English edition
Articles are written first in Portuguese, then editorially rewritten in English — not machine-translated. The full catalogue is in the Portuguese edition.
World 1 of AI
- World 1-1: where it all begins
AI is an assistant — junior or senior, depending on who runs it. The four possible uses, the method each one needs, and the question that comes before any tool: how will you check the result?
- World 1-2: underground
Why AI needs data, context and structure; what tokens are and why more context isn't better context; where hallucination comes from — and the method rules that came out of running this every day.
- World 1-3: up high
It's not about which tool to use. At each level — chat, chat with context, AI inside Office, an agent on your computer, automation with no interface — the AI does more without you watching. The right question is who checks, what can break and what a mistake costs.
- World 1-4: what holds the platforms up
What makes AI work in a company isn't the model, it's what you build around it: what it sees, where it looks things up, how it acts, how you know it got it right and what stops it from getting it wrong. And when RAG or fine-tuning is worth it.
- World 1-5: the castle
Big model or small, open or closed, train, fine-tune or distil, run it in the cloud or on your own computer. What each choice changes, what it costs, and why the model is the piece that decides the least.