Claude Code 101: Demystifying Language Models

What is a token The context window How models generate text The attention mechanism How the model picks between options Model families What models can't do How much it costs Final thoughts 🇧🇷 Leia em português In the previous article , we built the entire factory: the evolution from manual production to autonomous machines, the ecosystem of agentic tools, the three pillars (prompt, context, and harness engineering). You know what the factory does, who works in it, and even how much revenue it pulls in. But the machines in the factory build things. And to understand how they build, the best analogy I know is LEGO. Standardized pieces that snap together one at a time, following (or ignoring) a manual, on a desk with limited space. This is the second article in the Claude Code 101 series, and here we take that mechanic apart. What tokens are, how the context window works, why models generate text the way they do, and why they sometimes get things wrong with unsettling confidence. What is