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.)
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.
Getting a book is my goto to learn anything new. I taught myself PHP and ended up finding a book 4 years later when I was looking for an answer to something. On the next page was something that would have saved me tons of time so I read the whole book.
Since then I’ve read books on Ruby, Go, Elixir, Docker, K8s and a lot more. By far the best way to get a semi complete understanding of anything without scraping together data from the internet yourself, because you won’t easily know the gaps.
I have a gut feeling that human as a creature learns better when looking at the information from several different angles, both physically and mentally. Been physically I mean looking at the same concept on screen and on hard copy books, perhaps taking notes and mark relevent sentences with a highlighter. Similarly, seeing a concept on physical book and write some short code snippet is viewing the concept from different mental angles. Though I don't have a proof for that and have yet to find a formal research on this topic.
I learn the same way but there are different people who learn in different ways. Also some people come with some concepts already understood from past experience or education, it’s easier for to pick things up without needing various angles.
I came here to say almost the same thing. I've been learning rust in my free time because I don't do enough programming at work to scratch that itch any more, and I've been using the rust book as a reference.
Thanks for the rec for "rust for rustaceans" I'll have a look into it.
I've only been using chatgpt for points where i'd normally go ask another dev for another set of eyes to debug something, otherwise all my learning and doing has been mostly the rust book, crates, and blogs about rust, ecs, roguelikes etc etc. It's been so fun!
My method is to get some hands-on experience first with the technology: Tutorials, Getting Started pages, the first chapter of a book. The fumble for some time and getting exposed to various sources of information. Then I take a book and skim it from start to cover, stopping at interesting bits. Then I reread various parts.
Sometimes I start with reading the books. But I already known I won't retain anything deeply. But it will gives me all the right keywords to learn more about the technology.
It always amaze me when I see fellow programmers struggle with problems that could have been solved easily by just reading that introductory book on the subject.
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 :).
Apparently the guy who wrote the Camel book on Perl made less than $1000 from that book. I was shocked when I heard about that because back in the day when I was learning that book was incredibly popular and seemed to be everywhere.
Agreed, Books on specific programming language were indeed tricky.
I found books on architecture, systems, or patterns, were more available. E.g. On relational database optimization principles, or Unix system administration, or graphics algorithms and rendering math, etc :)
I've commented on this development before on HN, so I'm glad to see this post on the front page. From a few months back:
"...the fact of the matter is that kids getting into high tech and programming mostly don't read books anymore. How do I know? Recently I was hanging out with a bunch of high school students who asked me how I learned. I said it was mostly via books and man pages. "Yeah, don't sleep on high quality written material. O'Reilly. Wiley. Addison-Wesley. Manning. MIT. No Starch Press. &c...
"Well. You should have seen the look on their faces. I might as well have morphed into the Steve Buscemi meme "How do you do, fellow kids?" They looked at me like I was a total relic or greybeard and said things like "Nah, nobody reads tech books anymore; I learned Typescript from YouTube videos."
Any question asked would be edited beyond recognition (and usually into brash rudeness). Half the answers were demanding ever increasing proof of work, and the other half told the OP that they shouldn't even be trying to do what they're doing. The only useful thing were opinion based posts from people with domain expertise, and SO kept trying to ban and remove those. It was the least helpful place online, but the most accessible, and it survived for lack of alternatives.
I'm no AI booster, but answering simple questions about well understood topics is a perfect fit for it. Good riddance to StackOverflow.
Stack Overflow was a nice experience for me because I was able to hit 2k reputation fairly quickly, in just 30 days of posting and 6 weeks calendar time. That being said, it never had the community feel of places I spent during my formative years, which were more on forums and IRC.
Here's a conference talk I gave on how to gain Stack Overflow reputation from back in 2018, selected out of 5 submitted talks. It's amazing how fast times have changed from before, during, and now after.
I had a pretty good time asking a question about Prolog. It was a really interesting experience knowing that there's someone out there that high proficiency in a very niche language, patiently explaining to me an issue that they have probably heard a million times from yet another imperative programmer. They even have their own website advocating for Prolog, etc.
Now, I could imagine an LLM would be able to do the same. However, I understand that this is only possible because of people like them. I don't think the youngins that started with LLM directly would appreciate the humongous amount of data and discussions online that enables that. The internet is so much bigger than just Google, Facebook, Youtube and Twitter.
