AI

This area covers building software on top of language models. The pages group into three clusters, and the clusters map to three widening scopes of control: the prompt, the single input you hand a model. The context window, the full set of information the model sees at inference. The agentic system, the autonomous loop that drives the model across many steps. Those three scopes are the generator. Every page in this area works one of them.

Each scope contains the one before it. A prompt sits inside a context window, and a context window sits inside the loop an agent runs. The vocabulary carries across those boundaries. A pattern you learn for prompts reappears when you assemble context, and again when an agent composes both.

flowchart TB
    subgraph sys["Agentic system"]
        subgraph ctx["Context window"]
            p["Prompt"]
        end
    end

Clusters

• Prompt Engineering: shaping the text input to a model so the output is more useful and more reliable
• Context Engineering: curating the full set of information a model sees at inference, across a finite token window and workloads that span many turns
• Agentic Development: building software where autonomous, model-driven tools act as collaborators rather than autocomplete. This cluster reaches past the systems pillar into the self pillar, because delegating to an agent reshapes how you work, not only what you build.

Prompt engineering works the innermost scope, context engineering the middle, and agentic development the outermost. Reading them in that order builds each idea on the one before it.