, if it still exists, is just things like _Excel for Dummies_, and the philosophy books have all been pushed out by self help and dime-store "metaphysics".
But I've started reading programming books again recently, on my e-reader and on my laptop. People are still writing them, and they're still good. We should all go buy some!
For my own usage, I don't see chatbots as supplanting textbooks. If anything, they pair well; reading a book from cover to cover gives me the breadth and depth I want, but LLMs are there for tangents and questions that come up along the way. I was reading a book and chatting with Claude like tihs just yesterday, for a few hours.
I think it might have been a cognitive development thing, but at some point in high school, Stroustrup's "The C++ Programming Language" just kinda clicked for me, like I hadn't been reading it properly before.
I convinced my mom to buy me a book on C++ when I was 13 (25+ years ago). I made it to page ~75 or so before I got bored of reading, and needed to start building stuff to stay interested. I don't think I ever looked at it or any other language book again after that.
Reading a well written programming book will put you ahead of 99% of other programmers using that language. Most programmers learn some subset of a language to get things working, and never learn more than that.
With a programming book, you invested time to thoroughly learn something new. The language/tool and its associated way of thinking were now part of your toolbox. You were simply a better programmer going forward.
With AI that toolbox doesn’t grow. You don’t become better, everything from now on will be wrangled into shape with the frozen set of skills you have at this point.
For some people this is fine. For me the endless learning and sharpening of the blade is one of this most appealing parts of programming, so I hate this.
Curiously, I do buy and read tech books. My hobby is legacy OS kernel research so I bought some second handed books on old Linux (kernel 1.2) and NT (3.1). It is fun to research so I don’t use AI often for side projects.
This predates LLMs. The internet has been the primary source of programming knowledge for decades.
Books are still good for the fundamentals of course.
imo books for programming language should be roughly a guide to docs, with better context collocation and more elaborate examples, otherwise it would be really painful to use language
you can't pick up c++ from the docs and the language itself is a monstrosity, and for that you must have book explaining why do you have 30 types of pointers, golang in the meantime have excellent official guide, and you don't really need any book
I've not read a programming book for years, even before LLMs came on the scene. Didn't see the need to when there's so much information online.
These days, I don't use LLMs for actual programming but will ask them questions in lieu of doing a web search. It's like documentation I can chat to. Basically a more efficient blog post or book chapter that happens to be dedicated to whatever it is I'm working on.
I started learning software in the early 2010s and I read a lot of software books like the ones mentioned in the article. I continued reading them as the years went on, but the last one I bought was probably 4 or 5 years ago. Naturally, I probably don't need books as much as I used to -- I can generally pick up something new and know where to find what I need to find, "learning to learn" and all that. I also think they are better for foundational knowledge; many times the books become outdated very quickly. So if I was gonna attempt to write a database or learn distributed programming theory, I'd probably pick up a book, but if I wanted to learn a specific tool (or most languages) I'd probably stick to the web.
I had a taste of OReilly books through ACM/IEEE. A few years pass and I'm working on some obscure legacy code and a book saved me. Needed it in a hurry so signed up for OReilly books for a year even though only needed it for four weeks. I have renewed it since and enjoy it, being able to dip into topics, just for fun, written by people telling a story. The long form of a book is quite different from SO and new books keep coming out. I'm hooked.
The problem is knowledge gaps. You don't know what you don't know.
A good book deals with that.
For example, my Effective Pandas book teaches best practices that most never learn because the blog posts that train the models are actually espousing anti patterns.
(As an author I'm keenly aware that fewer folks are buying books.)
This post feels misleading or possibly just nostalgic. The books referenced still exist because the people creating the technology are still writing them. They're also creating video and attending conferences (virtual or otherwise). That's not going away anytime soon. But perhaps what has changed is how the information is accessed.