I did originally, when it collected a bunch of obscure knowledge and made it searchable and useful. It was fun and rewarding to put things you knew into the common knowledge pool, and everyone celebrated a successful competitor to Experts Exchange. The SO model had a few major flaws that became impossible to ignore after it was entrenched. First, the reward scheme rewards the exact opposite of what it should incentivise: common questions are hit by many users and therefore attract lots of upvotes while answering the really hard stuff often meant you didn't even get your answer marked as "accepted" (because the OP had given up and stopped checking the site). Second, the site deliberately cultivated an "editor caste" in the Wikipedia style before the failure modes of that model were well-known: well-intentioned newbies get shut down by miserable yet untouchable people who play (and sometimes help write) the site's rules. Third, the stated desire to identify canonical answers to questions had no clear way to handle the evolution of the software people were talking about. So you'd have highly upvoted answers that might have been referencing deprecated libraries, and it was very hard for the newer answer to gain traction via either internal or external search.
It was also unfortunately before the retro boom of the 2020s, so questions about older arcana were often vulnerable to being closed instead of answered.
Yes, I liked it. When it debuted it was a massive improvement over expertsexchange, who had previously dominated the Google searches with bait and switch previews.
It may not have aged well but to say it was always crap is rewriting history.
StackOverflow was great when I was a very junior dev working on JavaScript apps. Anytime I ran into a roadblock, there was often a relevant post there to help me. As I become more competent though, I realized that reading the documentation directly was usually a much better way to get answers to my questions, and I stopped visiting.
I have found StackOverflow useful on rare occasion, but the friction, idiotic moderation culture, and high noise-to-signal ratio usually made it somewhere I didn't want to return to.
I was just thinking to my self the other day how it's nice I don't need to stop what I'm doing to make a question that's answerable by someone else. Ai can answer my question without me spending time recreating the problem and stripping out all of the irrelevant context
It was pretty dang useful when there was no alternative, and I’m sure that many people physically could not have performed their jobs without copy-pasting from it.
But yeah, I don’t know how anyone could have any affection or nostalgia for it, people were massive jerks and it was not a pleasant place.
I thought StackOverflow was pretty great. This is an unpopular opinion but I think a lot of the questions that were closed really deserved to be closed. Otherwise it would have been a firehose of the same basic questions over and over again. For every person who posted a question and got mad that it was closed, there were probably 100 people who googled something and found a useful StackOverflow answer that was relevant and useful to them although they never posted their own question or even made an account on the site.
This is why you never pay $1.8 billion for a social media company.
It never ends well for the new owner. Not just Stack Overflow but also Tumblr, Vine, MySpace, Twitter, and more. Instagram might be the only exception.
Good job on the founders for selling at the peak though.
Insta and YT were bought fairly early on the ascent and then pumped hard with some resources and autonomy before being folded in completely. WhatsApp is a legit counterpoint though.
A lot of companies don't hire at the end of the year because of holidays breaking up schedules, coupled with strapped budgets by that point. New year starts, budgets are refreshed, everyone is back to work so hiring pipelines can roll again. Get hired in early February, on-board for a couple of weeks, really start to dig into work at the end of February and early March. You don't want to look like you don't know anything, so instead of asking your coworker sitting next to you or just a Slack DM away, you throw a question up on Stack Overflow and hope to get an answer that gets you unstuck.
Both before my time. These days it seems like every site is able to withstand pretty much every controversy. Facebook should have died about 5 times by now but the company is as strong as ever.
> 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.
At work we had a dispute over if AI should be allowed in the technical interview, we resolved it by both running an AI allowed and not allowed interview. Something interesting we found is that every candidate either passed or failed both. People who could not program manually without AI were not able to get the agent to complete the tasks either.
I've seen people type questions in to the LLM and get the answer they asked for but not the one they needed/wanted because they didn't have the correct terminology.
> Surely the desired state isn't that nobody knows how to write code any more right?
Shaping up like that in my org. At least one mid-career dev says he no longer looks at code.
I still look at code and find that agents work best when I write the foundation and then vibe on top of my hand-written code. Works extremely well because agent picks up my style accurately.
Hopefully your management is trying to answer the following question: is said middle-career dev outproducing their past self, and others who still look at code, with:
1) submitted changes that don't need any more revision than their previous human-written ones when it comes to code review?
