Manning Publications Co.

AI agents can sometimes feel like magic hidden behind frameworks and black box libraries. But once you understand what's happening under the hood, they become much easier to reason about, debug, and improve. Build an AI Agent (From Scratch) by Jungjun Hur 허정준 and Younghee Song walks you through designing, building, and deploying a system of autonomous AI agents step by step. You'll create a sophisticated research agent in Python capable of tackling complex, multi step tasks from the GAIA (General AI Assistants) benchmark. Instead of treating agent frameworks as magic, you'll see how the entire system works end to end so you can confidently build your own. Read a chapter for free: https://hubs.la/Q04rvl0d0 Learn more: https://hubs.la/Q04rvfGv0

AI isn't just changing what software can do. It's changing how we have to secure it. Traditional application security doesn't fully prepare you for systems that generate content, make decisions, and interact with the world. AI introduces new attack surfaces, where attackers target prompts, context, models, and intent instead of just code. In A Complete Plan for AI Security, Derek Fisher explains how to approach security across the AI lifecycle. You'll learn how to design secure-by-default AI applications, understand emerging attack techniques, and build systems that are resilient, compliant, and ready for production. Get the liveVideo for $10 during our Summer Sale: https://hubs.la/Q04rr9Mj0

Most developers agree that business logic belongs at the center of an application, independent of frameworks, databases, UIs, and external services. The challenge is translating that principle into code. Clean Applications with Hexagonal Architecture by Enrique Medina Montenegro bridges that gap. Starting with the concepts behind hexagonal architecture, it shows you how to apply them in a real Java and Spring Boot application, building outward from a domain-focused core. If you've ever found yourself wondering, "Where does this code belong?" this book gives you a practical framework for making those decisions with confidence. Now available through the Manning Early Access Program, and 50% off through August 13th: https://hubs.la/Q04rqXjx0

Reducing inference costs isn't just about buying cheaper hardware. It's about understanding which optimizations actually move the needle. Tomorrow at The Fifth Elephant Conference by Hasgeek, Vivek Kalyanarangan will show how IDfy reduced ML inference costs by up to 10× across a fleet of 40+ production models using quantization and graph fusion without sacrificing latency, throughput, or accuracy. Rather than treating GPU costs as fixed, the talk focuses on systematically matching each workload to the least expensive hardware that still meets production requirements. If you're building or operating ML systems at scale, it's a practical look at making inference more efficient through repeatable engineering rather than one-off optimizations. Learn more about the session: https://hubs.la/Q04rmVJR0 Can't make the conference? Quantization and Fast Inference covers the techniques in depth. Get 45% off through August 22nd with code tfecon26: https://hubs.la/Q04rmK1Q0

Bayesian statistics can seem intimidating if math isn't your first language. But understanding uncertainty, probability, and inference is becoming increasingly valuable in fields ranging from data science to AI. Grokking Bayes by Quan Nguyen builds that intuition first through clear explanations, visual illustrations, and relatable examples before translating those ideas into practical Python code with PyMC and ArviZ. Along the way, you'll see how Bayesian thinking connects to modern AI topics, including uncertainty-aware deep learning and LLM applications. Get it for 50% off through tomorrow during our Summer Sale: https://hubs.la/Q04rlNgT0

Most AI conversations still revolve around prompts. But if you want to build AI that solves real problems, that's only the beginning. The next step is understanding how modern AI systems are built, from training and aligning models to designing agentic applications and using AI to navigate large, unfamiliar codebases. If you're looking for a practical learning path, these four books are a strong place to start: • Reinforcement Learning from Human Feedback by Nathan Lambert: https://hubs.la/Q04rgQTH0 • Build a Large Language Model (From Scratch) by Sebastian Raschka, PhD: https://hubs.la/Q04rgKkf0 • AI Agents and Applications by Roberto Infante: https://hubs.la/Q04rgMks0 • Crack Any Codebase with AI by Zachary H.: https://hubs.la/Q04rgJLR0 They're all 50% off through tomorrow as part of our Summer Sale, along with the rest of the Manning catalogue. Also on sale: • 20% off Manning Online Pro annual subscriptions • 20% off Teams Manning Online subscriptions • $10 liveProjects and liveVideos

The ability to think statistically isn't just for data analysts anymore. It shapes sports debates, AI systems, software development, business strategy, and countless everyday decisions. On the latest Digital Social Hour with Sean Kelly, Gary Sutton explains why statistical thinking has become such a valuable skill and how the same principles show up across disciplines you might not expect. The conversation: https://hubs.la/Q04r9tGX0 If you'd like to go deeper, Gary's Manning books on statistics, algorithms, and AI are all included with a Manning Online Pro Annual subscription, now 20% off through Monday: https://hubs.la/Q04r8c230 (Individual book links are in the comments.)

Do platforms have an agenda? Are they favoring one type of content over another? And, who's in charge of such a decision? The answer lies in understanding recommender systems, the Hidden Influences behind our digital lives. As it turns out, it's more complicated than it seems, as it includes platform policy, human psychology and technical reasons. On August 12th at 12pm ET, Dr. Luca Belli will help us demystify the tools that have become an invisible part of our life. Register for the LinkedIn Live below ↓

Ever wonder what leads someone to build a company worth a reported $5 billion investment before it has even revealed what it's building? NVIDIA has reportedly invested $5 billion in Ilya Sutskever's Safe Superintelligence (SSI), along with access to its next-generation Vera Rubin systems. SSI inc remains largely secretive, but the investment says a lot about the confidence the industry has in Sutskever's vision. Reuters has the story: https://hubs.la/Q04r7VfD0 If you're curious how Sutskever approaches AI, Rich Heimann's Sutskever's List offers a useful perspective. The book explores the collection of research papers Sutskever once described as covering "90% of what matters" in AI, unpacking the breakthroughs, debates, and ideas that helped shape modern machine learning. It won't reveal SSI's roadmap. But it does provide context for understanding the thinking behind one of AI's most influential researchers. Now on Amazon: https://hubs.la/Q04r8pVr0

Ever wish your LLM app could improve without endlessly rewriting prompts? That's the idea behind DSPy (Community). Instead of manually crafting and tweaking prompt strings, you define your task in clean, modular Python code and let DSPy optimize the instructions and examples that produce the best results. It treats prompts as something to optimize rather than something to hand-tune. If you're curious how that approach works in practice, Building LLM Applications with DSPy by Serj Smorodinsky and Brett Kennedy walks through the concepts and shows you how to build with the framework. Alain AIROM (Ayrom) recently reviewed the book and noted that even in its MEAP stage, it's already a valuable resource for developers looking to get started: https://hubs.la/Q04qVQdl0 Learn more about the book: https://hubs.la/Q04qVwd10