// learn
What you actually need
to know about AI.
Not the hype. Not the buzzwords. A clear, honest guide to understanding what's happening in AI — and how to get your own setup running.
01 / start here
Three concepts that actually matter.
What is a language model?
A language model is software trained on vast amounts of text to understand and generate human language. When you chat with ChatGPT or Claude, you're talking to a language model. They don't "think" — they predict the most useful next word, billions of times per second.
Local vs Cloud AI
Cloud AI (ChatGPT, Claude, Gemini) runs on someone else's servers. You send your data to them, pay per use, and have no control over what happens to it. Local AI runs entirely on your own hardware — private, fast, and free after the initial setup.
What is an AI agent?
An AI agent is a language model that can take actions — browse the web, run code, send emails, manage files. Instead of just answering questions, agents complete multi-step tasks autonomously. This is where AI gets genuinely powerful.
02 / the stack
What you need to run local AI
It's simpler than you think. Here's the full picture in 4 steps.
The hardware
A modern computer with enough RAM (16GB minimum, 32–48GB recommended). Apple Silicon (M1–M4) and NVIDIA RTX GPUs are the best options for local AI in 2026.
The runtime
Ollama is the easiest way to run local models. It handles downloading, managing, and running open-source AI models with a single command.
The models
Open-source models like Llama 4, Mistral, and Phi-3 are free to download and run. Different models are good at different things — we'll help you pick.
The agent layer
Tools like OpenClaw turn a local model into an agent that can take actions, use skills, and operate autonomously. This is the layer that makes it useful.
03 / model cheat sheet
Models worth knowing
Open-source models you can run today.
Llama 4 Scout
Meta
Best for: general tasks, long context
Meta's latest — strong all-rounder
Mistral Small 3.1
Mistral AI
Best for: tool use, coding, fast responses
Apache 2.0, best-in-class tool calling
Phi-3 Mini
Microsoft
Best for: edge devices, low-RAM setups
Punches way above its weight
Gemma 2 9B
Best for: balanced performance, everyday tasks
Google's best open-source model
Ready to get started?
We've done the hardware research so you don't have to.
Browse our compatibility-tested kits — hardware + setup script + model recommendations, all in one place.