The computer book section isn't gone yet. But it's getting smaller. In some stores it's down to a rack of six titles, three of which are about ChatGPT.
The numbers are stark. Computer book sales fell 16.9% year-over-year in the first nine months of 2023. The "professional books" segment — the category that covers technical reference, the stuff your employer used to buy you — was down 22.3% in August 2025. And then, quietly, Publishers Weekly simply stopped reporting the category. Not a press conference. Not a Napster moment. Just silence.
"The category doesn't die, it just stops being talked about."
What actually changed
The demand hasn't disappeared — it's been rerouted. ChatGPT now has 900 million monthly active users. GitHub Copilot has 4.7 million paying subscribers as of January 2026, up roughly 75% in a year. Stack Overflow is receiving about 3,800 questions a month — which is exactly what it was getting in 2008, before it had even finished launching.
The chatbots absorbed the demand that programming books used to serve. You have a question about idempotency or regex or SQL indexes — you ask the model, get a precise answer in the number of words you need, and close the tab.
The thing that's actually going away
The author of the original post lands on something worth sitting with:
"Knowledge, for working programmers, was always the residue of typing. Of doing. The typing was the practice. What is going away is the typing."
The programming book was the wrong format for the content — printed static text describing dynamic software, requiring readers to retype examples by hand. But that friction was the point. You couldn't fake your way through 400 pages. The slowness was the mechanism by which knowledge stuck.
The chatbot has read every book and forgotten the point of every one of them. It'll explain anything in exactly the words you need, instantly, in a way that requires no effort from you — and so leaves no residue.
What this means for how developers learn
This isn't a doom narrative. The kid learning to code by chatting with an agent isn't a worse programmer — they're a different one. Working at a higher level of abstraction from day one. That'll produce things we can't predict.
But the shift matters for anyone thinking about developer education, onboarding, or how teams build shared knowledge:
- The passive consumption trap is real. Getting an answer from a model and getting that answer into your head are different things. Deliberate practice still matters — the form just has to change.
- Depth still requires friction. Deep dives, side projects, building things that break: these are the replacements for the 400-page book. The mechanism is different; the need is the same.
- Technical publishing isn't dead, it's transforming. What survives will be the stuff models can't replace: opinionated takes, hard-won experience, context that isn't in the training data.
The O'Reilly animal books were always a workaround — an imperfect medium for a problem that now has a better solution. What we're figuring out, collectively, is what the better solution costs us.
Read the original: unix.foo
✏️ Drafted with KewBot (AI), edited and approved by Drew.
Top comments (0)