The module starts with a guiding question: what can AI do that is useful for your specific practice, and what can't it?
Language models don't think or design. They predict, suggest, and execute with consistency and speed. Understanding how an LLM works, what context is, why they hallucinate, and when an image model produces what you expect — is what enables using them well. Without that framework, use oscillates between disappointment and uncritical dependency.
The second part builds the working environment: Python and dependency management, API keys, Claude Code as a scripting and automation environment, and MCP as the layer connecting language models with the architect's tools (files, design software, geospatial databases). Prompting is worked as a practical skill, developing instructions that produce useful results, persistent project context, and a personal knowledge base.