What Actually Changed in Ollama
Ollama has quietly stopped being "just" a model runner. Recent releases added an interactive agent mode, a one-command way to launch a full coding agent, and a cloud proxy that can route some requests off your machine when you want it to. Worth knowing about, and worth being clear-eyed about what actually changes for you.
Running bare ollama now starts an agent
Type ollama with no arguments today and, on a current install, you land in an interactive agent that can chat, write code, search the web, and delegate tasks - not just a bare prompt waiting for a model name. Ollama's own release notes describe it plainly: running ollama now launches an agent to help you code and delegate work. It's a real capability shift from "a tool you point at a model" toward "a tool with opinions about what you're trying to do."
A new cloud proxy, opt-in
Recent releases also add cloud-routed models - the release notes reference commands like ollama launch pointing at cloud-hosted models alongside your local ones, using the same commands and the same local port. It's opt-in, not a default - but it's worth knowing the line between "local" and "cloud" now runs through a setting, not just through which software you installed.
Where this guide's approach still holds up
None of this changes what we teach in Setting Up a Mini PC for Local AI - pull a model explicitly, confirm it responds, then add a real interface with Open WebUI. That's still the better first path for understanding what's actually running on your machine, rather than starting from an agent that's already making decisions for you. Once you understand the pieces, the new agent mode is worth exploring - just not where we'd point someone on day one.