This is a real AI model. My WiFi is off.

The exact clicks

1. A "terminal" is just a plain window where you type text commands instead of clicking icons — every Mac and Windows computer already has one built in, nothing to download for this part. Open it — Mac: click the magnifying glass icon top-right of the screen (Spotlight), type "Terminal", press Enter; Windows: click the Start menu (bottom-left), type "PowerShell", press Enter. A plain black or white window opens — that's it, that's the terminal.
2. Copy this exact line for your system, paste it into that window, and press Enter. Mac/Linux: curl -fsSL https://ollama.com/install.sh | sh — Windows: powershell -c "irm https://ollama.com/install.ps1 | iex". This is the official installer from Ollama's own site (ollama.com) — safe to paste. You'll see several lines of text scroll by as it downloads and installs; when the lines stop and you see a new blank line waiting for input, it's finished — that can take a minute or two, depending on your internet.
3. Type ollama run llama3 and press Enter. The first time, this downloads the actual AI model (a few gigabytes — needs internet, and can take several minutes depending on your connection). You'll see a progress bar while it downloads. When it's done, the model starts up and the window shows ">>>" waiting for you to type something — that's it running, fully on this computer.
4. Optional, if you want to see the "offline" claim for yourself: turn off WiFi (or unplug the cable) now, then run ollama run llama3 again in a new terminal window — it still answers, because the model already lives on this machine and never needs the internet again.
5. Type any question directly after the ">>>" and press Enter — it answers right there in the same window.
6. To leave the chat and get your terminal back, type /bye and press Enter.
7. To try a smaller, faster model instead of llama3: type ollama run phi3 (smaller and faster, less capable) or ollama run gemma2 (a middle ground) — same steps as above, just that name instead of llama3.

What changes

• Ollama is free, open source, and actively maintained — confirmed via its own GitHub and release history, not assumed from an old memory of the tool.
• Local use needs no account and no API key at all — the model runs entirely on the machine once downloaded, with nothing to send anywhere.
• A normal laptop with no dedicated graphics card runs a 7B-class model (Llama 3, Gemma 2, Phi-3) at a usable speed — a GPU helps, but this doesn't require one.

What it will not do

• Ollama separately offers optional paid cloud tiers for cloud-hosted inference — this episode is about the free, local, offline path specifically, not Ollama's whole product line.
• A small local model won't match the biggest paid models on hard reasoning tasks — the claim here is a real, private, free, offline assistant, not a like-for-like replacement.
• The first ollama run of any model needs an internet connection to download it once — "offline" describes every run after that, not the very first one.