Do you need to debug some ancient perl? Sure, ask Claude. You'll get an answer and move on. But if you're looking to learn how to use the next technology before it's mainstream, you'll go looking for that material. And it's there, where you expect it to be. Do you still watch network television or haunt Blockbuster? Times change and the market moves on. The interesting thing is, people like books and they're also available for those looking for a physical artifact to hold. Most of what's available is POD. Depending on the title, you're hitting the print button when you place the order.
Nobody code anymore!!
Before the rise of AI, developers were basically doing copy/paste from StackOverflow. There are few developers who knows how to code.
Even DevOps engineer, I worked with CI/CD "specialist" who couldn't work for sht, if you asked him anything outside StackOverflow, he couldn't answer.
But there is a silver line for everything.....
I am not a developer but I learned to code with Perplexity AI, but not copy/paste, anything I didn't understand I asked it to explain why.
I wrote my first python app with classes, functions, 94% code quality coverade, the mock unittest was 4x bigger than the actual code. I can start a python script from scratch without looking at my own examples.
I would never be able to do that within a few weeks by looking at forums that often have worse response than AI hallucination.
I type this comment flanked by my shelves of computer books. Most I keep for sentimental or historical value. My first edition K&R, the first PostScript Red Book, the volumes of The Art of Computer Programming, and many animal books. In addition to what the author said, what saddens me is that most developers today will not have these milestones of their growth as a software engineer to look back upon. Our history is literally evaporating into the cloud.
It doesn't matter so much if the LLMs are better or worse for learning things. I tend to think much worse in the long term. The problem is the reduction in choice. Soon we won't have web search. No user generated content. No genuine personal interaction. No blogs. No personal computer industry. No tech book publishers. Its all going to be LLM generated content owned by a small group of people. It sucks. I am in the opt out group wondering what exactly will be left to opt out into.
Hot take: I'm reading programming books more now. There is so much to know about any technological topic and an LLM can tell you all of it, but it's overwhelming. What a book does is disciple and structure what you need to know, and what order to learn it in. Start with a book, grow your knowledge and put it into practice with an LLM.
I recently purchased Computer Systems: A Programmer's Perspective [0] and am currently working through it with pen and paper.
I've only had peripheral exposure to writing in assembly and "systems level" programming so I'm really quite enjoying it.
[0] https://www.amazon.com/Computer-Systems-Programmers-Perspect...
It used to be that you could buy a book and use it as a reference for years. That stopped being true sometime in the 1990s, as the half life of book value declined rapidly.
One persistent internet and Altavista became available it was just a matter of time, and now we're there. The whole move fast and break things culture won.
Like Chesterton's fence, you don't know what you're got until it's gone.
I still have my copy of Learning Perl. Mostly because it represents a milestone in my learning. I have kept and obtained a number of other books simply because they are antiquated, special and/or classics that are interesting to read even if they are not that useful to me, like Codd's relational book, or Calendrical Calculations. I hope the AI is trained on these sorts of books, so that the knowledge can live on in a different way.
The “higher level of abstraction” phrase used to describe what LLMs are in relation to programming… That phrase needs to die.
I printed a ton of books from libgen in the past 10 years.
Using paper just works better for me.
I do use LLMs for asking questions, and other learning tools.
I used to read a book or two when diving into a new language. But I think the last time I did that was in 2017 when I learned Swift. That was supplemented with a lot of Stackoverflow.
I think the next deep dive was in 2022, when I learned Go. But that was completely from online sources.
I use AI to summarize a book because I need to use it asap for my professional works. But I do reading because I enjoy learning process, the knowledge is a bonus, not the goal
> They were thick, they cost about $50, and they had titles like “Learning React” and “HTTP: The Definitive Guide”.
The most effective way to make money from open soruce was (for a time at least) to be Tim O'Reilly, Amazon, or Google.
I still even now feel that K&R C should be a mandatory reading for CS students, but alas.
The most transformative book for me has been SICP and Uncle Bob's work.
> Nobody cracks open a programming book anymore Not true for me. I still read the "Learning Rust in a month of lunches" although I ask AI to write Rust code all the times.
