Nobody cracks open a programming book anymore
Programming books and traditional resources like man pages and Stack Overflow are losing ground to AI coding assistants and web tutorials, raising questions about how new developers learn. Many argue that books still provide irreplaceable structured, big-picture understanding and shared vocabulary, while LLMs excel at answering narrow, contextual questions but can entrench shallow or piecemeal knowledge. Others counter that hands-on, interactive learning has long surpassed books for most people and that AI is simply the latest tool, though concerns remain about long-term skill development, code quality, and the sustainability of technical publishing.
Role of Programming Books vs. Other Resources
- Many say books are no longer primary; web docs, tutorials, and now LLMs cover most day‑to‑day needs.
- Others report buying and reading more books recently because they provide curated scope and order, compared to overwhelming, fragmented online info.
- Several distinguish: use LLMs/docs for specific problems, books for fundamentals, architecture, and “big picture.”
- Some argue books were never ideal for learning languages; interactive platforms, REPLs, and hands‑on projects work better for many.
Learning Styles and Mental Models
- Strong theme: people learn differently. Some need interactive exercises; others happily learn by reading dense texts.
- Advocates of books emphasize:
- Deliberate sequencing of concepts.
- Building a coherent mental model, avoiding random “piecemeal” knowledge and gaps.
- Exposure to peripheral ideas you wouldn’t think to search for.
- Critics counter that many language books are shallow, syntax‑heavy, or quickly obsolete; theoretical and systems books age better.
LLMs, Skills, and Code Quality
- Common practice: pair books with LLMs—read a chapter, then query AI for clarification, tangents, or debugging help.
- Concern that LLM‑driven workflows can freeze or erode genuine skill growth; people may ship “first drafts” and accumulate debt.
- Multiple comments stress juniors still need to learn to code and understand concepts to:
- Guide agents with correct vocabulary.
- Evaluate and review AI‑generated code.
- Interview anecdote: candidates who couldn’t code manually also failed with AI assistance.
Economics, Piracy, and Decline of Book Culture
- Reported sales data for one language book show tens of thousands of copies, but revenue is modest; most income now comes from subscription platforms.
- High retail prices and cheaper or pirated PDFs (Libgen, Anna’s Archives) push younger devs away from buying.
- Several note LLMs were trained on “pirated” books and online Q&A; some see this as theft, others as analogous to earlier digital copying.
Stack Overflow and AI
- SO question volume has dropped sharply; some see it as “dying,” replaced by LLMs for routine questions.
- Views on SO are mixed: invaluable in its prime but plagued by harsh moderation and aging, obsolete answers.
- Worry that as sites and blogs decline, LLM training data and the broader technical ecosystem will suffer.
Nostalgia and Timeless Texts
- Many cherish classic books (C, OS, networking, algorithms, SICP, etc.) as milestones and enduring references.
- Some fear that with AI and fewer books, future developers will lack that shared canon and personal “history on the shelf.”