New to Artificial Intelligence? Read the Artificial Intelligence article
Most self-hosted LLM deployments run at 10–30% utilization — full retail price for GPUs that mostly sit idle. Tokens per Dollar is a deliberately short field guide for platform engineers: tune vLLM (2–4x throughput on the same hardware), quantize to halve your GPU count, cache the prompt tokens you keep re-paying for, and autoscale on real token traffic instead of CPU. Working YAML throughout, and a case study that cuts a 70B deployment from $29,200 to $8,600 a month with better latency. Read it in an evening; it pays for itself the first day you apply it.
What if your browser could run AI without sending your data to a server? AI in the Browser shows you how to build fast, private and fully local AI applications using WebGPU, WebAssembly and on-device LLMs. From GPU compute to a complete local chat interface, you’ll build everything yourself with practical code.
AI is changing how people discover information, products and businesses. The AEO Playbook shows you how to get found, understood, cited and recommended by AI. From practical tactics to technical implementation, it gives you the frameworks and tools to stay visible as search evolves.
AI can produce more work in less time. Making that work reliable takes discipline. Disciplined AI Development explores how to build workflows that combine AI capabilities with human judgment, clear boundaries, and rigorous verification.
What if AI could help you find flaws in software logic before they become costly bugs? This practical guide shows you how to use Claude Code alongside proven verification tools to test assumptions, uncover hidden issues and build confidence in your code. From first principles to production workflows, learn a smarter way to verify software with AI.
Abandona el ensayo y error en la IA generativa. Un marco práctico y metodológico basado en el falsacionismo de Karl Popper para auditar, estresar y dominar la interacción con modelos de lenguaje.
Build real machine learning applications with TensorFlow.js, from your first model to production-ready systems. Learn how to train and deploy models in the browser and Node.js, tackle real-world challenges and explore modern techniques with practical JavaScript and TypeScript examples you can run and adapt.
Generative AI can write code fast, but fast does not always mean safe. This practical guide shows software teams how to verify, test, secure and govern AI-assisted code across the full development lifecycle. Packed with real-world examples and practical workflows, it offers a clear path to adopting AI without lowering the bar for software quality.
Forecasting is changing fast. This practical guide takes you from ARIMA and exponential smoothing to Transformers, PatchTST and foundation models like Chronos and TimesFM. With clear explanations, hands-on Python examples and an honest look at what works and what fails, you’ll learn how to build forecasting systems that hold up in the real world.
Build production-ready applications with TypeScript in the age of AI. Go beyond the basics to master type-safe development, modern full-stack architectures, LLMs, intelligent agents, tool calling and the Model Context Protocol. With practical, runnable examples throughout, this book shows you how to build, test, secure and deploy serious TypeScript AI systems.
A hands-on guide for engineers and architects to build, optimize, and deploy scalable production-grade AI systems and LLM infrastructure.
My pitch was that rules written in CLAUDE.md get skipped, so you need enforcement. Then I measured it. In 22 trials with the ban written in, the banned command was never attempted. In 17 without it, 17 out of 17. This is the whole record: the instrument, every trial, the limits, and the harness so you can check me.
The second half of a from-scratch AI/ML course that refuses to skip the hard parts: LSTMs, real attention, a working transformer, LoRA fine-tuning derived and measured, and a genuine multi-tool, multi-agent system — every mechanism built by hand, then checked against PyTorch and real numbers.
A model that crashes is a good day. The dangerous failures return answers — plausible, fluent, and wrong. 100 failures. 63 diagnostic instruments. One rule: inspect what actually happened.
"I am not qualified to endorse theology. But I can witness: the structure is real. It was discovered, not invented."