2) no increase in bug incidents
3) no slow-down in peer work or future work caused by humans-or-agents having to fight increasingly overly-complicated, poorly-factored, copypasta-style code or god methods? (this might not be evident yet)
(Another question is how well is this person doing their job as a reviewer, making sure to keep the product quality bar high, without looking at code?)
Anyone in an org with coworkers no longer writing code needs to be making sure their managers have a pulse on the long-term health of the product to see who's doing it well (lots of test coverage, shipping only super-high-quality, refined-from-multiple angles stuff) or just being lazy (shipping first drafts that continually add debt to various files and methods).
It's generally and simply an encoding of what amounts to binary machine code which you translate via assembly code acting as a deterministic compiler from assembly to machine code if you are doing it manually.
LLMs aren't a deterministic process and human languages aren't as clear as machine code and assembly.
If I transported you to the 1960s and gave you a wizard that could punch cards for you with a chance of making a mistake, would you still bother to learn how to operate a punch card?
What would you do if the wizard gets stuck? Coarse the wizard into making the black box work through somebody else's direct perspective on the problem?
It's more like a restaurant. You give an order and a little while later, a finished dish appears.
The difference between a Chipotle and a Michelin starred establishment is that Chipotle is just assembling a mass produced good. A Michelin chef knows their ingredients inside and out; knows the science of how those ingredients work; knows varied techniques to extract flavors, create textures, etc.
Anyone can work in a Chipotle; few can achieve a Michelin star.
I remember! You created a control card, with tab stops and other controls, wrapped it around a control drum, and then had an easy time punching your source FORTRAN!
I just looked and found my old control drum, in the back of my junk drawer. But I can't find an old punch card machine in there, most have lost it somehow.
And to operate a self-driving car safely you need to keep your attention on the road so you can take over quickly when needed.
But that's not how human nature works. Most people take the path of least resistance. Especially when the primary purpose of the invention is to offer convenience.
I was wondering about this myself, but given everything I know about AI. Won't the vocabulary slowly and subtly change as common people try to develop software, not knowing the jargon? Won't the AI systems learn from the prompts and adjust their understanding of what's trying to be accomplished?
> 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 read programming books and use LLMs for different purposes. With books, it's usually not to find a solution to the very specific problem I'm working on. That's what I use LLMs for because they give very focused answers. Books, on the other hand, provide much broader context that help me learn a language. Whereas with LLMs I get a solution yet tend to retain nothing. YMMV.
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.
This is a good investment. Your fingers will remember things long down the road and you will be better at having an AI bullshit detector for code development.
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
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 tried to get started with programming with books. But I just didn't seem to be getting anything, I'd read the chapters and not really learn or understand it. What really worked was interactive education like Codecademy and some others I have forgotten the name of.
Reading a small paragraph and then immediately putting it in to action made everything clear far better than books did.
It's fascinating the different ways human minds work and learn. I'm the same way.
It also shows up in other areas like language learning where some people prefer classes and grammar books, and others prefer to just learn via exposure to a lot of content.
I think this is probably just the common experience. Programming is probably best learned hands on rather than through a book, which is why the use of programming books has fallen off a cliff once we got other options. Even before AI I think programming books had already fallen off in popularity.
There would be some things books can provide that are probably better than other options, but for a lot of hands on skills it seems best to learn in a hands on way.
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'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.
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.
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.
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.
It depends on what kind of reference book you get.
A reference book on a particular language is going to have a pretty short useful lifetime, since any language of significant popularity will evolve relatively quickly.
In contrast, a reference book on general programming knowledge (e.g., design patterns) could very well last a lifetime.
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.
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.
I enjoy reading really old programming books, the 1997 edition of Learning Perl mentioned in the article being a perfect example. I don't fret over the exercises, but if it's well-written it gives a glimpse into how people thought about technology/code/computers at that point in time, like the tech equivalent of flipping through old newspapers.
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 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.
I almost feel nostalgic for the bittorrent era, when piracy (or unauthorized distribution) was done the old-fashioned way!
Though it's a shame that current copyright law is incompatible with building an effective digital library that isn't crippled with restrictions designed to impose the limitations of paper books (or worse) onto ebooks (while removing benefits such as first-sale doctrine.)
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.
> 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.
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.
