You search for a book on prompt engineering and get a page of listings, half of them self-published and some with titles that sound like they were generated. Meanwhile the field changes monthly, so a book from 2023 can feel like a museum piece.
I checked eight books against publisher or author pages (and, where those blocked me, library and bookseller catalog records) so every title, author and year below is real. I have not read all of them cover to cover, and I say so per book. The goal is a short list with an honest answer to "who should skip this one".
How I verified these
Date: 2026-10-08. For each book I tried the publisher page first. Manning's and Penguin Random House's pages loaded, and so did Chip Huyen's own book page. O'Reilly's pages returned HTTP 403 to my fetch tool, so for the O'Reilly titles I cross-checked Open Library records and search results that quoted O'Reilly's catalog. Where sources disagreed on the exact month, I give the year or the month that O'Reilly's own listing was reported to use, and say so. I did not read the books for this post. The "what it's for" notes come from publisher descriptions and the titles' stated scope, and I label my own judgment as judgment.
Prices are omitted. They vary by country and format, and you should look them up wherever you buy.
Quick picks
| If you want to... | Start with |
|---|---|
| Understand how LLMs work, with code | Hands-On Large Language Models |
| Build LLM products end to end | AI Engineering |
| Learn prompting as an engineering practice | Prompt Engineering for LLMs |
| Build agents in Python | AI Agents in Action |
| Understand models by building one | Build a Large Language Model (From Scratch) |
| Get a non-technical overview | Co-Intelligence |
Prompting and LLM application books
Prompt Engineering for LLMs, John Berryman and Albert Ziegler (O'Reilly, November 2024)
Subtitle: The Art and Science of Building Large Language Model-Based Applications. O'Reilly's listing was reported as November 2024; some retailer records show late December 2024 or a 2025 copyright year, probably print and cataloguing dates.
What it's for: Thinking about prompts as part of an application, not as one-off chat messages. The subtitle points at building LLM-based applications, which is the right frame if you ship software.
Skip it if you only chat with ChatGPT and want tips. You'll want something lighter. Aging: model-specific advice will date; the application-level thinking holds up better. I haven't read it.
Prompt Engineering for Generative AI, James Phoenix and Mike Taylor (O'Reilly, 2024)
Subtitle: Future-Proof Inputs for Reliable AI Outputs at Scale. Catalog records list it as a 2024 first edition.
What it's for: A broader treatment that includes text and image prompting, aimed at people producing reliable outputs repeatedly. A library catalog and summary sites describe it as covering prompts for both text and images.
Skip it if you only care about agents or about retrieval and evaluation in depth. It's a prompting book more than a systems book. Aging: image-model sections are the likeliest to be stale. I haven't read it, and I can't tell you which edition-level changes have been made since release.
AI Engineering, Chip Huyen (O'Reilly, 2025)
The author's own page lists O'Reilly as publisher, 2025 as the year, and describes it as a guide to building applications on existing foundation models, including a framework for developing and deploying an app and a survey of models, datasets, evaluation benchmarks and application patterns. Open Library lists a 2024 first-publish year, probably from an early-release date.
What it's for: The engineering around the model: how to evaluate, choose, and ship. If you'd otherwise end up reading ten scattered posts on evals and deployment, this is the consolidated route.
Skip it if you want copy-paste prompts. It's not that kind of book. Aging: benchmarks and model names will date fast; the evaluation mindset won't. My judgment: this is the book to buy if you're only buying one for building. Our own evaluation dataset guide covers a small slice of the same ground for free.
Hands-On Large Language Models, Jay Alammar and Maarten Grootendorst (O'Reilly, September 2024)
Search results quoting O'Reilly's catalog list September 2024 and about 428 pages. The scope described there includes Transformer architecture, semantic search, fine-tuning, and in-context learning.
What it's for: Understanding what's inside the box, with code, without needing a research background. It's the best fit here for the "why does it behave like that" questions that make you better at prompting.
