Ollama for Beginners

Most local-LLM walkthroughs end the same way: a folder of half-downloaded models, a chat window the reader opened twice, and a quiet drift back to the paid API tab. The tools work. The setup is the problem.\n\nThis is the Ollama setup the other way. Ten chapters, one binary, three models that fit the reader's hardware, a working chat UI, a Modelfile or two for recurring jobs, and the small set of habits that turn a folder of open-weights model files into a tool the reader actually uses six months later.\n\nWhat's inside: the three reasons local LLMs are worth running now (privacy, cost, reliability) without the AGI evangelism, what Ollama is and what it isn't, the install path on macOS, Windows, and Linux, the parameter-count and quantization math that tells the reader in advance which models their hardware can actually run, the Ollama CLI commands worth memorizing, the HTTP API every integration uses, a real chat UI (Open WebUI) installed in twenty minutes, custom Modelfiles for the jobs the reader does every week, an honest comparison of when a local model is good enough and when it isn't, and the maintenance rhythm that keeps the setup useful over the long run.\n\nFor knowledge workers, hobbyist developers, IT generalists, privacy-conscious consultants, and curious tinkerers who have a recent laptop or desktop and are comfortable copy-pasting commands into a terminal. Not for machine-learning engineers, model trainers, or readers looking for a deep dive on transformer internals.\n\nLength: roughly 22,500 words. Approach: anti-hype, minimum-useful-version framing, no AGI evangelism, no promises about productivity — just the setup and the habits that hold it together.