Context Engineering Explained

… and 5 Practical Tips

The best way to get good outputs from an LLM is by giving it the right inputs. This is the goal of context engineering. In this article, I’ll explain what this is and share five tips for curating the context window.

Photo by Med Badr Chemmaoui on Unsplash

Large language models (LLMs) predict what comes next. In other words, given input text, they generate a (reasonable) next piece of text (over and over again).

However, it’s not enough that LLMs generate reasonable text. We want them to generate text that is helpful to us.

Here, I’ll break down how we can achieve this via context engineering. But first, what is it?

What is Context Engineering?

Renowned AI researcher (and accidental meme lord) Andrej Karpathy defined context engineering as filling an LLM's context window with just the right information for the next step [1].

This involves more than just giving LLMs clear instructions. It also includes curating examples, reference material, tool descriptions…