How people, repositories, knowledge systems, evaluation frameworks and AI work together to improve engineering, research, creativity and decision-making.
View the Project on GitHub dermot-r-cochran/engineering-intelligence
Context engineering is the deliberate selection, organization, and maintenance of the information and resources available to an AI system for a task. Context can include instructions, repository files, examples, conversation history, tools, and constraints. Its quality affects what the system can notice and how well its output fits the task.
Start from the task and identify the information needed to complete it safely and correctly. Prefer relevant, authoritative, current sources over large volumes of loosely related material. Make priorities, constraints, and expected outputs explicit, and preserve links back to source material where possible.
Context is not a substitute for evaluation. Information may be incomplete, stale, or contradictory; generated outputs may still be wrong. Review important sources, expose uncertainty, and update context as the task or environment changes. Avoid including sensitive information unless it is necessary and authorized.
Context is a core part of repository-aware development and human-AI systems.
Where this is practised: see the evidence page.