My career kicked into high gear some time around 2008. I saw somewhere online where a publisher was seeking a volunteer book reviewer / junior editor.
I volunteered, did the best job I could, and posted an honest review via blog. I got more review requests, and a few other publishers contacted me for the same.
I didn’t really master much, because I didn’t put hands on keyboard for a lot of it. But I got a good view of the technical landscape, and I accumulated a nice paperback library.
Before too long, the free books became free ebooks and some of my contacts needed renewing as natural career progression took place. I let my ‘hobby’ die off as I dug deeper in the topics that interested me.
So that era passed. I still have several books with my name in the credits, sort of a souvenir set from the time.
It's a bit different than that analogy would suggest. Learning things piecemeal can leave strange gaps in one's knowledge, in my experience. A book is often much quicker.
You can't just buy a horse and park it in a garage. You need to exercise it, give it vet care, shoes, feed it, deal with poop, etc. Or, pay someone to do that.
Unless you live in a place with dirt roads, or really love horses, I think a beater Toyota would win in terms of time and cost.
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.)
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.
Getting a book is my goto to learn anything new. I taught myself PHP and ended up finding a book 4 years later when I was looking for an answer to something. On the next page was something that would have saved me tons of time so I read the whole book.
Since then I’ve read books on Ruby, Go, Elixir, Docker, K8s and a lot more. By far the best way to get a semi complete understanding of anything without scraping together data from the internet yourself, because you won’t easily know the gaps.
I have a gut feeling that human as a creature learns better when looking at the information from several different angles, both physically and mentally. Been physically I mean looking at the same concept on screen and on hard copy books, perhaps taking notes and mark relevent sentences with a highlighter. Similarly, seeing a concept on physical book and write some short code snippet is viewing the concept from different mental angles. Though I don't have a proof for that and have yet to find a formal research on this topic.
I learn the same way but there are different people who learn in different ways. Also some people come with some concepts already understood from past experience or education, it’s easier for to pick things up without needing various angles.
- PAIP - The Art of of the Metaobject Protocol - Modern C
Those alone are the ones I've been re-reading this year.
I came here to say almost the same thing. I've been learning rust in my free time because I don't do enough programming at work to scratch that itch any more, and I've been using the rust book as a reference.
Thanks for the rec for "rust for rustaceans" I'll have a look into it.
I've only been using chatgpt for points where i'd normally go ask another dev for another set of eyes to debug something, otherwise all my learning and doing has been mostly the rust book, crates, and blogs about rust, ecs, roguelikes etc etc. It's been so fun!
My method is to get some hands-on experience first with the technology: Tutorials, Getting Started pages, the first chapter of a book. The fumble for some time and getting exposed to various sources of information. Then I take a book and skim it from start to cover, stopping at interesting bits. Then I reread various parts.
Sometimes I start with reading the books. But I already known I won't retain anything deeply. But it will gives me all the right keywords to learn more about the technology.
It always amaze me when I see fellow programmers struggle with problems that could have been solved easily by just reading that introductory book on the subject.
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 :).
Unfortunately even in the old days, a truly good programming book like you’re describing was depressingly rare.
Younger me really enjoyed some of the game programming books by Andre Lamothe.
Most “Learn Language X” books were terrible with over focus on syntax and very little thought into organization.
Apparently the guy who wrote the Camel book on Perl made less than $1000 from that book. I was shocked when I heard about that because back in the day when I was learning that book was incredibly popular and seemed to be everywhere.
Agreed, Books on specific programming language were indeed tricky.
I found books on architecture, systems, or patterns, were more available. E.g. On relational database optimization principles, or Unix system administration, or graphics algorithms and rendering math, etc :)
I've commented on this development before on HN, so I'm glad to see this post on the front page. From a few months back:
"...the fact of the matter is that kids getting into high tech and programming mostly don't read books anymore. How do I know? Recently I was hanging out with a bunch of high school students who asked me how I learned. I said it was mostly via books and man pages. "Yeah, don't sleep on high quality written material. O'Reilly. Wiley. Addison-Wesley. Manning. MIT. No Starch Press. &c...
"Well. You should have seen the look on their faces. I might as well have morphed into the Steve Buscemi meme "How do you do, fellow kids?" They looked at me like I was a total relic or greybeard and said things like "Nah, nobody reads tech books anymore; I learned Typescript from YouTube videos."
> 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...
Did anyone actually like StackOverflow?