"there are coffee stains where the caffeine blots are somehow still a valid Perl program"
Having coded Perl for years, I take that personally, pal...
I do feel that programming is more accessible than ever though.
I've bought (and cracked open) more programming books in the past year than I had in the previous 10. I'm nobody?
Don’t worry, once LLMs poison the well enough by disincentivizing sharing content online, technical books will thrive again.
Remember man pages to learn an write C. Guided AI is good if it learns from a book not crap code found on GitHub.
People still use programming books, they just don't purchase new retail books as often. Books have become magnitudes cheaper for publishers to produce but the books didn't get cheaper for consumers. I went to O'Reilly's site, clicked on the first book that popped up and it costs $67.99 for a digitally printed paperback book:
https://www.ebooks.com/en-us/book/345913182/ai-engineering/c...
Take a guess why so many younger devs will opt to pirate a PDF rather than purchase retail programming books. Publishers are pricing themselves out of the market.
Llm learning fosters mediocrity
I bought a book on C++20 last year when we started a project on it and I read it.
I don't think programming books are going anywhere, because they still contain a concise directory full of information on different languages or frameworks. If I try to learn a language piecemeal through chatgpt or blogs I risk missing important details or platform-specific knowledge. I'd believe books on vim are going away but books on languages or other job-essential tools have a use in the market and I can't imagine they'll go away.
I love these books. I would like to own every single one of them, as crystallizations of a moment. But let's be real -- some of these books are trash.
I still maintain an O'Reilly.com subscription, because it's good to read aan edited book on a topic, and the Google search has just gone to seed.
This corporate messaging of "just use AI, cut as many corners as possible, only retain the essential people and force them to sling slop 7 days a week" is unsustainable.
It's wrong for so many reasons. It disrupts talent pipelines. The staff+ people probably don't want to work twice as hard to cover the cut headcount. In general, people prefer to work on systems that are well architected and not some slop that got vibe coded up in a weekend.
They (corporate upper management) could've just done nothing and the end result would've been better than whatever the fuck is happening right now
Disregarding the issue of AI for a moment, I don’t really think books were ever the ideal way to learn programming.
It’s so obviously better to learn programming in a web based medium. Not just for tutorials or code-running environments, but also for having up-to-date manuals and references for tooling as new releases come out.
Or, if you don’t like that, e-books are again vastly superior with the ability to search easily without flipping through indexes, copy/select text, etc.
Books become out of date so fast, and you live in a hell of manual transcription, which is not actually that helpful for learning despite being highly manual. I also remember dealing with typos and mistakes that were hard to fix as a new learner. Let’s hope someone sent a letter to the author and that the book sold well enough to get a second edition, which I’d then have to buy…but by then it was too late, I’d have moved on.
There was a huge bookshelf because there was no better option. Just like Blockbuster video, something far better came around.
I'm the author of O'Reilly's "Learning Go". Here are the last 13 months of paperback book sales:
- Mar 2026: 124
- Feb 2026: 140
- Jan 2026: 157
- Dec 2025: 306
- Nov 2025: 484
- Oct 2025: 218
- Sep 2025: 176
- Aug 2025: 136
- Jul 2025: 317
- Jun 2025: 230
- May 2025: 237
- Apr 2025: 165
- Mar 2025: 367
Sales are certainly down, but it has gone up and down in the past.
Since the 1st edition came out in 2021, it has sold roughly 20,000 copies (about 10,500 English paperback copies, 3,800 ebooks, and 6,700 translated copies). The 2nd edition came out in 2024 and has sold roughly 13,000 copies (about 8,300 English paperback copies, about 3,000 ebooks, and about 1,600 translated copies).
Most of the money comes from O'Reilly's online platform, not from book sales. That has been declining lately, partially because the latest edition is now over 2 years old, but also I suspect that people are cancelling O'Reilly subscriptions and just relying on LLMs (which have indexed all of the books and used pirated copies to do so).