Skip it if you want a business or no-code view. It's for people comfortable running notebooks. Aging: architecture chapters are stable; library calls may need updating. I haven't read it. Our free RAG lesson is a lighter route into one of its topics.
Building and understanding models
Build a Large Language Model (From Scratch), Sebastian Raschka (Manning, September 2024)
Manning's page: 368 pages, ISBN 9781633437166. It teaches you to code a GPT-style model that runs on a laptop without leaning on existing LLM libraries, covering data preparation, pretraining and fine-tuning for classification and instruction-following. Prerequisites per Manning: intermediate Python and some machine-learning knowledge.
What it's for: Learning by building. If you want to know what attention actually computes, this is the way.
Skip it if your goal is shipping an app next month. Building a small model teaches fundamentals but won't get you a product faster. Aging: the fundamentals age slowly, which is the argument for it. I haven't read it.
AI agents books
Agent tooling moves faster than book production. Treat any framework-specific code in these as a starting point and check current docs.
AI Agents in Action, Micheal Lanham (Manning, February 2025)
Manning's page: 344 pages, ISBN 9781633436343, aimed at intermediate Python programmers. It covers building LLM-powered agents and assistants, memory and knowledge systems, multi-agent orchestration, and uses tools including LangChain, Prompt Flow, AutoGen and CrewAI. It also covers the OpenAI Assistants API.
What it's for: A broad hands-on tour of agent frameworks in Python.
Skip it if you want one clean, current stack. A book that tours several frameworks and a specific vendor API (the Assistants API here) risks chapters that no longer match the current ecosystem. I haven't checked which, if any, have been overtaken; check the table of contents against the tools you plan to use. For free grounding first, start with building your first AI agent and AI agent design patterns.
Building Applications with AI Agents, Michael Albada (O'Reilly, 2025)
Subtitle: Designing and Implementing Multiagent Systems. Listings give the English edition ISBN 9781098176501. Dates differ by source: one O'Reilly platform listing said September 2025, while distributor records said 21 October 2025. I would say "autumn 2025" and not argue the day.
What it's for: The design side of multi-agent systems, per the subtitle. A reasonable next book after you've built one simple agent.
Skip it if you haven't yet built a single-agent loop. Start smaller. Also skip if you want a framework tutorial; I haven't read it and can't say how code-heavy it is. Our post on when multiple agents are worth it is a sensible check before you take on that complexity.
Non-technical
Co-Intelligence: Living and Working with AI, Ethan Mollick (Portfolio, 2 April 2024)
Penguin Random House's page lists the publisher as Portfolio and the on-sale date as April 2, 2024. It describes a practical guide to working with generative AI as co-worker, co-teacher and coach, drawing on business and education examples.
What it's for: Orienting a manager, teacher or curious reader, without code.
Skip it if you want prompt recipes or any technical depth. It's a way of thinking, not a manual. Aging: the 2024 model examples are dated; the argument is the point. I haven't read it in full.
Books I left out, and why
- Anything I couldn't confirm exists. Search results for this topic are full of self-published titles with templated names and a flood of repackaged content. If I couldn't find a publisher or author page or a library record, it isn't here.
- Older NLP textbooks. They're good, but this list is for prompting, applications and agents. Starting there is a different path.
- LLM Engineer's Handbook and similar. I found no problem with it, but I didn't verify it on a publisher page that loaded, so I'm not recommending a book I couldn't check.
Should you buy a book at all?
Often, no. Here's how I'd decide.
- Beginner who wants better answers from ChatGPT or Claude: skip books for now. Practise on real work, and use free resources like our prompt library and what is prompt engineering.
- Developer moving into LLM apps: AI Engineering first, then Hands-On Large Language Models if you want the internals.
- Developer who already ships apps and wants agents: read free posts on single agents first, then one agent book, after you've checked its table of contents against your stack. See the AI engineering career roadmap for how this fits a longer plan.
- Curious generalist: Co-Intelligence, then practice.
One last rule: read the sample chapter and the table of contents before buying anything on this fast-moving topic. If the book's code depends on a product that has since changed, you want to find out before you've paid for it.