Any question asked would be edited beyond recognition (and usually into brash rudeness). Half the answers were demanding ever increasing proof of work, and the other half told the OP that they shouldn't even be trying to do what they're doing. The only useful thing were opinion based posts from people with domain expertise, and SO kept trying to ban and remove those. It was the least helpful place online, but the most accessible, and it survived for lack of alternatives.
I'm no AI booster, but answering simple questions about well understood topics is a perfect fit for it. Good riddance to StackOverflow.
Stack Overflow was a nice experience for me because I was able to hit 2k reputation fairly quickly, in just 30 days of posting and 6 weeks calendar time. That being said, it never had the community feel of places I spent during my formative years, which were more on forums and IRC.
Here's a conference talk I gave on how to gain Stack Overflow reputation from back in 2018, selected out of 5 submitted talks. It's amazing how fast times have changed from before, during, and now after.
https://grokify.github.io/stackoverflow-the-hard-way/
I had a pretty good time asking a question about Prolog. It was a really interesting experience knowing that there's someone out there that high proficiency in a very niche language, patiently explaining to me an issue that they have probably heard a million times from yet another imperative programmer. They even have their own website advocating for Prolog, etc.
Now, I could imagine an LLM would be able to do the same. However, I understand that this is only possible because of people like them. I don't think the youngins that started with LLM directly would appreciate the humongous amount of data and discussions online that enables that. The internet is so much bigger than just Google, Facebook, Youtube and Twitter.
I did originally, when it collected a bunch of obscure knowledge and made it searchable and useful. It was fun and rewarding to put things you knew into the common knowledge pool, and everyone celebrated a successful competitor to Experts Exchange. The SO model had a few major flaws that became impossible to ignore after it was entrenched. First, the reward scheme rewards the exact opposite of what it should incentivise: common questions are hit by many users and therefore attract lots of upvotes while answering the really hard stuff often meant you didn't even get your answer marked as "accepted" (because the OP had given up and stopped checking the site). Second, the site deliberately cultivated an "editor caste" in the Wikipedia style before the failure modes of that model were well-known: well-intentioned newbies get shut down by miserable yet untouchable people who play (and sometimes help write) the site's rules. Third, the stated desire to identify canonical answers to questions had no clear way to handle the evolution of the software people were talking about. So you'd have highly upvoted answers that might have been referencing deprecated libraries, and it was very hard for the newer answer to gain traction via either internal or external search.
It was also unfortunately before the retro boom of the 2020s, so questions about older arcana were often vulnerable to being closed instead of answered.
Yes, I liked it. When it debuted it was a massive improvement over expertsexchange, who had previously dominated the Google searches with bait and switch previews.
It may not have aged well but to say it was always crap is rewriting history.
StackOverflow was great when I was a very junior dev working on JavaScript apps. Anytime I ran into a roadblock, there was often a relevant post there to help me. As I become more competent though, I realized that reading the documentation directly was usually a much better way to get answers to my questions, and I stopped visiting.
I have found StackOverflow useful on rare occasion, but the friction, idiotic moderation culture, and high noise-to-signal ratio usually made it somewhere I didn't want to return to.
I was just thinking to my self the other day how it's nice I don't need to stop what I'm doing to make a question that's answerable by someone else. Ai can answer my question without me spending time recreating the problem and stripping out all of the irrelevant context
It was pretty dang useful when there was no alternative, and I’m sure that many people physically could not have performed their jobs without copy-pasting from it.
But yeah, I don’t know how anyone could have any affection or nostalgia for it, people were massive jerks and it was not a pleasant place.
I thought StackOverflow was pretty great. This is an unpopular opinion but I think a lot of the questions that were closed really deserved to be closed. Otherwise it would have been a firehose of the same basic questions over and over again. For every person who posted a question and got mad that it was closed, there were probably 100 people who googled something and found a useful StackOverflow answer that was relevant and useful to them although they never posted their own question or even made an account on the site.
This is why you never pay $1.8 billion for a social media company.
It never ends well for the new owner. Not just Stack Overflow but also Tumblr, Vine, MySpace, Twitter, and more. Instagram might be the only exception.
Good job on the founders for selling at the peak though.
Instagram, YouTube, WhatsApp seemed to work out well.
