Local AI Engineering with Ollama

Why This Book Exists

Most people will only ever rent intelligence. They type into a box, a bill runs in the background, and the model that answers them lives on hardware they will never see, owned by a company that can change the price, the rules, or the model itself whenever it wants. That is the default now. I think it is a bad default.

This is not a theoretical risk. In August 2025, OpenAI retired GPT-4o overnight and pushed everyone onto GPT-5. People who had built their daily work around a fast, predictable model woke up to one that was slower and behaved differently, with no way to go back. In June 2026, a US government export control directive forced Anthropic to cut off Fable 5 and Mythos 5 for every customer at once, citing national security. Teams that had started building on those models lost them in an afternoon, for reasons that had nothing to do with their own work. None of these people did anything wrong. They just did not own what they depended on, so the model, the price, and the rules stayed in someone else's hands, and any of them could move without warning.

This book exists because I got tired of that arrangement. I wanted to own the thing that runs: take a model apart, change how it behaves, feed it my own data, and break it without a meter running. The first time a model answered me from my own machine, with the network unplugged, something clicked. It was mine. It would still be mine next year. The vendor could not revoke my access, change the terms, or deprecate the model out from under me. The limit was my own: I can run any model I want, when I want, where I want, and most importantly, how I want.

What surprised me was how few people knew this was possible, and how much of what little they could find was either a marketing page or a research paper. Almost nothing existed for the person in the middle: the developer who can read code and run commands but has no machine learning degree and wants none. So I wrote the book I needed when I started.