It’s always more complex than you think
Insta and YT were bought fairly early on the ascent and then pumped hard with some resources and autonomy before being folded in completely. WhatsApp is a legit counterpoint though.
Are YouTube and WhatsApp social media?
Interesting looking at the data, the questions always seemed to peek in March. Anyone have insight into that?
A lot of companies don't hire at the end of the year because of holidays breaking up schedules, coupled with strapped budgets by that point. New year starts, budgets are refreshed, everyone is back to work so hiring pipelines can roll again. Get hired in early February, on-board for a couple of weeks, really start to dig into work at the end of February and early March. You don't want to look like you don't know anything, so instead of asking your coworker sitting next to you or just a Slack DM away, you throw a question up on Stack Overflow and hope to get an answer that gets you unstuck.
Too early to call, but that is a crazy stat. Wow.
Too early to call? It's hit rock bottom. I've never seen a major site die so completely before.
Groupon? Did an IPO with a 12.7b valuation and was mostly dead within a year.
Google Search was looking bleak at one point, recently, based on deltas alone.
Digg was a major site that had a surprisingly fast turn of events. There are some good lessons from that.
MySpace comes to mind. Slashdot maybe.
Both before my time. These days it seems like every site is able to withstand pretty much every controversy. Facebook should have died about 5 times by now but the company is as strong as ever.
> 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.
At work we had a dispute over if AI should be allowed in the technical interview, we resolved it by both running an AI allowed and not allowed interview. Something interesting we found is that every candidate either passed or failed both. People who could not program manually without AI were not able to get the agent to complete the tasks either.
I've seen people type questions in to the LLM and get the answer they asked for but not the one they needed/wanted because they didn't have the correct terminology.
Junior devs should still read to learn how to write the code.
Surely the desired state isn't that nobody knows how to write code any more right?
I still look at code and find that agents work best when I write the foundation and then vibe on top of my hand-written code. Works extremely well because agent picks up my style accurately.
Hopefully your management is trying to answer the following question: is said middle-career dev outproducing their past self, and others who still look at code, with:
1) submitted changes that don't need any more revision than their previous human-written ones when it comes to code review?
2) no increase in bug incidents
3) no slow-down in peer work or future work caused by humans-or-agents having to fight increasingly overly-complicated, poorly-factored, copypasta-style code or god methods? (this might not be evident yet)
(Another question is how well is this person doing their job as a reviewer, making sure to keep the product quality bar high, without looking at code?)
Anyone in an org with coworkers no longer writing code needs to be making sure their managers have a pulse on the long-term health of the product to see who's doing it well (lots of test coverage, shipping only super-high-quality, refined-from-multiple angles stuff) or just being lazy (shipping first drafts that continually add debt to various files and methods).
Do you know how to operate a punch card?
Yes, and IBM has current documentation if you need to that has been updated in 2026: https://www.ibm.com/docs/en/zos/3.2.0?topic=considerations-u...
It's generally and simply an encoding of what amounts to binary machine code which you translate via assembly code acting as a deterministic compiler from assembly to machine code if you are doing it manually.
LLMs aren't a deterministic process and human languages aren't as clear as machine code and assembly.
Yes. But Python isn't punch cards behind the scenes so it's not the same thing at all.
Besides. You're not asking <AGENT OF THE WEEK> to produce punch cards to jam into the PDP.
If I transported you to the 1960s and gave you a wizard that could punch cards for you with a chance of making a mistake, would you still bother to learn how to operate a punch card?
What would you do if the wizard gets stuck? Coarse the wizard into making the black box work through somebody else's direct perspective on the problem?
I don't think this is comparable.
It's more like a restaurant. You give an order and a little while later, a finished dish appears.
The difference between a Chipotle and a Michelin starred establishment is that Chipotle is just assembling a mass produced good. A Michelin chef knows their ingredients inside and out; knows the science of how those ingredients work; knows varied techniques to extract flavors, create textures, etc.
Anyone can work in a Chipotle; few can achieve a Michelin star.
> Do you know how to operate a punch card?
I remember! You created a control card, with tab stops and other controls, wrapped it around a control drum, and then had an easy time punching your source FORTRAN!
I just looked and found my old control drum, in the back of my junk drawer. But I can't find an old punch card machine in there, most have lost it somehow.
Do you maintain a system in which punch cards play a critical role?
Do you let your Jenkins re-inference your entire program from markdown files on each push?
And to operate a self-driving car safely you need to keep your attention on the road so you can take over quickly when needed.
But that's not how human nature works. Most people take the path of least resistance. Especially when the primary purpose of the invention is to offer convenience.
I was wondering about this myself, but given everything I know about AI. Won't the vocabulary slowly and subtly change as common people try to develop software, not knowing the jargon? Won't the AI systems learn from the prompts and adjust their understanding of what's trying to be accomplished?
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.
This is a good investment. Your fingers will remember things long down the road and you will be better at having an AI bullshit detector for code development.
would you mind sharing the names of the books you bought?
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
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 tried to get started with programming with books. But I just didn't seem to be getting anything, I'd read the chapters and not really learn or understand it. What really worked was interactive education like Codecademy and some others I have forgotten the name of.
Reading a small paragraph and then immediately putting it in to action made everything clear far better than books did.
It's fascinating the different ways human minds work and learn. I'm the same way.
It also shows up in other areas like language learning where some people prefer classes and grammar books, and others prefer to just learn via exposure to a lot of content.
I think this is probably just the common experience. Programming is probably best learned hands on rather than through a book, which is why the use of programming books has fallen off a cliff once we got other options. Even before AI I think programming books had already fallen off in popularity.
There would be some things books can provide that are probably better than other options, but for a lot of hands on skills it seems best to learn in a hands on way.
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'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.
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.
There's a wide world outside of tech jobs. Most of it isn't immensely lucrative, but I hear it can be quite fulfilling
Hey if it means anything there's other people here in the opt out group with you
Maybe we can build something else in parallel for the other opt-outers
This predates LLMs. The internet has been the primary source of programming knowledge for decades.
Books are still good for the fundamentals of course.
The “higher level of abstraction” phrase used to describe what LLMs are in relation to programming… That phrase needs to die.
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.
"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...
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.
It depends on what kind of reference book you get.
A reference book on a particular language is going to have a pretty short useful lifetime, since any language of significant popularity will evolve relatively quickly.
In contrast, a reference book on general programming knowledge (e.g., design patterns) could very well last a lifetime.
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 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...
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.
I enjoy reading really old programming books, the 1997 edition of Learning Perl mentioned in the article being a perfect example. I don't fret over the exercises, but if it's well-written it gives a glimpse into how people thought about technology/code/computers at that point in time, like the tech equivalent of flipping through old newspapers.
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 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.
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 almost feel nostalgic for the bittorrent era, when piracy (or unauthorized distribution) was done the old-fashioned way!
Though it's a shame that current copyright law is incompatible with building an effective digital library that isn't crippled with restrictions designed to impose the limitations of paper books (or worse) onto ebooks (while removing benefits such as first-sale doctrine.)
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.
> 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.
> 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.
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.
LLMs are to a large degree trained on pirated books.
Remember man pages to learn an write C. Guided AI is good if it learns from a book not crap code found on GitHub.
I still even now feel that K&R C should be a mandatory reading for CS students, but alas.
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.
My career kicked into high gear some time around 2008. I saw somewhere online where a publisher was seeking a volunteer book reviewer / junior editor.
I volunteered, did the best job I could, and posted an honest review via blog. I got more review requests, and a few other publishers contacted me for the same.
I didn’t really master much, because I didn’t put hands on keyboard for a lot of it. But I got a good view of the technical landscape, and I accumulated a nice paperback library.
Before too long, the free books became free ebooks and some of my contacts needed renewing as natural career progression took place. I let my ‘hobby’ die off as I dug deeper in the topics that interested me.
So that era passed. I still have several books with my name in the credits, sort of a souvenir set from the time.
Nobody uses a horse and cart as an every day method of commuting anymore.
It's a bit different than that analogy would suggest. Learning things piecemeal can leave strange gaps in one's knowledge, in my experience. A book is often much quicker.
People still exercise and make up physical activities to compensate for the more sedentary lifestyle, though.
Horse and cart might be antiquated and slow, but my god are they so much cheaper and more reliable than the modern car.
Many parts of the world still rely on horse and cart today, even modern societies.
You can't just buy a horse and park it in a garage. You need to exercise it, give it vet care, shoes, feed it, deal with poop, etc. Or, pay someone to do that.
Unless you live in a place with dirt roads, or really love horses, I think a beater Toyota would win in terms of time and cost.